Automated brain extraction of multisequence MRI using artificial neural networks

IdenBAT: Disentangled representation learning for identity-preserved brain age transformation

Brain age transformation aims to convert reference brain images into synthesized images that accurately reflect the age-specific features of a target age group. The primary objective of this task is to modify only the age-related attributes of the reference image while preserving all other age-irrelevant attributes. However, achieving this goal poses substantial challenges due to the inherent entanglement of various image attributes within features extracted from a backbone encoder, resulting in simultaneous alterations during image generation. To address this challenge, we propose a novel architecture that employs disentangled representation learning for identity-preserved brain age transformation, called IdenBAT. This approach facilitates the decomposition of image features, ensuring the preservation of individual traits while selectively transforming age-related characteristics to match those of the target age group. Through comprehensive experiments conducted on both 2D and full-size 3D brain datasets, our method adeptly converts input images to target age while retaining individual characteristics accurately. Furthermore, our approach demonstrates superiority over existing state-of-the-art regarding performance fidelity. The code is available at: https://github.com/ku-milab/IdenBAT.

Automated neuroradiological support systems for multiple cerebrovascular disease markers - A systematic review and meta-analysis

Cerebrovascular diseases (CVD) can lead to stroke and dementia. Stroke is the second leading cause of death world wide and dementia incidence is increasing by the year. There are several markers of CVD that are visible on brain imaging, including: white matter hyperintensities (WMH), acute and chronic ischaemic stroke lesions (ISL), lacunes, enlarged perivascular spaces (PVS), acute and chronic haemorrhagic lesions, and cerebral microbleeds (CMB). Brain atrophy also occurs in CVD. These markers are important for patient management and intervention, since they indicate elevated risk of future stroke and dementia. We systematically reviewed automated systems designed to support radiologists reporting on these CVD imaging findings. We considered commercially available software and research publications which identify at least two CVD markers. In total, we included 29 commercial products and 13 research publications. Two distinct types of commercial support system were available: those which identify acute stroke lesions (haemorrhagic and ischaemic) from computed tomography (CT) scans, mainly for the purpose of patient triage; and those which measure WMH and atrophy regionally and longitudinally. In research, WMH and ISL were the markers most frequently analysed together, from magnetic resonance imaging (MRI) scans; lacunes and PVS were each targeted only twice and CMB only once. For stroke, commercially available systems largely support the emergency setting, whilst research systems consider also follow-up and routine scans. The systems to quantify WMH and atrophy are focused on neurodegenerative disease support, where these CVD markers are also of significance. There are currently no openly validated systems, commercially, or in research, performing a comprehensive joint analysis of all CVD markers (WMH, ISL, lacunes, PVS, haemorrhagic lesions, CMB, and atrophy).

AutoDPS: An unsupervised diffusion model based method for multiple degradation removal in MRI

BACKGROUND AND OBJECTIVE

Diffusion models have demonstrated their ability in image generation and solving inverse problems like restoration. Unlike most existing deep-learning based image restoration techniques which rely on unpaired or paired data for degradation awareness, diffusion models offer an unsupervised degradation independent alternative. This is well-suited in the context of restoring artifact-corrupted Magnetic Resonance Images (MRI), where it is impractical to exactly model the degradations apriori. In MRI, multiple corruptions arise, for instance, from patient movement compounded by undersampling artifacts from the acquisition settings.

METHODS

To tackle this scenario, we propose AutoDPS, an unsupervised method for corruption removal in brain MRI based on Diffusion Posterior Sampling. Our method (i) performs motion-related corruption parameter estimation using a blind iterative solver, and (ii) utilizes the knowledge of the undersampling pattern when the corruption consists of both motion and undersampling artifacts. We incorporate this corruption operation during sampling to guide the generation in recovering high-quality images.

RESULTS

Despite being trained to denoise and tested on completely unseen corruptions, our method AutoDPS has shown ∼ 1.63 dB of improvement in PSNR over baselines for realistic 3D motion restoration and ∼ 0.5 dB of improvement for random motion with undersampling. Additionally, our experiments demonstrate AutoDPS's resilience to noise and its generalization capability under domain shift, showcasing its robustness and adaptability.

CONCLUSION

In this paper, we propose an unsupervised method that removes multiple corruptions, mainly motion with undersampling, in MRI images which are essential for accurate diagnosis. The experiments show promising results on realistic and composite artifacts with higher improvement margins as compared to other methods. Our code is available at https://github.com/arunima101/AutoDPS/tree/master.

Noninvasive blood-brain barrier integrity mapping in patients with high-grade glioma and metastasis by multi-echo time-encoded arterial spin labeling

PURPOSE

In brain tumors, disruption of the blood-brain barrier (BBB) indicates malignancy. Clinical assessment is qualitative; quantitative evaluation is feasible using the K2 leakage parameter from dynamic susceptibility contrast MRI. However, contrast agent-based techniques are limited in patients with renal dysfunction and insensitive to subtle impairments. Assessing water transport times across the BBB (Tex) by multi-echo arterial spin labeling promises to detect BBB impairments noninvasively and potentially more sensitively. We hypothesized that reduced Tex indicates impaired BBB. Furthermore, we assumed higher sensitivity for Tex than dynamic susceptibility contrast-based K2, because arterial spin labeling uses water as a freely diffusible tracer.

METHODS

We acquired 3T MRI data from 28 patients with intraparenchymal brain tumors (World Health Organization Grade 3 & 4 gliomas [n = 17] or metastases [n = 11]) and 17 age-matched healthy controls. The protocol included multi-echo and single-echo Hadamard-encoded arterial spin labeling, dynamic susceptibility contrast, and conventional clinical imaging. Tex was calculated using a T2-dependent multi-compartment model. Areas of contrast-enhancing tissue, edema, and normal-appearing tissue were automatically segmented, and parameter values were compared across volumes of interest and between patients and healthy controls.

RESULTS

Tex was significantly reduced (-20.3%) in contrast-enhancing tissue compared with normal-appearing gray matter and correlated well with |K2| (r = -0.347). Compared with healthy controls, Tex was significantly lower in tumor patients' normal-appearing gray matter (Tex,tumor = 0.141 ± 0.032 s vs. Tex,HC = 0.172 ± 0.036 s) and normal-appearing white matter (Tex,tumor = 0.116 ± 0.015 vs. Tex,HC = 0.127 ± 0.017 s), whereas |K2| did not differ significantly. Receiver operating characteristic analysis showed a larger area under the curve for Tex (0.784) than K2 (0.604).

CONCLUSION

Tex is sensitive to pathophysiologically impaired BBB. It agrees with contrast agent-based K2 in contrast-enhancing tissue and indicates sensitivity to subtle leakage.

Mapping glioblastoma-induced neurological deficits: A brain atlas

BACKGROUND

Identifying radiological characteristics and brain regions associated with neurological deficits in glioblastoma patients can improve diagnostic evaluation and understanding of the disease's impact on neurological function.

METHODS

The retrospective study included 527 newly diagnosed glioblastoma patients. Eligibility criteria included pathologically confirmed IDH-wild type glioblastoma, availability of pre- and post-contrast MRIs, and detailed neurological examination reports. Contrast-enhancing tumors (CET) and non-contrast-enhancing lesions (NEL) were segmented from 3 Tesla MRI scans. Lesion volumes from patients without neurological deficits compared with symptomatic patients using either the Mann-Whitney test or Kruskal-Wallis test. Voxel-wise lesion-symptom mapping was conducted using Fisher-exact-test followed by random permutation analysis (ADIFFI) to identify brain regions with higher occurrences of deficit-associated lesions.

RESULTS

Location of CET and NEL within the brain were associated with specific neurological deficits. Larger CET and NEL volumes were associated with increased neurological deficits (CET: rs = 0.15, p = 0.0006; NEL: rs = 0.22, p < 0.0001). Lesion volumes were smaller in patients without neurological deficits (CET: 4.97 ± 0.69 ml vs. 20.0 ± 0.9 ml, p < 0.0001). Epilepsy-associated lesions were also smaller (CET: 4.59 ± 0.55 ml vs. 22.0 ± 0.9 ml, p < 0.0001).

CONCLUSION

The study highlights that neurological and epilepsy status at pre-treatment provide estimates of glioblastoma lesion volumes and locations. The correlation between lesion volumes and neurological deficits underscores the significance of comprehensive radiological assessments in glioblastoma patients. These findings support the use of detailed lesion-symptom mapping to guide clinical management and prognosis evaluation in glioblastoma.

Combined peritumoral radiomics and clinical features predict 12-month progression free survival in glioblastoma

PURPOSE

Analyzing post-treatment MRIs from glioblastoma patients can be challenging due to similar radiological presentations of disease progression and treatment effects. Identifying radiomics features (RFs) revealing progressive glioblastoma can contribute to an improved evaluation of the response assessment.

METHODS

3 Tesla MRI data from 560 glioblastoma patients (mean age 58.1 years) after treatment according to Stupp's protocol were analyzed retrospectively. A total of 418 RFs were extracted from contrast-enhancing tumors, non-enhancing lesions, peritumoral regions (PeriCET) and normal-appearing white matter as regions of interest using PyRadiomics. Dataset was initially split into a training (70%) and a validation (30%) cohort. The training cohort was used for feature selection and model-optimization. Logistic regression was used as a machine-learning model to identify patients with progression-free survival (PFS) as defined by the RANO criteria at 6 and 12 months after treatment. Models were trained with (i) clinical features only, (ii) RFs only, and (iii) a combination of clinical and radiomics features. The performance of each model was evaluated with the validation cohort.

RESULTS

The predictive performances of the model trained with only RFs from the PeriCET were AUC = 0.61 (95%-CI: 0.51-0.70) and AUC = 0.71 (95%-CI: 0.61-0.81) for 6-months and 12-months PFS respectively. Combining clinical and RFs from PeriCET resulted in overall best performance in predicting patients with progression within 12-months AUC = 0.75 (95%-CI: 0.65-0.85).

CONCLUSION

RFs from peritumoral region combined with clinical features including age, sex, and MGMT status can identify patients with 12-months PFS, suggesting the important role of peritumoral regions for the progression of glioblastoma.

Comparison of the Correlation Between Cerebral [18F]FDG Metabolism as Assessed by Two Asymmetry Indices and Clinical Neurological Score in Patients with Ischemic Cerebrovascular Disease

PURPOSE

Two types of asymmetry index (AI) have been utilized in evaluating cerebral function in ischemic cerebrovascular disease, however, few data exist on the differences between these AI measures. This study aimed to compare the two AIs in assessing PET cerebral metabolism and their correlation with clinical scales, to explore their potential value and applications in clinical settings.

PROCEDURES

Seventy patients diagnosed with subacute and chronic ischemic stroke were retrospectively analyzed. All patients underwent 2-deoxy- 2-[18F]fluoro-D-glucose ([18F]FDG) PET/MR scans and were assessed using the National Institutes of Health Stroke Scale (NIHSS) and the Modified Rankin Scale (mRS). Nineteen patients underwent a repeat [18F]FDG PET/MR scan one year later. Two voxel-wise AI methods, designated as AI1 and AI2, were calculated based on standardized uptake value ratio (SUVR). The hypometabolism on affected side assessed by different AI methods were compared. The correlations between the hypometabolism and the clinical scores were analyzed.

RESULTS

The volume and percentage of decreased [18F]FDG metabolism assessed by AI2 was larger than that obtained from AI1 (all p < 0.0001). The correlation coefficients between the clinical scores and the decreased metabolism in temporal and parietal lobes assessed by AI1 method were all higher than those from AI2. In addition, the improved follow-up patients showed more pronounced metabolic improvement as assessed by AI1.

CONCLUSIONS

The assessment of cerebral [18F]FDG metabolism in patients with unilateral internal carotid/middle cerebral artery steno-occlusion to reflect clinical neurological function using the AI1 method demonstrated superior performance in comparison to the AI2 method.

Imaging-Based Molecular Characterization of Adult-Type Diffuse Glioma Using Diffusion and Perfusion MRI in Pre- and Post-Treatment Stage Considering Spatial and Temporal Heterogeneity

BACKGROUND

Imaging-based molecular characterization is important for identifying treatment targets in adult-type diffuse gliomas.

PURPOSE

To assess isocitrate dehydrogenase (IDH) mutation and epidermal growth factor receptor (EGFR) amplification status in primary and recurrent gliomas using diffusion and perfusion MRI, addressing spatial and temporal heterogeneity.

STUDY TYPE

Retrospective.

SUBJECTS

Three-hundred and twelve newly diagnosed (cross-sectional set, 57.9 ± 13.2 years, 52.2% male, 235 IDH-wildtype, 71 EGFR-amplified) and 38 recurrent (longitudinal set, 53.1 ± 13.4 years, 44.7% male, 30 IDH-wildtype, 13 EGFR-amplified) adult-type diffuse glioma patients.

FIELD STRENGTH/SEQUENCE

3.0T; diffusion weighted and dynamic susceptibility contrast-perfusion weighted imaging.

ASSESSMENT

Radiomics features from contrast-enhancing tumors (CET) and non-enhancing lesions (NEL) were extracted from apparent diffusion coefficient and perfusion maps. Spatial heterogeneity was assessed using intersection and Bhattacharyya distance between CET and NEL. Stable imaging features were identified in patients with unchanged genetic profiles in the longitudinal set. The "best model," using features from the cross-sectional set (n = 312), and the "concordant model," using stable features identified in the longitudinal set (n = 38), were constructed using the LASSO for IDH and EGFR status.

STATISTICAL TESTS

The area under the receiver-operating-characteristic curve (AUC).

RESULTS

For IDH mutations, both best and concordant models demonstrated high AUCs in the cross-sectional set (0.936; 95% confidence interval [CI]: 0.903-0.969 and 0.964 [0.943-0.986], respectively). Only the concordant model maintained strong performance in recurrent tumors (AUC, 0.919 vs. 0.656). For EGFR amplification in IDH-wildtype, the best and concordant models showed AUCs of 0.821 (95% CI: 0.761-0.881) and 0.746 (95% CI: 0.675-0.817) in newly diagnosed gliomas, but poor performance in recurrent tumors with AUCs of 0.503 (95% CI: 0.34-0.665) and 0.518 (95% CI: 0.357-0.678).

DATA CONCLUSION

Diffusion and perfusion MRI characterized IDH status in both newly diagnosed and recurrent gliomas, but showed limited diagnostic performance for EGFR, especially for recurrent tumors.

EVIDENCE LEVEL

3 TECHNICAL EFFICACY: Stage 3.

Exercise modulates brain pulsatility: insights from q-aMRI and MRI-based flow methods

This study investigates intracranial dynamics following the Monro-Kellie doctrine, depicting how brain pulsatility, cerebrospinal fluid (CSF) flow and cerebral blood flow (CBF) interact under resting and exercise conditions. Using quantitative amplified magnetic resonance imaging (q-aMRI) alongside traditional MRI flow metrics, we measured and analysed blood flow, CSF dynamics and brain displacement in a cohort of healthy adults both at rest and during low-intensity handgrip exercise. Exercise was found to reduce pulsatility in CBF while increasing CSF flow and eliminating CSF regurgitation, highlighting a shift towards more sustained forward flow patterns (from cranial to spinal compartments). Displacement analysis using q-aMRI revealed a consistent trend of reduced whole brain motion during exercise, though as the sample of data that met quality control was low (n = 5), this was not a significant result. There was an observable decrease in the motion of third and fourth ventricles, linking ventricular displacement to CSF flow alterations. These findings suggest that exercise may not only affect the rate and directionality of CSF flow but also modulate brain tissue motion, supporting cerebral homeostasis. This study offers insights into how the brain adapts dynamically under varying conditions, with implications for understanding intracranial pressure regulation in humans and diagnostic contexts.

An External, Independent Validation of an O-(2-[18F]Fluoroethyl)-l-Tyrosine PET Automatic Segmentation Network on a Single-Center, Prospective Dataset of Patients with Glioblastoma

The goal of this study was to conduct an external, independent validation of an O-(2-[18F]fluoroethyl)-l-tyrosine ([18F]FET) PET automatic segmentation network on a cohort of patients with glioblastoma. Methods: Twenty-four patients with glioblastoma were included in this study who underwent a total of 52 [18F]FET PET scans (preradiotherapy, n = 23; preradiotherapy retest, n = 9; follow-up, n = 20). Biologic tumor volume (BTV) delineation was performed by an expert nuclear medicine physician and an automatic segmentation network. Physician and automated quantitative metrics (BTV, mean tumor-to-background ratio [TBRmean], lesion SUVmean, and background SUVmean) were assessed with Pearson correlation and Bland-Altman analysis (bias, limits of agreement [LoA]). Automated and physician segmentation overlap was assessed with spatial and distance-based metrics. Results: BTV and TBRmean Pearson correlation was excellent for all time points (range, 0.92-0.98). In 2 patients with frontal lobe lesions, the network segmented the transverse sinus. Bland-Altman analysis showed network underestimation of physician-derived BTVs (absolute bias, 2.7 cm3, LoA, -13.1-18.5 cm3; relative bias, 27.9%, LoA, -95.3%-151.2%) and deviations for TBRmean were small (absolute bias, 0.03, LoA, -0.25-0.30; relative bias, 0.83%, LoA -14.27%-15.93%). Median Dice similarity coefficient, surface Dice similarity coefficient, Hausdorff distance, 95th percentile Hausdorff distance, and mean absolute surface distance were 0.83, 0.95, 10.94 mm, 3.62 mm, and 0.88 mm, respectively. Conclusion: Automated quantitative analysis was highly correlated with physician assessment; however, volume underestimation and erroneous segmentations may impact radiotherapy treatment planning and response assessment. Further training on a representative local dataset would likely be required for multicenter implementation.

A magnetic resonance imaging (MRI)-based deep learning radiomics model predicts recurrence-free survival in lung cancer patients after surgical resection of brain metastases

AIM

To develop and validate a magnetic resonance imaging (MRI)-based deep learning radiomics model (DLRM) to predict recurrence-free survival (RFS) in lung cancer patients after surgical resection of brain metastases (BrMs).

MATERIALS AND METHODS

A total of 215 lung cancer patients with BrMs confirmed by surgical pathology were retrospectively included in five centres, 167 patients were assigned to the training cohort, and 48 to the external test cohort. All patients underwent regular follow-up brain MRIs. Clinical and morphological MRI models for predicting RFS were built using univariate and multivariate Cox regressions, respectively. Handcrafted and deep learning (DL) signatures were constructed from BrMs pretreatment MR images using the least absolute shrinkage and selection operator (LASSO) method, respectively. A DLRM was established by integrating the clinical and morphological MRI predictors, handcrafted and DL signatures based on the multivariate Cox regression coefficients. The Harrell C-index, area under the receiver operating characteristic curve (AUC), and Kaplan-Meier's survival analysis were used to evaluate model performance.

RESULTS

The DLRM showed satisfactory performance in predicting RFS and 6- to 18-month intracranial recurrence in lung cancer patients after BrMs resection, achieving a C-index of 0.79 and AUCs of 0.84-0.90 in the training set and a C-index of 0.74 and AUCs of 0.71-0.85 in the external test set. The DLRM outperformed the clinical model, morphological MRI model, handcrafted signature, DL signature, and clinical-morphological MRI model in predicting RFS (P < 0.05). The DLRM successfully classified patients into high-risk and low-risk intracranial recurrence groups (P < 0.001).

CONCLUSION

This MRI-based DLRM could predict RFS in lung cancer patients after surgical resection of BrMs.

Comprehensive Segmentation of Gray Matter Structures on T1-Weighted Brain MRI: A Comparative Study of Convolutional Neural Network, Convolutional Neural Network Hybrid-Transformer or -Mamba Architectures

BACKGROUND AND PURPOSE

Recent advances in deep learning have shown promising results in medical image analysis and segmentation. However, most brain MRI segmentation models are limited by the size of their data sets and/or the number of structures they can identify. This study evaluates the performance of 6 advanced deep learning models in segmenting 122 brain structures from T1-weighted MRI scans, aiming to identify the most effective model for clinical and research applications.

MATERIALS AND METHODS

A total of 1510 T1-weighted MRIs were used to compare 6 deep learning models for the segmentation of 122 distinct gray matter structures: nnU-Net, SegResNet, SwinUNETR, UNETR, U-Mamba_BOT, and U-Mamba_ Enc. Each model was rigorously tested for accuracy by using the dice similarity coefficient (DSC) and the 95th percentile Hausdorff distance (HD95). Additionally, the volume of each structure was calculated and compared between normal controls (NCs) and patients with Alzheimer disease (AD).

RESULTS

U-Mamba_Bot achieved the highest performance with a median DSC of 0.9112 (interquartile range [IQR]: 0.8957, 0.9250). nnU-Net achieved a median DSC of 0.9027 [IQR: 0.8847, 0.9205], and had the highest HD95 of 1.392 [IQR: 1.174, 2.029]. The value of each HD95 (<3 mm) indicates its superior capability in capturing detailed brain structures accurately. Following segmentation, volume calculations were performed, and the resultant volumes of NCs and patients with AD were compared. The volume changes observed in 13 brain substructures were all consistent with those reported in existing literature, reinforcing the reliability of the segmentation outputs.

CONCLUSIONS

This study underscores the efficacy of U-Mamba_Bot as a robust tool for detailed brain structure segmentation in T1-weighted MRI scans. The congruence of our volumetric analysis with the literature further validates the potential of advanced deep learning models to enhance the understanding of neurodegenerative diseases such as AD. Future research should consider larger data sets to validate these findings further and explore the applicability of these models in other neurologic conditions.

A quantitatively interpretable model for Alzheimer's disease prediction using deep counterfactuals

Deep learning (DL) for predicting Alzheimer's disease (AD) has provided timely intervention in disease progression yet still demands attentive interpretability to explain how their DL models make definitive decisions. Counterfactual reasoning has recently gained increasing attention in medical research because of its ability to provide a refined visual explanatory map. However, such visual explanatory maps based on visual inspection alone are insufficient unless we intuitively demonstrate their medical or neuroscientific validity via quantitative features. In this study, we synthesize the counterfactual-labeled structural MRIs using our proposed framework and transform it into a gray matter density map to measure its volumetric changes over the parcellated region of interest (ROI). We also devised a lightweight linear classifier to boost the effectiveness of constructed ROIs, promoted quantitative interpretation, and achieved comparable predictive performance to DL methods. Throughout this, our framework produces an "AD-relatedness index" for each ROI. It offers an intuitive understanding of brain status for an individual patient and across patient groups concerning AD progression.

"Chocolate Chip Sign" on Susceptibility-Weighted Imaging: A Novel Neuroimaging Biomarker for HTRA1-Related Cerebral Small Vessel Disease

Background and Objectives

HTRA1-related cerebral small vessel disease (HRSVD) is a rare hereditary form of cerebral small vessel disease (CSVD) caused by HTRA1 pathogenic variants. Diagnosing HRSVD without genetic testing is challenging because of the lack of distinctive imaging features and clinical symptoms, and even family history can be unclear in some cases with HRSVD. This study investigates whether susceptibility-weighted imaging (SWI) can identify useful diagnostic findings for HRSVD.

Methods

This retrospective study included 8 patients with HRSVD, 12 with cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL), and 14 with sporadic CSVD (sCSVD). Two neurologists blinded to clinical data counted the number of hypointense dots around the midbrain on SWI. Receiver operating characteristic curve analysis evaluated the optimal threshold of the number that can distinguish HRSVD and CADASIL or sCSVD. In addition, histopathologic analysis including measurement of leptomeningeal vessel diameter and type III collagen deposition was performed on autopsied brains from 3 cases each of HRSVD, CADASIL, and sCSVD and control participants.

Results

Patients with HRSVD exhibited a significantly higher number of hypointense dots around the midbrain on SWI compared with CADASIL and sCSVD groups. A threshold of 5 or more dots, termed the "Chocolate Chip Sign," well distinguished HRSVD from CADASIL and sCSVD (area under the curve: 0.817, 95% confidence interval: 0.624-1.00). Three-dimensional SWI reconstruction and 7T MRI confirmed these dots as dilated extraparenchymal vessels. Histopathologic analysis revealed pronounced dilation of leptomeningeal veins with type III collagen accumulation specifically, in HRSVD brains.

Discussion

The Chocolate Chip Sign on SWI represents a novel and promising neuroimaging biomarker for HRSVD. This finding holds significant potential for facilitating early diagnosis, prompting timely genetic testing, and appropriate family screening for this rare genetic disorder.

Coupling of state space modules and attention mechanisms: An input-aware multi-contrast MRI synthesis method

BACKGROUND

Medical imaging plays a pivotal role in the real-time monitoring of patients during the diagnostic and therapeutic processes. However, in clinical scenarios, the acquisition of multi-modal imaging protocols is often impeded by a number of factors, including time and economic costs, the cooperation willingness of patients, imaging quality, and even safety concerns.

PURPOSE

We proposed a learning-based medical image synthesis method to simplify the acquisition of multi-contrast MRI.

METHODS

We redesigned the basic structure of the Mamba block and explored different integration patterns between Mamba layers and Transformer layers to make it more suitable for medical image synthesis tasks. Experiments were conducted on the IXI (a total of 575 samples, training set: 450 samples; validation set: 25 samples; test set: 100 samples) and BRATS (a total of 494 samples, training set: 350 samples; validation set: 44 samples; test set: 100 samples) datasets to assess the synthesis performance of our proposed method in comparison to some state-of-the-art models on the task of multi-contrast MRI synthesis.

RESULTS

Our proposed model outperformed other state-of-the-art models in some multi-contrast MRI synthesis tasks. In the synthesis task from T1 to PD, our proposed method achieved the peak signal-to-noise ratio (PSNR) of 33.70 dB (95% CI, 33.61, 33.79) and the structural similarity index (SSIM) of 0.966 (95% CI, 0.964, 0.968). In the synthesis task from T2 to PD, the model achieved a PSNR of 33.90 dB (95% CI, 33.82, 33.98) and SSMI of 0.971 (95% CI, 0.969, 0.973). In the synthesis task from FLAIR to T2, the model achieved PSNR of 30.43 dB (95% CI, 30.29, 30.57) and SSIM of 0.938 (95% CI, 0.935, 0.941).

CONCLUSIONS

Our proposed method could effectively model not only the high-dimensional, nonlinear mapping relationships between the magnetic signals of the hydrogen nucleus in tissues and the proton density signals in tissues, but also of the recovery process of suppressed liquid signals in FLAIR. The model proposed in our work employed distinct mechanisms in the synthesis of images belonging to normal and lesion samples, which demonstrated that our model had a profound comprehension of the input data. We also proved that in a hierarchical network, only the deeper self-attention layers were responsible for directing more attention on lesion areas.

Automated Quantification of Cerebral Microbleeds in SWI: Association with Vascular Risk Factors, White Matter Hyperintensity Burden, and Cognitive Function

BACKGROUND AND PURPOSE

The amount and distribution of cerebral microbleeds (CMB) are important risk factors for cognitive impairment. Our objective was to train and validate a deep learning (DL)-based segmentation model for cerebral microbleeds (CMBs) on SWI and to find associations among CMB, cognitive impairment, and vascular risk factors.

MATERIALS AND METHODS

Participants in this single-institution retrospective study underwent brain MRI to evaluate cognitive impairment between January and September 2023. For training the DL model, the nnU-Net framework was used without modifications. The performance of the DL model was evaluated on independent internal and external validation data sets. Linear regression analysis was used to find associations among log-transformed CMB numbers, cognitive function (Mini-Mental Status Examination [MMSE]), white matter hyperintensity (WMH) burden, and clinical vascular risk factors (age, sex, hypertension, diabetes, lipid profiles, and body mass index).

RESULTS

Training of the DL model (n = 287) resulted in a robust segmentation performance with an average Dice score of 0.73 (95% CI, 0.67-0.79) in an internal validation set (n = 67) and modest performance in an external validation set (Dice score = 0.46; 95% CI, 0.33-0.59; n = 68). In a temporally independent clinical data set (n = 448), older age, hypertension, and WMH burden were significantly associated with CMB numbers in all distributions (total, lobar, deep, and cerebellar; all P < . 01). The MMSE was significantly associated with hyperlipidemia (β = 1.88; 95% CI, 0.96-2.81; P < . 001), WMH burden (β = -0.17 per 1% WMH burden, 95% CI, -0.27-0.08; P < . 001), and total CMB number (β = -0.01 per 1 CMB, 95% CI, -0.02-0.001; P = .04) after adjusting for age and sex.

CONCLUSIONS

The DL model showed a robust segmentation performance for CMB. In all distributions, CMB had significant positive associations with WMH burden. Increased WMH burden and CMB numbers were associated with decreased cognitive function.

Magnetization transfer imaging using non-balanced SSFP at ultra-low field

PURPOSE

Ultra-low field MRI scanners have the potential to improve health care delivery, both through improved access in areas where there are few MRI scanners and allowing more frequent monitoring of disease progression and treatment response. This may be particularly true in white matter disorders, including leukodystrophies and multiple sclerosis, in which frequent myelin-sensitive imaging, such as magnetization transfer (MT) imaging, might improve clinical care and patient outcomes.

METHODS

We implemented an on-resonance approach to MT imaging on a commercial point-of-care 64 mT scanner using a non-balanced steady-state free precession sequence. Phantom and in vivo experiments were used to evaluate and optimize the sequence sensitivity and reproducibility, and to demonstrate in vivo performance and inter-site reproducibility.

RESULTS

From phantom experiments, T1 and T2 effects were determined to have a negligible effect on the differential MT weighting. MT ratio (MTR) values in white matter were 23.1 ± 1.0% from 10 healthy volunteers, with an average reproducibility coefficient of variation of 1.04%. Normal-appearing white matter MTR values in a multiple sclerosis participant (21.5 ± 6.2%) were lower, but with a similar spread of values, compared to an age-matched healthy volunteer (23.3 ± 6.2%).

CONCLUSION

An on-resonance MT imaging approach was developed at 64 mT that can be performed in as little as 4 min. A semi-quantitative myelin-sensitive imaging biomarker at this field strength is available for assessing both myelination and demyelination.

MRI distortion correction is associated with improved local control in stereotactic radiotherapy for brain metastases

Distortions in brain MRI caused by gradient nonlinearities may reach several millimeters, thus distortion correction is strongly recommended for radiotherapy treatment planning. However, the significance of MRI distortion correction on actual clinical outcomes has not been described yet. Therefore, we investigated the impact of planning MRI distortion correction on subsequent local control in a historic series of 419 brain metastases in 189 patients treated with stereotactic radiotherapy between 01/2003 and 04/2015. Local control was evaluated using a volumetric extension of the RANO-BM criteria. The predictive significance of distortion correction was assessed using competing risk analysis. In this cohort, 2D distortion-corrected MRIs had been used for treatment planning in 52.5% (220/419) of lesions, while uncorrected MRIs had been employed in 47.5% (199/419) of metastases. 2D distortion correction was associated with improved local control (Cumulative incidence of local progression at 12 months: 14.3% vs. 21.2% and at 24 months: 18.7% vs. 28.6%, p = 0.038). In multivariate analysis, adjusting for histology, baseline tumor volume, interval between MRI and treatment delivery, year of planning MRI, biologically effective dose and adjuvant Whole-brain radiotherapy, use of distortion correction remained significantly associated with improved local control (HR 0.55, p = 0.020). This is the first study to clinically evaluate the impact of MRI gradient nonlinearity distortion correction on local control in stereotactic radiotherapy for brain metastases. In this historic series, we found significantly higher local control when using 2D corrected vs. uncorrected MRI studies for treatment planning. These results stress the importance of assuring that MR images used for radiotherapy treatment planning are properly distortion-corrected.

Interpretable multimodal transformer for prediction of molecular subtypes and grades in adult-type diffuse gliomas

Molecular subtyping and grading of adult-type diffuse gliomas are essential for treatment decisions and patient prognosis. We introduce GlioMT, an interpretable multimodal transformer that integrates imaging and clinical data to predict the molecular subtype and grade of adult-type diffuse gliomas according to the 2021 WHO classification. GlioMT is trained on multiparametric MRI data from an institutional set of 1053 patients with adult-type diffuse gliomas to predict the IDH mutation status, 1p/19q codeletion status, and tumor grade. External validation on the TCGA (200 patients) and UCSF (477 patients) shows that GlioMT outperforms conventional CNNs and visual transformers, achieving AUCs of 0.915 (TCGA) and 0.981 (UCSF) for IDH mutation, 0.854 (TCGA) and 0.806 (UCSF) for 1p/19q codeletion, and 0.862 (TCGA) and 0.960 (UCSF) for grade prediction. GlioMT enhances the reliability of clinical decision-making by offering interpretability through attention maps and contributions of imaging and clinical data.

Quantification of NAD+ T1 and T2 Relaxation Times Using Downfield 1H MRS at 7 T in Human Brain In Vivo

The purpose of this study was to measure T1 and T2 relaxation times of NAD+ proton resonances in the downfield 1H MRS spectrum in human brain at 7 T in vivo and to assess the propagation of relaxation time uncertainty in NAD+ quantification. Downfield spectra from eight healthy volunteers were acquired at multiple echo times to measure T2 relaxation times, and saturation recovery data were acquired to measure T1 relaxation times. The downfield acquisition used a spectrally selective 90° sinc pulse for excitation centered at 9.1 ppm with a bandwidth of 2 ppm, followed by a 180° spatially selective Shinnar-Le Roux refocusing pulse for localization. Uncertainty propagation analysis on metabolite quantification was performed analytically and with Monte Carlo simulation. [NAD+] was quantified in five participants. The mean ± standard deviation of T1 relaxation times of the H2, H6, and H4 NAD+ protons were 205.6 ± 25.7, 211.6 ± 33.5, and 237.3 ± 42.4 ms, respectively. The mean ± standard deviation of T2 relaxation times of the H2, H6, and H4 protons were 33.6 ± 7.4, 29.1 ± 4.7, and 42.3 ± 11.6 ms, respectively. The relative uncertainty in NAD+ concentration due to relaxation time uncertainty was 8.4%-11.4%, and measured brain [NAD+] (N = 5) was 0.324 ± 0.050 mM. Using downfield spectrally selective spectroscopy with single-slice localization, we found T1 and T2 relaxation times averaged across all NAD+ resonances to be approximately 218 and 35 ms, respectively, in the human brain in vivo at 7 T.

Shape matters: unsupervised exploration of IDH-wildtype glioma imaging survival predictors

OBJECTIVES

This study examines clustering based on shape radiomic features and tumor volume to identify IDH-wildtype glioma phenotypes and assess their impact on overall survival (OS).

MATERIALS AND METHODS

This retrospective study included 436 consecutive patients diagnosed with IDH-wt glioma who underwent preoperative MR imaging. Alongside the total tumor volume, nine distinct shape radiomic features were extracted using the PyRadiomics framework. Different imaging phenotypes were identified using partition around medoids (PAM) clustering on the training dataset (348/436). The prognostic efficacy of these phenotypes in predicting OS was evaluated on the test dataset (88/436). External validation was performed using the public UCSF glioma dataset (n = 397). A decision-tree algorithm was employed to determine the relevance of features associated with cluster affiliation.

RESULTS

PAM clustering identified two clusters in the training dataset: Cluster 1 (n = 233) had a higher proportion of patients with higher sphericity and elongation, while Cluster 2 (n = 115) had a higher proportion of patients with higher maximum 3D diameter, surface area, axis lengths, and tumor volume (p < 0.001 for each). OS differed significantly between clusters: Cluster 1 showed a median OS of 23.8 compared to 11.4 months of Cluster 2 in the holdout test dataset (p = 0.002). Multivariate Cox regression showed improved performance with cluster affiliation over clinical data alone (C index 0.67 vs 0.59, p = 0.003). Cluster-based models outperformed the models with tumor volume alone (evidence ratio: 5.16-5.37).

CONCLUSION

Data-driven clustering reveals imaging phenotypes, highlighting the improved prognostic power of combining shape-radiomics with tumor volume, thereby outperforming predictions based on tumor volume alone in high-grade glioma survival outcomes.

CLINICAL RELEVANCE STATEMENT

Shape-radiomics and volume-based cluster analyses of preoperative MRI scans can reveal imaging phenotypes that improve the prediction of OS in patients with IDH-wild type gliomas, outperforming currently known models based on tumor size alone or clinical parameters.

KEY POINTS

Shape radiomics and tumor volume clustering in IDH-wildtype gliomas are investigated for enhanced prognostic accuracy. Two distinct phenotypic clusters were identified with different median OSs. Integrating shape radiomics and volume-based clustering enhances OS prediction in IDH-wildtype glioma patients.

Characterizing the Effect of Repetitive Head Impact Exposure and mTBI on Adolescent Collision Sports Players' Brain with Diffusion Magnetic Resonance Imaging

Athletes in collision sports frequently sustain repetitive head impacts (RHI), which, while not individually severe enough for a clinical mild traumatic brain injury (mTBI) diagnosis, can compromise neuronal organization by transferring mechanical energy to the brain. Although numerous studies target athletes with mTBI, there is a lack of longitudinal research on young collision sport participants, highlighting an unaddressed concern regarding cumulative RHI effects on brain microstructures. Therefore, this study aimed to investigate the microstructural changes in the brains' of high school rugby players due to repeated head impacts and to establish a correlation between clinical symptoms, cumulative effects of RHI exposure, and changes in the brain's microstructure. We conducted a longitudinal magnetic resonance imaging (MRI) study on 36 male high school rugby players across a season using 3D T1-weighted and multi-shell diffusion MRI sequences, comparing them with 20 matched controls. Players with concussions were separately tracked up to 6 weeks post-injury with three-times scans within this period. The Sport Concussion Assessment Tool (SCAT5) symptom scale assessed mTBI symptoms, and mouthguard-embedded kinematic sensors recorded head impacts. No significant volumetric changes in subcortical structures were found post-rugby season. However, there were substantial differences in mean diffusivity (MD) and axial diffusivity (AD) between the rugby players and controls across widespread brain regions. Diffusion metrics, especially AD, MD, and radial diffusivity of certain brain tracts, displayed strong correlations with SCAT5 symptom severity. Repeated head impacts during a rugby season may adversely affect the structural organization of the brain's white matter. The observed diffusion changes, closely tied to SCAT5 symptom burden, stress the profound effects of seasonal head impacts and highlight individual variability in response to repetitive head impact exposure. To better manage sports-related mTBI and guide return-to-play decisions, comprehensive studies on brain injury mechanisms and recovery post-mTBI/RHI exposure are required.

Principles of artificial intelligence in radiooncology

PURPOSE

In the rapidly expanding field of artificial intelligence (AI) there is a wealth of literature detailing the myriad applications of AI, particularly in the realm of deep learning. However, a review that elucidates the technical principles of deep learning as relevant to radiation oncology in an easily understandable manner is still notably lacking. This paper aims to fill this gap by providing a comprehensive guide to the principles of deep learning that is specifically tailored toward radiation oncology.

METHODS

In light of the extensive variety of AI methodologies, this review selectively concentrates on the specific domain of deep learning. It emphasizes the principal categories of deep learning models and delineates the methodologies for training these models effectively.

RESULTS

This review initially delineates the distinctions between AI and deep learning as well as between supervised and unsupervised learning. Subsequently, it elucidates the fundamental principles of major deep learning models, encompassing multilayer perceptrons (MLPs), convolutional neural networks (CNNs), recurrent neural networks (RNNs), transformers, generative adversarial networks (GANs), diffusion-based generative models, and reinforcement learning. For each category, it presents representative networks alongside their specific applications in radiation oncology. Moreover, the review outlines critical factors essential for training deep learning models, such as data preprocessing, loss functions, optimizers, and other pivotal training parameters including learning rate and batch size.

CONCLUSION

This review provides a comprehensive overview of deep learning principles tailored toward radiation oncology. It aims to enhance the understanding of AI-based research and software applications, thereby bridging the gap between complex technological concepts and clinical practice in radiation oncology.

Guided reconstruction with conditioned diffusion models for unsupervised anomaly detection in brain MRIs

The application of supervised models to clinical screening tasks is challenging due to the need for annotated data for each considered pathology. Unsupervised Anomaly Detection (UAD) is an alternative approach that aims to identify any anomaly as an outlier from a healthy training distribution. A prevalent strategy for UAD in brain MRI involves using generative models to learn the reconstruction of healthy brain anatomy for a given input image. As these models should fail to reconstruct unhealthy structures, the reconstruction errors indicate anomalies. However, a significant challenge is to balance the accurate reconstruction of healthy anatomy and the undesired replication of abnormal structures. While diffusion models have shown promising results with detailed and accurate reconstructions, they face challenges in preserving intensity characteristics, resulting in false positives. We propose conditioning the denoising process of diffusion models with additional information derived from a latent representation of the input image. We demonstrate that this conditioning allows for accurate and local adaptation to the general input intensity distribution while avoiding the replication of unhealthy structures. We compare the novel approach to different state-of-the-art methods and for different data sets. Our results show substantial improvements in the segmentation performance, with the Dice score improved by 11.9%, 20.0%, and 44.6%, for the BraTS, ATLAS and MSLUB data sets, respectively, while maintaining competitive performance on the WMH data set. Furthermore, our results indicate effective domain adaptation across different MRI acquisitions and simulated contrasts, an important attribute for general anomaly detection methods. The code for our work is available at https://github.com/FinnBehrendt/Conditioned-Diffusion-Models-UAD.

WAND: A multi-modal dataset integrating advanced MRI, MEG, and TMS for multi-scale brain analysis

This paper introduces the Welsh Advanced Neuroimaging Database (WAND), a multi-scale, multi-modal imaging dataset comprising in vivo brain data from 170 healthy volunteers (aged 18-63 years), including 3 Tesla (3 T) magnetic resonance imaging (MRI) with ultra-strong (300 mT/m) magnetic field gradients, structural and functional MRI and nuclear magnetic resonance spectroscopy at 3 T and 7 T, magnetoencephalography (MEG), and transcranial magnetic stimulation (TMS), together with trait questionnaire and cognitive data. Data are organised using the Brain Imaging Data Structure (BIDS). In addition to raw data, we provide brain-extracted T1-weighted images, and quality reports for diffusion, T1- and T2-weighted structural data, and blood-oxygen level dependent functional tasks. Reasons for participant exclusion are also included. Data are available for download through our GIN repository, a data access management system designed to reduce storage requirements. Users can interact with and retrieve data as needed, without downloading the complete dataset. Given the depth of neuroimaging phenotyping, leveraging ultra-high-gradient, high-field MRI, MEG and TMS, this dataset will facilitate multi-scale and multi-modal investigations of the healthy human brain.

WDRIV-Net: a weighted ensemble transfer learning to improve automatic type stratification of lumbar intervertebral disc bulge, prolapse, and herniation

The degeneration of the intervertebral discs in the lumbar spine is the common cause of neurological and physical dysfunctions and chronic disability of patients, which can be stratified into single-(e.g., disc herniation, prolapse, or bulge) and comorbidity-type degeneration (e.g., simultaneous presence of two or more conditions), respectively. A sample of lumbar magnetic resonance imaging (MRI) images from multiple clinical hospitals in China was collected and used in the proposal assessment. We devised a weighted transfer learning framework WDRIV-Net by ensembling four pre-trained models including Densenet169, ResNet101, InceptionV3, and VGG19. The proposed approach was applied to the clinical data and achieved 96.25% accuracy, surpassing the benchmark ResNet101 (87.5%), DenseNet169 (82.5%), VGG19 (88.75%), InceptionV3 (93.75%), and other state-of-the-art (SOTA) ensemble deep learning models. Furthermore, improved performance was observed as well for the metric of the area under the curve (AUC), producing a ≥ 7% increase versus other SOTA ensemble learning, a ≥ 6% increase versus most-studied models, and a ≥ 2% increase versus the baselines. WDRIV-Net can serve as a guide in the initial and efficient type screening of complex degeneration of lumbar intervertebral discs (LID) and assist in the early-stage selection of clinically differentiated treatment options.

Enhancing deep learning methods for brain metastasis detection through cross-technique annotations on SPACE MRI

BACKGROUND

Gadolinium-enhanced "sampling perfection with application-optimized contrasts using different flip angle evolution" (SPACE) sequence allows better visualization of brain metastases (BMs) compared to "magnetization-prepared rapid acquisition gradient echo" (MPRAGE). We hypothesize that this better conspicuity leads to high-quality annotation (HAQ), enhancing deep learning (DL) algorithm detection of BMs on MPRAGE images.

METHODS

Retrospective contrast-enhanced (gadobutrol 0.1 mmol/kg) SPACE and MPRAGE data of 157 patients with BM were used, either annotated on MPRAGE resulting in normal annotation quality (NAQ) or on coregistered SPACE resulting in HAQ. Multiple DL methods were developed with NAQ or HAQ using either SPACE or MRPAGE images and evaluated on their detection performance using positive predictive value (PPV), sensitivity, and F1 score and on their delineation performance using volumetric Dice similarity coefficient, PPV, and sensitivity on one internal and four additional test datasets (660 patients).

RESULTS

The SPACE-HAQ model reached 0.978 PPV, 0.882 sensitivity, and 0.916 F1-score. The MPRAGE-HAQ reached 0.867, 0.839, and 0.840, the MPRAGE NAQ 0.964, 0.667, and 0.798, respectively (p ≥ 0.157). Relative to MPRAGE-NAQ, the MPRAGE-HAQ F1-score detection increased on all additional test datasets by 2.5-9.6 points (p < 0.016) and sensitivity improved on three datasets by 4.6-8.5 points (p < 0.001). Moreover, volumetric instance sensitivity improved by 3.6-7.6 points (p < 0.001).

CONCLUSION

HAQ improves DL methods without specialized imaging during application time. HAQ alone achieves about 40% of the performance improvements seen with SPACE images as input, allowing for fast and accurate, fully automated detection of small (< 1 cm) BMs.

RELEVANCE STATEMENT

Training with higher-quality annotations, created using the SPACE sequence, improves the detection and delineation sensitivity of DL methods for the detection of brain metastases (BMs)on MPRAGE images. This MRI cross-technique transfer learning is a promising way to increase diagnostic performance.

KEY POINTS

Delineating small BMs on SPACE MRI sequence results in higher quality annotations than on MPRAGE sequence due to enhanced conspicuity. Leveraging cross-technique ground truth annotations during training improved the accuracy of DL models in detecting and segmenting BMs. Cross-technique annotation may enhance DL models by integrating benefits from specialized, time-intensive MRI sequences while not relying on them. Further validation in prospective studies is needed.

Physical activity may protect myelin via modulation of high-density lipoprotein

INTRODUCTION

Physical activity is associated with greater myelin content in older individuals with cerebral small vessel disease (CSVD), a condition marked by demyelination. However, potential mechanisms underlying this relationship remain understudied.

METHODS

We assessed cross-sectionally whether serum high-density lipoprotein (HDL), low-density lipoprotein, and triglycerides moderated the association between physical activity and in vivo myelin in older individuals with CSVD and mild cognitive impairment.

RESULTS

We included 81 highly educated, community-dwelling older individuals (mean age 74.57 years), 64% of whom were female. Regression models revealed that HDL levels significantly moderated the relationship between physical activity and myelin in the sagittal stratum, wherein higher physical activity levels were linked to greater myelin levels for those with average or high HDL (standardized B [95% CI] = 0.289 [0.087 to 0.491], p = 0.006).

DISCUSSION

Physical activity may promote myelin health partly through HDL. Data from longitudinal studies are needed to confirm our findings.

HIGHLIGHTS

Myelin loss is common in individuals with cerebral small vessel disease (CSVD). Physical activity was positively associated with myelin in older adults with CSVD. High-density lipoproteins (HDL) levels were also positively related to myelin. Physical activity effects on myelin were moderated by HDL levels.

Using partially shared radiomics features to simultaneously identify isocitrate dehydrogenase mutation status and epilepsy in glioma patients from MRI images

Prediction of isocitrate dehydrogenase (IDH) mutation status and epilepsy occurrence are important to glioma patients. Although machine learning models have been constructed for both issues, the correlation between them has not been explored. Our study aimed to exploit this correlation to improve the performance of both of the IDH mutation status identification and epilepsy diagnosis models in patients with glioma II-IV. 399 patients were retrospectively enrolled and divided into a training (n = 279) and an independent test (n = 120) cohort. Multi-center dataset (n = 228) from The Cancer Imaging Archive (TCIA) was used for external test for identification of IDH mutation status. Region of interests comprising the entire tumor and peritumoral edema were automatically segmented using a pre-trained deep learning model. Radiomic features were extracted from T1-weighted, T2-weighted, post-Gadolinium T1 weighted, and T2 fluid-attenuated inversion recovery images. We proposed an iterative approach derived from LASSO to select features shared by two tasks and features specific to each task, before using them to construct the final models. Receiver operating characteristic (ROC) analysis was employed to evaluate the model. The IDH mutation identification model achieved area under the ROC curve (AUC) values of 0.948, 0.946 and 0.860 on the training, internal test, and external test cohorts, respectively. The epilepsy diagnosis model achieved AUCs of 0.924 and 0.880 on the training and internal test cohorts, respectively. The proposed models can identify IDH status and epilepsy with fewer features, thus having better interpretability and lower risk of overfitting. This not only improves its chance of application in clinical settings, but also provides a new scheme to study multiple correlated clinical tasks.

Uncertainty quantification in multi-parametric MRI-based meningioma radiotherapy target segmentation

Purpose

This work investigates the use of a spherical projection-based U-Net (SPU-Net) segmentation model to improve meningioma segmentation performance and allow for uncertainty quantification.

Methods

A total of 76 supratentorial meningioma patients treated with radiotherapy were studied. Gross tumor volumes (GTVs) were contoured by a single experienced radiation oncologist on high-resolution contrast-enhanced T1 MRI scans (T1ce), and both T1 and T1ce images were utilized for segmentation. SPU-Net, an adaptation of U-Net incorporating spherical image projection to map 2D images onto a spherical surface, was proposed. As an equivalence of a nonlinear image transform, projections enhance locoregional details while maintaining the global field of view. By employing multiple projection centers, SPU-Net generates various GTV segmentation predictions, the variance indicating the model's uncertainty. This uncertainty is quantified on a pixel-wise basis using entropy calculations and aggregated through Otsu's method for a final segmentation.

Results/conclusion

The SPU-Net model poses an advantage over traditional U-Net models by providing a quantitative method of displaying segmentation uncertainty. Regarding segmentation performance, SPU-Net demonstrated comparable results to a traditional U-Net in sensitivity (0.758 vs. 0.746), Dice similarity coefficient (0.760 vs. 0.742), reduced mean Hausdorff distance (mHD) (0.612 cm vs 0.744 cm), and reduced 95% Hausdorff distance (HD95) (2.682 cm vs 2.912 cm). SPU-Net not only is comparable to U-Net in segmentation performance but also offers a significant advantage by providing uncertainty quantification. The added SPU-Net uncertainty mapping revealed low uncertainty in accurate segments (e.g., within GTV or healthy tissue) and higher uncertainty in problematic areas (e.g., GTV boundaries, dural tail), providing valuable insights for potential manual corrections. This advancement is particularly valuable given the complex extra-axial nature of meningiomas and involvement with dural tissue. The capability to quantify uncertainty makes SPU-Net a more advanced and informative tool for segmentation, without sacrificing performance.

Balance recovery and its link to vestibular agnosia in traumatic brain injury: a longitudinal behavioural and neuro-imaging study

BACKGROUND

Vestibular dysfunction causing imbalance affects c. 80% of acute hospitalized traumatic brain injury (TBI) cases. Poor balance recovery is linked to worse return-to-work rates and reduced longevity. We previously showed that white matter network disruption, particularly of right inferior longitudinal fasciculus, mediates the overlap between imbalance and impaired vestibular perception of self-motion (i.e., vestibular agnosia) in acute hospitalized TBI. However, there are no prior reports tracking the acute-longitudinal trajectory of objectively measured vestibular function for hospitalized TBI patients. We hypothesized that recovery of vestibular agnosia and imbalance is linked and mediated by overlapping brain networks.

METHODS

We screened 918 acute major trauma in-patients, assessed 146, recruited 39 acutely, and retested 34 at 6 months. Inclusion criteria were 18-65-year-old adults hospitalized for TBI with laboratory-confirmed preserved peripheral vestibular function. Benign paroxysmal positional vertigo and migraine were treated prior to testing. Vestibular agnosia was quantified by participants' ability to perceive whole-body yaw plane rotations via an automated rotating-chair algorithm. Subjective symptoms of imbalance (via questionnaires) and objective imbalance (via posturography) were also assessed.

RESULTS

Acute vestibular agnosia predicted poor balance recovery at 6 months. Recovery of vestibular agnosia and linked imbalance was mediated by bihemispheric fronto-posterior cortical circuits. Recovery of subjective symptoms of imbalance and objective imbalance were not correlated.

CONCLUSION

Vestibular agnosia mediates balance recovery post-TBI. The link between subjective dizziness and brain injury recovery, although important, is unclear. Therapeutic trials of vestibular recovery post-TBI should target enhancing bi-hemispheric connectivity and linked objective clinical measures (e.g., posturography).

A lifespan-generalizable skull-stripping model for magnetic resonance images that leverages prior knowledge from brain atlases

In magnetic resonance imaging of the brain, an imaging-preprocessing step removes the skull and other non-brain tissue from the images. But methods for such a skull-stripping process often struggle with large data heterogeneity across medical sites and with dynamic changes in tissue contrast across lifespans. Here we report a skull-stripping model for magnetic resonance images that generalizes across lifespans by leveraging personalized priors from brain atlases. The model consists of a brain extraction module that provides an initial estimation of the brain tissue on an image, and a registration module that derives a personalized prior from an age-specific atlas. The model is substantially more accurate than state-of-the-art skull-stripping methods, as we show with a large and diverse dataset of 21,334 lifespans acquired from 18 sites with various imaging protocols and scanners, and it generates naturally consistent and seamless lifespan changes in brain volume, faithfully charting the underlying biological processes of brain development and ageing.

The potential of GPT-4 advanced data analysis for radiomics-based machine learning models

Background

This study aimed to explore the potential of the Advanced Data Analytics (ADA) package of GPT-4 to autonomously develop machine learning models (MLMs) for predicting glioma molecular types using radiomics from MRI.

Methods

Radiomic features were extracted from preoperative MRI of n = 615 newly diagnosed glioma patients to predict glioma molecular types (IDH-wildtype vs IDH-mutant 1p19q-codeleted vs IDH-mutant 1p19q-non-codeleted) with a multiclass ML approach. Specifically, ADA was used to autonomously develop an ML pipeline and benchmark performance against an established handcrafted model using various MRI normalization methods (N4, Zscore, and WhiteStripe). External validation was performed on 2 public glioma datasets D2 (n = 160) and D3 (n = 410).

Results

GPT-4 achieved the highest accuracy of 0.820 (95% CI = 0.819-0.821) on the D3 dataset with N4/WS normalization, significantly outperforming the benchmark model's accuracy of 0.678 (95% CI = 0.677-0.680) (P < .001). Class-wise analysis showed performance variations across different glioma types. In the IDH-wildtype group, GPT-4 had a recall of 0.997 (95% CI = 0.997-0.997), surpassing the benchmark's 0.742 (95% CI = 0.740-0.743). For the IDH-mut 1p/19q-non-codel group, GPT-4's recall was 0.275 (95% CI = 0.272-0.279), lower than the benchmark's 0.426 (95% CI = 0.423-0.430). In the IDH-mut 1p/19q-codel group, GPT-4's recall was 0.199 (95% CI = 0.191-0.206), below the benchmark's 0.730 (95% CI = 0.721-0.738). On the D2 dataset, GPT-4's accuracy was significantly lower (P < .001) than the benchmark's, with N4/WS achieving 0.668 (95% CI = 0.666-0.671) compared with 0.719 (95% CI = 0.717-0.722) (P < .001). Class-wise analysis revealed the same pattern as observed in D3.

Conclusions

GPT-4 can autonomously develop radiomics-based MLMs, achieving performance comparable to handcrafted MLMs. However, its poorer class-wise performance due to unbalanced datasets shows limitations in handling complete end-to-end ML pipelines.

Structural-based uncertainty in deep learning across anatomical scales: Analysis in white matter lesion segmentation

This paper explores uncertainty quantification (UQ) as an indicator of the trustworthiness of automated deep-learning (DL) tools in the context of white matter lesion (WML) segmentation from magnetic resonance imaging (MRI) scans of multiple sclerosis (MS) patients. Our study focuses on two principal aspects of uncertainty in structured output segmentation tasks. First, we postulate that a reliable uncertainty measure should indicate predictions likely to be incorrect with high uncertainty values. Second, we investigate the merit of quantifying uncertainty at different anatomical scales (voxel, lesion, or patient). We hypothesize that uncertainty at each scale is related to specific types of errors. Our study aims to confirm this relationship by conducting separate analyses for in-domain and out-of-domain settings. Our primary methodological contributions are (i) the development of novel measures for quantifying uncertainty at lesion and patient scales, derived from structural prediction discrepancies, and (ii) the extension of an error retention curve analysis framework to facilitate the evaluation of UQ performance at both lesion and patient scales. The results from a multi-centric MRI dataset of 444 patients demonstrate that our proposed measures more effectively capture model errors at the lesion and patient scales compared to measures that average voxel-scale uncertainty values. We provide the UQ protocols code at https://github.com/Medical-Image-Analysis-Laboratory/MS_WML_uncs.

IGUANe: A 3D generalizable CycleGAN for multicenter harmonization of brain MR images

In MRI studies, the aggregation of imaging data from multiple acquisition sites enhances sample size but may introduce site-related variabilities that hinder consistency in subsequent analyses. Deep learning methods for image translation have emerged as a solution for harmonizing MR images across sites. In this study, we introduce IGUANe (Image Generation with Unified Adversarial Networks), an original 3D model that leverages the strengths of domain translation and straightforward application of style transfer methods for multicenter brain MR image harmonization. IGUANe extends CycleGAN by integrating an arbitrary number of domains for training through a many-to-one architecture. The framework based on domain pairs enables the implementation of sampling strategies that prevent confusion between site-related and biological variabilities. During inference, the model can be applied to any image, even from an unknown acquisition site, making it a universal generator for harmonization. Trained on a dataset comprising T1-weighted images from 11 different scanners, IGUANe was evaluated on data from unseen sites. The assessments included the transformation of MR images with traveling subjects, the preservation of pairwise distances between MR images within domains, the evolution of volumetric patterns related to age and Alzheimer's disease (AD), and the performance in age regression and patient classification tasks. Comparisons with other harmonization and normalization methods suggest that IGUANe better preserves individual information in MR images and is more suitable for maintaining and reinforcing variabilities related to age and AD. Future studies may further assess IGUANe in other multicenter contexts, either using the same model or retraining it for applications to different image modalities. Codes and the trained IGUANe model are available at https://github.com/RocaVincent/iguane_harmonization.git.

Advanced imaging reveals enhanced malignancy in glioblastomas involving the subventricular zone: evidence of increased infiltrative growth and perfusion

BACKGROUND

Glioblastoma's infiltrative growth and heterogeneity are influenced by neural, molecular, genetic, and immunological factors, with the precise origin of these tumors remaining elusive. Neurogenic zones might serve as the tumor stem cells' nest, with tumors in contact with these zones exhibiting worse outcomes and more aggressive growth patterns. This study aimed to determine if these characteristics are reflected in advanced imaging, specifically diffusion and perfusion data.

METHODS

In this monocentric retrospective study, 137 glioblastoma therapy-naive patients (IDH-wildtype, grade 4) with advanced preoperative MRI, including perfusion and diffusion imaging, were analyzed. Tumors and neurogenic zones were automatically segmented. Advanced imaging metrics, including cerebral blood volume (CBV) from perfusion imaging, tissue volume mask (TVM), and free water corrected fractional anisotropy (FA-FWE) from diffusion imaging, were extracted.

RESULTS

SVZ infiltration positively correlated with CBV, indicating higher perfusion in tumors. Significant CBV differences were noted between high and low SVZ infiltration cases at specific percentiles. Negative correlation was observed with TVM and positive correlation with FA-FWE, suggesting more infiltrative tumor growth. Significant differences in TVM and FA-FWE values were found between high and low SVZ infiltration cases.

DISCUSSION

Glioblastomas with SVZ infiltration exhibit distinct imaging characteristics, including higher perfusion and lower cell density per voxel, indicating a more infiltrative growth and higher vascularization. Stem cell-like characteristics in SVZ-infiltrating cells could explain the increased infiltration and aggressive behavior. Understanding these imaging and biological correlations could enhance the understanding of glioblastoma evolution.

AI-guided virtual biopsy: Automated differentiation of cerebral gliomas from other benign and malignant MRI findings using deep learning

Background

This study aimed to develop an automated algorithm to noninvasively distinguish gliomas from other intracranial pathologies, preventing misdiagnosis and ensuring accurate analysis before further glioma assessment.

Methods

A cohort of 1280 patients with a variety of intracranial pathologies was included. It comprised 218 gliomas (mean age 54.76 ± 13.74 years; 136 males, 82 females), 514 patients with brain metastases (mean age 59.28 ± 12.36 years; 228 males, 286 females), 366 patients with inflammatory lesions (mean age 41.94 ± 14.57 years; 142 males, 224 females), 99 intracerebral hemorrhages (mean age 62.68 ± 16.64 years; 56 males, 43 females), and 83 meningiomas (mean age 63.99 ± 13.31 years; 25 males, 58 females). Radiomic features were extracted from fluid-attenuated inversion recovery (FLAIR), contrast-enhanced, and noncontrast T1-weighted MR sequences. Subcohorts, with 80% for training and 20% for testing, were established for model validation. Machine learning models, primarily XGBoost, were trained to distinguish gliomas from other pathologies.

Results

The study demonstrated promising results in distinguishing gliomas from various intracranial pathologies. The best-performing model consistently achieved high area-under-the-curve (AUC) values, indicating strong discriminatory power across multiple distinctions, including gliomas versus metastases (AUC = 0.96), gliomas versus inflammatory lesions (AUC = 1.0), gliomas versus intracerebral hemorrhages (AUC = 0.99), gliomas versus meningiomas (AUC = 0.98). Additionally, across all these entities, gliomas had an AUC of 0.94.

Conclusions

The study presents an automated approach that effectively distinguishes gliomas from common intracranial pathologies. This can serve as a quality control upstream to further artificial-intelligence-based genetic analysis of cerebral gliomas.

Integrated Deep Learning Model for the Detection, Segmentation, and Morphologic Analysis of Intracranial Aneurysms Using CT Angiography

Purpose To develop a deep learning model for the morphologic measurement of unruptured intracranial aneurysms (UIAs) based on CT angiography (CTA) data and validate its performance using a multicenter dataset. Materials and Methods In this retrospective study, patients with CTA examinations, including those with and without UIAs, in a tertiary referral hospital from February 2018 to February 2021 were included as the training dataset. Patients with UIAs who underwent CTA at multiple centers between April 2021 and December 2022 were included as the multicenter external testing set. An integrated deep learning (IDL) model was developed for UIA detection, segmentation, and morphologic measurement using an nnU-Net algorithm. Model performance was evaluated using the Dice similarity coefficient (DSC) and intraclass correlation coefficient (ICC), with measurements by senior radiologists serving as the reference standard. The ability of the IDL model to improve performance of junior radiologists in measuring morphologic UIA features was assessed. Results The study included 1182 patients with UIAs and 578 controls without UIAs as the training dataset (median age, 55 years [IQR, 47-62 years], 1012 [57.5%] female) and 535 patients with UIAs as the multicenter external testing set (median age, 57 years [IQR, 50-63 years], 353 [66.0%] female). The IDL model achieved 97% accuracy in detecting UIAs and achieved a DSC of 0.90 (95% CI: 0.88, 0.92) for UIA segmentation. Model-based morphologic measurements showed good agreement with reference standard measurements (all ICCs > 0.85). Within the multicenter external testing set, the IDL model also showed agreement with reference standard measurements (all ICCs > 0.80). Junior radiologists assisted by the IDL model showed significantly improved performance in measuring UIA size (ICC improved from 0.88 [95% CI: 0.80, 0.92] to 0.96 [95% CI: 0.92, 0.97], P < .001). Conclusion The developed integrated deep learning model using CTA data showed good performance in UIA detection, segmentation, and morphologic measurement and may be used to assist less experienced radiologists in morphologic analysis of UIAs. Keywords: Segmentation, CT Angiography, Head/Neck, Aneurysms, Comparative Studies Supplemental material is available for this article. © RSNA, 2024 See also the commentary by Wang in this issue.

Automatic detection and multi-component segmentation of brain metastases in longitudinal MRI

Manual segmentation of lesions, required for radiotherapy planning and follow-up, is time-consuming and error-prone. Automatic detection and segmentation can assist radiologists in these tasks. This work explores the automated detection and segmentation of brain metastases (BMs) in longitudinal MRIs. It focuses on several important aspects: identifying and segmenting new lesions for screening and treatment planning, re-segmenting lesions in successive images using prior lesion locations as an additional input channel, and performing multi-component segmentation to distinguish between enhancing tissue, edema, and necrosis. The key component of the proposed approach is to propagate the lesion mask from the previous time point to improve the detection performance, which we refer to as "re-segmentation". The retrospective data includes 518 metastases in 184 contrast-enhanced T1-weighted MRIs originating from 49 patients (63% male, 37% female). 131 time-points (36 patients, 418 BMs) are used for cross-validation, the remaining 53 time-points (13 patients, 100 BMs) are used for testing. The lesions were manually delineated with label 1: enhancing lesion, label 2: edema, and label 3: necrosis. One-tailed t-tests are used to compare model performance including multiple segmentation and detection metrics. Significance is considered as p < 0.05. A Dice Similarity Coefficient (DSC) of 0.79 and F 1 -score of 0.80 are obtained for the segmentation of new lesions. On follow-up scans, the re-segmentation model significantly outperforms the segmentation model (DSC and F 1 0.78 and 0.88 vs 0.56 and 0.60). The re-segmentation model also significantly outperforms the simple segmentation model on the enhancing lesion (DSC 0.76 vs 0.53) and edema (0.52 vs 0.47) components, while similar scores are obtained on the necrosis component (0.62 vs 0.63). Additionally, we analyze the correlation between lesion size and segmentation performance, as demonstrated in various studies that highlight the challenges in segmenting small lesions. Our findings indicate that this correlation disappears when utilizing the re-segmentation approach and evaluating with the unbiased normalized DSC. In conclusion, the automated segmentation of new lesions and subsequent re-segmentation in follow-up images was achievable, with high level of performance obtained for single- and multiple-component segmentation tasks.

Amide proton transfer weighted MRI measurements yield consistent and repeatable results in patients with gliomas: a prospective test-retest study

OBJECTIVES

Chemical exchange saturation transfer (CEST) imaging has emerged as a promising imaging biomarker, but its reliability for clinical practice remains uncertain. This study aimed to investigate the robustness of CEST parameters in healthy volunteers and patients with brain tumours.

METHODS

A total of n = 52 healthy volunteers and n = 52 patients with histologically confirmed glioma underwent two consecutive 3-T MRI scans separated by a 1-min break. The CEST measurements were reconstructed using two models: with and without fluid suppression and included the evaluation of both amide (amidePTw) and amine (aminePTw) offsets. Mean intensity values in healthy volunteers were compared from volumetric segmentations (VOI) of grey matter, white matter, and the whole brain. Mean intensity values in brain tumour patients were assessed from VOI of the contrast-enhancing, non-enhancing and whole tumour, as well as from the normal-appearing white matter. Test-retest reliability was assessed using ICC and Bland-Altman plots.

RESULTS

The amidePTw/aminePTw signal intensity distribution was significantly affected by fluid suppression (p < 0.001 for each VOI). Test-retest reliability in healthy volunteers showed fair to excellent agreement (ICC = 0.53-0.74), with the highest signal intensity values observed by amidePTw (ICC = 0.73-0.74). In patients, an excellent agreement of both amidePTw and aminePTw measurements was observed across different tumour regions (ICC = 0.76-0.89), with the highest ICC for contrast-enhancing tumour measurements. Bland-Altman analysis indicated negligible systematic bias and no proportional bias in measurement errors.

CONCLUSION

Measurements from amide/aminePTw imaging obtained from an adequately powered test-retest study yield consistent and reproducible results in glioma patients, as a prerequisite for robust imaging biomarker discovery in neuro-oncology.

KEY POINTS

Question The clinical reliability of chemical exchange saturation transfer imaging remains uncertain, necessitating further investigation to establish its robustness as a biomarker in neuro-oncology. Findings This study demonstrates that amide/amine proton transfer imaging provides repeatable, high-agreement measurements in glioma patients, particularly in contrast-enhancing tumour regions. Clinical relevance This test-retest study demonstrates that chemical exchange saturation transfer imaging using two models and assessing amide and amine offsets yield consistent and repeatable results in glioma patients, as a prerequisite for robust imaging biomarker discovery for neuro-oncology studies and clinical practice.

The Brain Tumor Segmentation - Metastases (BraTS-METS) Challenge 2023: Brain Metastasis Segmentation on Pre-treatment MRI

The translation of AI-generated brain metastases (BM) segmentation into clinical practice relies heavily on diverse, high-quality annotated medical imaging datasets. The BraTS-METS 2023 challenge has gained momentum for testing and benchmarking algorithms using rigorously annotated internationally compiled real-world datasets. This study presents the results of the segmentation challenge and characterizes the challenging cases that impacted the performance of the winning algorithms. Untreated brain metastases on standard anatomic MRI sequences (T1, T2, FLAIR, T1PG) from eight contributed international datasets were annotated in stepwise method: published UNET algorithms, student, neuroradiologist, final approver neuroradiologist. Segmentations were ranked based on lesion-wise Dice and Hausdorff distance (HD95) scores. False positives (FP) and false negatives (FN) were rigorously penalized, receiving a score of 0 for Dice and a fixed penalty of 374 for HD95. The mean scores for the teams were calculated. Eight datasets comprising 1303 studies were annotated, with 402 studies (3076 lesions) released on Synapse as publicly available datasets to challenge competitors. Additionally, 31 studies (139 lesions) were held out for validation, and 59 studies (218 lesions) were used for testing. Segmentation accuracy was measured as rank across subjects, with the winning team achieving a LesionWise mean score of 7.9. The Dice score for the winning team was 0.65 ± 0.25. Common errors among the leading teams included false negatives for small lesions and misregistration of masks in space. The Dice scores and lesion detection rates of all algorithms diminished with decreasing tumor size, particularly for tumors smaller than 100 mm3. In conclusion, algorithms for BM segmentation require further refinement to balance high sensitivity in lesion detection with the minimization of false positives and negatives. The BraTS-METS 2023 challenge successfully curated well-annotated, diverse datasets and identified common errors, facilitating the translation of BM segmentation across varied clinical environments and providing personalized volumetric reports to patients undergoing BM treatment.

A foundation model for generalized brain MRI analysis

Artificial intelligence (AI) applied to brain magnetic resonance imaging (MRI) has the potential to improve disease diagnosis and management but requires algorithms with generalizable knowledge that can perform well in a variety of clinical scenarios. The field has been constrained, thus far, by limited training data and task-specific models that do not generalize well across patient populations and medical tasks. Foundation models, by leveraging self-supervised learning, pretraining, and targeted adaptation, present a promising paradigm to overcome these limitations. Here, we present Brain Imaging Adaptive Core (BrainIAC), a novel foundation model designed to learn generalized representations from unlabeled brain MRI data and serve as a core basis for diverse downstream application adaptation. Trained and validated on 48,519 brain MRIs across a broad spectrum of tasks, we demonstrate that BrainIAC outperforms localized supervised training and other pretrained models, particularly in low-data settings and high-difficulty tasks, allowing for application in scenarios otherwise infeasible. BrainIAC can be integrated into imaging pipelines and multimodal frameworks and may lead to improved biomarker discovery and AI clinical translation.

Acute nicotinamide riboside supplementation increases human cerebral NAD+ levels in vivo

PURPOSE

The purpose of this study was to determine the effect of acute nicotinamide riboside (NR) supplementation on cerebral nicotinamide adenine dinucleotide (NAD+) levels in the human brain in vivo by means of downfield proton MRS (DF 1H MRS).

METHODS

DF 1H MRS was performed on 10 healthy volunteers in a 7.0 T MRI scanner with spectrally selective excitation and spatially selective localization to determine cerebral NAD+ levels on two back-to-back days: once after an overnight fast (baseline) and once 4 h after oral ingestion of nicotinamide riboside (900 mg). Additionally, two more baseline scans were performed following the same paradigm to assess test-retest reliability of the NAD+ levels in the absence of NR.

RESULTS

NR supplementation increased mean NAD+ concentration compared to the baseline (0.458 ± 0.053 vs. 0.392 ± 0.058 mM; p < 0.001). The additional two baseline scans demonstrated no differences in mean NAD+ concentrations (0.425 ± 0.118 vs. 0.405 ± 0.082 mM; p = 0.45), and no difference from the first baseline scan (F(2, 16) = 0.907; p = 0.424).

CONCLUSION

These preliminary results confirm that acute NR supplementation increases cerebral NAD+ levels in healthy human volunteers and shows the promise of DF 1H MRS utility for robust detection of NAD+ in humans in vivo.

IVIM parameters mapping with artificial neural network based on mean deviation prior

BACKGROUND

The diffusion and perfusion parameters derived from intravoxel incoherent motion (IVIM) imaging provide promising biomarkers for noninvasively quantifying and managing various diseases. Nevertheless, due to the distribution gap between simulated and real datasets, the out-of-distribution (OOD) problem occurred in supervised learning-based methods degrades their performance and hinders their real applications.

PURPOSE

To address the OOD problem in supervised methods and to further improve the accuracy and stability of IVIM parameter estimation, this work proposes a novel learning framework called IterANN, based on mean deviation prior (MDP) between training and estimated IVIM parameters on the test set.

METHODS

Specifically, MDP indicates that the mean of the estimated IVIM parameters always locates between the mean of IVIM parameters in the test and train sets. In IterANN, we adopt a very simple artificial neural network (ANN) architecture of two hidden layers with 12 neurons per hidden layer, an input layer containing the signals acquired at multiple b-values and an output layer composed of three IVIM parameters ( D $D$ , F $F$ andD S t a r $DStar$ ). Inspired by MDP, the distribution of IVIM parameters in the training set (simulated data) is iteratively updated so that their mean gradually approaches the predicted values of the real data. This aims to achieve a strong correlation between the simulated data and the real data. To validate the effectiveness of IterANN, we compare it with several methods on both simulation and real acquisition datasets, including 21 healthy and 3 tumor subjects, in terms of residual errors of IVIM parameters or DW signals, the coefficients of variation (CV) of IVIM parameters, and the parameter contrast-to-noise ratio (PCNR) between normal and tumor tissues.

RESULTS

On two simulation datasets, the proposed IterANN achieves the lowest residual error in IVIM parameters, especially in the case of low signal-to-noise ratio (SNR = 10), the residual error of D $D$ , F $F$ andD S t a r $DStar$ is decreased by15.82 % / 14.92 % , 81.19 % / 74.04 % , 50.77 % / 1.549 % $15.82\%/14.92\%, 81.19\%/74.04\%, 50.77\%/1.549\%$ (Gaussian distribution /realistic distribution) respectively comparing to the suboptimal method. On real dataset, the IterANN achieves the highest PCNR when comparing the normal and tumor regions. Additionally, the proposed IterANN demonstrated better stability, with its CV being significantly lower than that of other methods in the vast majority of cases (p < 0.01 $p<0.01$ , paired-sample Student's t-test).

CONCLUSIONS

The superior performance of IterANN demonstrates that updating the distribution of the train set based on MDP can effectively solve the OOD problem, which allows us not only to improve the accuracy and stability of the estimated IVIM parameters, but also to increase the potential of IVIM in disease diagnosis.

Diffusion MRI is superior to quantitative T2-FLAIR mismatch in predicting molecular subtypes of human non-enhancing gliomas

PURPOSE

This study compared the classification performance of normalized apparent diffusion coefficient (nADC) with percentage T2-FLAIR mismatch-volume (%T2FM-volume) for differentiating between IDH-mutant astrocytoma (IDHm-A) and other glioma molecular subtypes.

METHODS

A total of 105 non-enhancing gliomas were studied. T2-FLAIR digital subtraction maps were used to identify T2FM and T2-FLAIR non-mismatch (T2FNM) subregions within tumor volumes of interest (VOIs). Median nADC from the whole tumor, T2FM, and T2NFM subregions and %T2FM-volume were obtained. IDHm-A classification analyses using receiver-operating characteristic curves and multiple logistic regression were performed in addition to exploratory survival analyses.

RESULTS

T2FM subregions had significantly higher nADC than T2FNM subregions within IDHm-A with ≥ 25% T2FM-volume (P < 0.0001). IDHm-A with ≥ 25% T2FM-volume demonstrated significantly higher whole tumor nADC compared to IDHm-A with < 25% T2FM-volume (P < 0.0001), and both IDHm-A subgroups demonstrated significantly higher nADC compared to IDH-mutant oligodendroglioma and IDH-wild-type gliomas (P < 0.05). For classification of IDHm-A vs. other gliomas, the area under curve (AUC) of nADC was significantly greater compared to the AUC of %T2FM-volume (P = 0.01, nADC AUC = 0.848, %T2FM-volume AUC = 0.714) along with greater sensitivity. In exploratory survival analyses within IDHm-A, %T2FM-volume was not associated with overall survival (P = 0.2), but there were non-significant trends for nADC (P = 0.07) and tumor volume (P = 0.051).

CONCLUSION

T2-FLAIR subtraction maps are useful for characterizing IDHm-A imaging characteristics. nADC outperforms %T2FM-volume for classifying IDHm-A amongst non-enhancing gliomas with preserved high specificity and increased sensitivity, which may be related to inherent diffusivity differences regardless of T2FM. In line with previous findings on visual T2FM-sign, quantitative %T2FM-volume may not be prognostic.

TransferGWAS of T1-weighted brain MRI data from UK Biobank

Genome-wide association studies (GWAS) traditionally analyze single traits, e.g., disease diagnoses or biomarkers. Nowadays, large-scale cohorts such as UK Biobank (UKB) collect imaging data with sample sizes large enough to perform genetic association testing. Typical approaches to GWAS on high-dimensional modalities extract predefined features from the data, e.g., volumes of regions of interest. This limits the scope of such studies to predefined traits and can ignore novel patterns present in the data. TransferGWAS employs deep neural networks (DNNs) to extract low-dimensional representations of imaging data for GWAS, eliminating the need for predefined biomarkers. Here, we apply transferGWAS on brain MRI data from UKB. We encoded 36, 311 T1-weighted brain magnetic resonance imaging (MRI) scans using DNN models trained on MRI scans from the Alzheimer's Disease Neuroimaging Initiative, and on natural images from the ImageNet dataset, and performed a multivariate GWAS on the resulting features. We identified 289 independent loci, associated among others with bone density, brain, or cardiovascular traits, and 11 regions having no previously reported associations. We fitted polygenic scores (PGS) of the deep features, which improved predictions of bone mineral density and several other traits in a multi-PGS setting, and computed genetic correlations with selected phenotypes, which pointed to novel links between diffusion MRI traits and type 2 diabetes. Overall, our findings provided evidence that features learned with DNN models can uncover additional heritable variability in the human brain beyond the predefined measures, and link them to a range of non-brain phenotypes.

Radiomics features for the discrimination of tuberculomas from high grade gliomas and metastasis: a multimodal study

BACKGROUND

Tuberculomas are prevalent in developing countries and demonstrate variable signals on MRI resulting in the overlap of the conventional imaging phenotype with other entities including glioma and brain metastasis. An accurate MRI diagnosis is important for the early institution of anti-tubercular therapy, decreased patient morbidity, mortality, and prevents unnecessary neurosurgical excision. This study aims to assess the potential of radiomics features of regular contrast images including T1W, T2W, T2W FLAIR, T1W post contrast images, and ADC maps, to differentiate between tuberculomas, high-grade-gliomas and metastasis, the commonest intra parenchymal mass lesions encountered in the clinical practice.

METHODS

This retrospective study includes 185 subjects. Images were resampled, co-registered, skull-stripped, and zscore-normalized. Automated lesion segmentation was performed followed by radiomics feature extraction, train-test split, and features reduction. All machine learning algorithms that natively support multiclass classification were trained and assessed on features extracted from individual modalities as well as combined modalities. Model explainability of the best performing model was calculated using the summary plot obtained by SHAP values.

RESULTS

Extra tree classifier trained on the features from ADC maps was the best classifier for the discrimination of tuberculoma from high-grade-glioma and metastasis with AUC-score of 0.96, accuracy-score of 0.923, Brier-score of 0.23.

CONCLUSION

This study demonstrates that radiomics features are effective in discriminating between tuberculoma, metastasis, and high-grade-glioma with notable accuracy and AUC scores. Features extracted from the ADC maps surfaced as the most robust predictors of the target variable.

A minimalistic approach to classifying Alzheimer's disease using simple and extremely small convolutional neural networks

BACKGROUND

There is a broad interest in deploying deep learning-based classification algorithms to identify individuals with Alzheimer's disease (AD) from healthy controls (HC) based on neuroimaging data, such as T1-weighted Magnetic Resonance Imaging (MRI). The goal of the current study is to investigate whether modern, flexible architectures such as EfficientNet provide any performance boost over more standard architectures.

METHODS

MRI data was sourced from the Alzheimer's Disease Neuroimaging Initiative (ADNI) and processed with a minimal preprocessing pipeline. Among the various architectures tested, the minimal 3D convolutional neural network SFCN stood out, composed solely of 3x3x3 convolution, batch normalization, ReLU, and max-pooling. We also examined the influence of scale on performance, testing SFCN versions with trainable parameters ranging from 720 up to 2.9 million.

RESULTS

SFCN achieves a test ROC AUC of 96.0% while EfficientNet got an ROC AUC of 94.9 %. SFCN retained high performance down to 720 trainable parameters, achieving an ROC AUC of 91.4%.

COMPARISON WITH EXISTING METHODS

The SFCN is compared to DenseNet and EfficientNet as well as the results of other publications in the field.

CONCLUSIONS

The results indicate that using the minimal 3D convolutional neural network SFCN with a minimal preprocessing pipeline can achieve competitive performance in AD classification, challenging the necessity of employing more complex architectures with a larger number of parameters. This finding supports the efficiency of simpler deep learning models for neuroimaging-based AD diagnosis, potentially aiding in better understanding and diagnosing Alzheimer's disease.

Atlas-Based Structural Disconnectomes Are Associated with Cognitive Performance in Brain Tumors

Background: Brain tumors are associated with impaired cognitive functioning, which may result from disruptions in brain structural connectivity. Estimating structural disconnections is a more advantageous representation of tumor impact and can be performed indirectly through normative brain atlases. Materials and Methods: Using a publicly available dataset of glioma and meningioma patient MRI scans and tumor masks, latent correlations were estimated between measures of structural disconnection and attention-based cognitive functioning. These measures included gray matter (GM) parcel damage, white matter tract damage, GM parcel-to-parcel disconnections, and reaction time (RTI) as part of the Cambridge Neuropsychological Test Automated Battery to assess attention. Results: Preprocessing pipelines with two different methods of minimizing the pathology impact on MRI normalization were utilized: cost-function masking and lesion filling. The results across both pipelines were nearly consistent, with significant correlations mainly found between RTI measures and the damage to the left inferior fronto-occipital and uncinate fasciculus, as well as the left prefrontal-visual disconnections. Conclusions: This alludes to the importance of left-hemispheric prefrontal-visual coupling in attention-based tasks, particularly those involving object- and feature-based attention.

Amide proton transfer-weighted CEST MRI for radiotherapy target delineation of glioblastoma: a prospective pilot study

BACKGROUND

Extensive glioblastoma infiltration justifies a 15-mm margin around the gross tumor volume (GTV) to define the radiotherapy clinical target volume (CTV). Amide proton transfer (APT)-weighted imaging could enable visualization of tumor infiltration, allowing more accurate GTV delineation. We quantified the impact of integrating APT-weighted imaging into GTV delineation of glioblastoma and compared two APT-weighted quantification methods-magnetization transfer ratio asymmetry (MTRasym) and Lorentzian difference (LD) analysis-for target delineation.

METHODS

Nine glioblastoma patients underwent an extended imaging protocol prior to radiotherapy, yielding APT-weighted MTRasym and LD maps. From both maps, biological tumor volumes were generated (BTVMTRasym and BTVLD) and added to the conventional GTV to generate biological GTVs (GTVbio,MTRasym and GTVbio,LD). Wilcoxon signed-rank tests were performed for comparisons.

RESULTS

The GTVbio,MTRasym and GTVbio,LD were significantly larger than the conventional GTV (p ≤ 0.022), with a median volume increase of 9.3% and 2.1%, respectively. The GTVbio,MTRasym and GTVbio,LD were significantly smaller than the CTV (p = 0.004), with a median volume reduction of 72.1% and 70.9%, respectively. There was no significant volume difference between the BTVMTRasym and BTVLD (p = 0.074). In three patients, BTVMTRasym delineation was affected by elevated signals at the brain periphery due to residual motion artifacts; this elevation was absent on the APT-weighted LD maps.

CONCLUSION

Larger biological GTVs compared to the conventional GTV highlight the potential of APT-weighted imaging for radiotherapy target delineation of glioblastoma. APT-weighted LD mapping may be advantageous for target delineation as it may be more robust against motion artifacts.

RELEVANCE STATEMENT

The introduction of APT-weighted imaging may, ultimately, enhance visualization of tumor infiltration and eliminate the need for the substantial 15-mm safety margin for target delineation of glioblastoma. This could reduce the risk of radiation toxicity while still effectively irradiating the tumor.

TRIAL REGISTRATION

NCT05970757 (ClinicalTrials.gov).

KEY POINTS

Integration of APT-weighted imaging into target delineation for radiotherapy is feasible. The integration of APT-weighted imaging yields larger GTVs in glioblastoma. APT-weighted LD mapping may be more robust against motion artifacts than APT-weighted MTRasym.