The ANTsX ecosystem for quantitative biological and medical imaging

Prenatal Household Income Instability and Infant Subcortical Brain Volumes

Prenatal stress exposure may negatively influence the development of the amygdala and hippocampus. Although there is significant income instability during pregnancy, and it can increase stress among pregnant parents, the impact of income instability on brain development is not well understood. The present study examined the association between household income losses during pregnancy and hippocampus and amygdala volumes in early infancy. A total of 63 infants from a prospective longitudinal study of pregnant individuals and their infants completed an MRI during natural sleep. The total number of negative month-to-month earnings shocks (defined as an arc percent change [APC] of -25 or greater) was significantly associated with smaller right hippocampal and right amygdala volumes. This suggests that household income losses during the perinatal period are associated with infant brain structure after birth. These findings provide support for the development of public programs that prioritize financial consistency, especially during sensitive periods like pregnancy.

Association of Hypoxemia Due to Obstructive Sleep Apnea With White Matter Hyperintensities and Temporal Lobe Changes in Older Adults

BACKGROUND AND OBJECTIVES

Cerebral small vessel disease (CSVD) is a leading cause of cognitive decline and functional loss in older adults. Obstructive sleep apnea (OSA) is common in older adults, can increase cerebrovascular disease risk, and is linked to medial temporal lobe (MTL) degeneration and cognitive impairment. However, the interaction between OSA features and CSVD burden and their combined effect on MTL structure and function are not well understood. This study tested the hypothesis that CSVD burden is a candidate mechanism linking OSA to MTL degeneration and impaired memory in older adults.

METHODS

Cognitively unimpaired older adults from the Biomarker Exploration in Aging, Cognition, and Neurodegeneration cohort were recruited for an observational, in-lab overnight polysomnography (PSG) study with emotional mnemonic discrimination ability assessed before and after sleep. Participants had no neurologic or psychiatric disorders and were not on sleep-affecting medications. PSG-derived OSA variables included apnea-hypopnea index (AHI), total arousal index, and minimum SpO2. MRI was used to assess global and lobar white matter hyperintensity (WMH) volumes and MTL structure (hippocampal volume; entorhinal cortex [ERC] thickness) at an earlier time point. Regressions were implemented while adjusting for age, sex, and concurrent use of antihyperlipidemic and/or antihypertensive medication. Minimum SpO2 was transformed into a Hypoxemia Severity Index for normality, in which lower SpO2 values indicated more severe hypoxemia.

RESULTS

Thirty-seven older adults were included in the study (age 72.5 ± 5.6 years, 23 women, AHI = 13.8 ± 18.0 [range 0-80]). Hypoxemia measures significantly predicted global WMH volume (bminSpO2 = 0.141 [0.001-0.282], bduration <90% = 0.008 [0.000-0.016]). This relationship was driven by hypoxemia severity during REM sleep (bREMminSpO2 = 0.143 [0.003-0.284]), which also predicted frontal (bREMminSpO2 = 0.101 [0.004-0.198]) and parietal (bREMminSpO2 = 0.121 [0.024-0.219]) WMH burden. Greater frontal WMH burden indirectly mediated the relationship between REM sleep hypoxemia and ERC thickness (indirect effect = -0.043, 95% CI -0.1174 to -0.00015). Reduced ERC thickness was, in turn, associated with worse overnight mnemonic discrimination ability (bleftERCthickness = 0.112 [0.014-0.211]).

DISCUSSION

These findings identify CSVD as a candidate mechanism linking OSA-related hypoxemia to MTL degeneration and impaired sleep-dependent memory in older adults, specifically implicating hypoxic events during REM sleep.

Characterization of Portable Ultra-Low Field MRI Scanners for Multi-Center Structural Neuroimaging

The lower infrastructure requirements of portable ultra-low field MRI (ULF-MRI) systems have enabled their use in diverse settings such as intensive care units and remote medical facilities. The UNITY Project is an international neuroimaging network harnessing this technology, deploying portable ULF-MRI systems globally to expand access to MRI for studies into brain development. Given the wide range of environments where ULF-MRI systems may operate, there are external factors that might influence image quality. This work aims to introduce the quality control (QC) framework used by the UNITY Project to investigate how robust the systems are and how QC metrics compare between sites and over time. We present a QC framework using a commercially available phantom, scanned with 64 mT portable MRI systems at 17 sites across 12 countries on four continents. Using automated, open-source analysis tools, we quantify signal-to-noise, image contrast, and geometric distortions. Our results demonstrated that the image quality is robust to the varying operational environment, for example, electromagnetic noise interference and temperature. The Larmor frequency was significantly correlated to room temperature, as was image noise and contrast. Image distortions were less than 2.5 mm, with high robustness over time. Similar to studies at higher field, we found that changes in pulse sequence parameters from software updates had an impact on QC metrics. This study demonstrates that portable ULF-MRI systems can be deployed in a variety of environments for multi-center neuroimaging studies and produce robust results.

Semi-Supervised Learning Allows for Improved Segmentation With Reduced Annotations of Brain Metastases Using Multicenter MRI Data

BACKGROUND

Deep learning-based segmentation of brain metastases relies on large amounts of fully annotated data by domain experts. Semi-supervised learning offers potential efficient methods to improve model performance without excessive annotation burden.

PURPOSE

This work tests the viability of semi-supervision for brain metastases segmentation.

STUDY TYPE

Retrospective.

SUBJECTS

There were 156, 65, 324, and 200 labeled scans from four institutions and 519 unlabeled scans from a single institution. All subjects included in the study had diagnosed with brain metastases.

FIELD STRENGTH/SEQUENCES

1.5 T and 3 T, 2D and 3D T1-weighted pre- and post-contrast, and fluid-attenuated inversion recovery (FLAIR).

ASSESSMENT

Three semi-supervision methods (mean teacher, cross-pseudo supervision, and interpolation consistency training) were adapted with the U-Net architecture. The three semi-supervised methods were compared to their respective supervised baseline on the full and half-sized training.

STATISTICAL TESTS

Evaluation was performed on a multinational test set from four different institutions using 5-fold cross-validation. Method performance was evaluated by the following: the number of false-positive predictions, the number of true positive predictions, the 95th Hausdorff distance, and the Dice similarity coefficient (DSC). Significance was tested using a paired samples t test for a single fold, and across all folds within a given cohort.

RESULTS

Semi-supervision outperformed the supervised baseline for all sites with the best-performing semi-supervised method achieved an on average DSC improvement of 6.3% ± 1.6%, 8.2% ± 3.8%, 8.6% ± 2.6%, and 15.4% ± 1.4%, when trained on half the dataset and 3.6% ± 0.7%, 2.0% ± 1.5%, 1.8% ± 5.7%, and 4.7% ± 1.7%, compared to the supervised baseline on four test cohorts. In addition, in three of four datasets, the semi-supervised training produced equal or better results than the supervised models trained on twice the labeled data.

DATA CONCLUSION

Semi-supervised learning allows for improved segmentation performance over the supervised baseline, and the improvement was particularly notable for independent external test sets when trained on small amounts of labeled data.

PLAIN LANGUAGE SUMMARY

Artificial intelligence requires extensive datasets with large amounts of annotated data from medical experts which can be difficult to acquire due to the large workload. To compensate for this, it is possible to utilize large amounts of un-annotated clinical data in addition to annotated data. However, this method has not been widely tested for the most common intracranial brain tumor, brain metastases. This study shows that this approach allows for data efficient deep learning models across multiple institutions with different clinical protocols and scanners.

LEVEL OF EVIDENCE

3 TECHNICAL EFFICACY: Stage 2.

Transformation trees - Documentation of multimodal image registration

Multimodal image registration plays a key role in creating digital patient models by combining data from different imaging techniques into a single coordinate system. This process often involves multiple sequential and interconnected transformations, which must be well-documented to ensure transparency and reproducibility. In this paper, we propose the use of transformation trees as a method for structured recording and management of these transformations. This approach has been implemented in the dpVision software and uses a dedicated .dpw file format to store hierarchical relationships between images, transformations, and motion data. Transformation trees allow precise tracking of all image processing steps, reduce the need to store multiple copies of the same data, and enable the indirect registration of images that do not share common reference points. This improves the reproducibility of the analyses and facilitates later processing and integration of images from different sources. The practical application of this method is demonstrated with examples from orthodontics, including the integration of 3D face scans, intraoral scans, and CBCT images, as well as the documentation of mandibular motion. Beyond orthodontics, this method can be applied in other fields that require systematic management of image registration processes, such as maxillofacial surgery, oncology, and biomechanical analysis. Maintaining long-term data consistency is essential for both scientific research and clinical practice. It enables easier comparison of results in longitudinal studies, improves retrospective analysis, and supports the development of artificial intelligence algorithms by providing standardized and well-documented datasets. The proposed approach enhances data organization, allows for efficient analysis, and facilitates the reuse of information in future studies and diagnostic procedures.

The effects of estradiol on subcortical brain volumes in perimenopausal-onset depression

BACKGROUND

Perimenopause is associated with increases in depressive and vasomotor symptoms (VMS), which can be alleviated with transdermal estradiol (TE2) administration. Subcortical brain regions are commonly implicated in depression, are dense with E2 receptors and are susceptible to volumetric changes resulting from E2 regulation of synaptic density. No studies have examined linkages among TE2 administration, perimenopausal-onset major depression (PO-MDD) and subcortical brain volumes.

METHODS

This is an exploratory data analysis of change in subcortical brain volumes measured via 3 T MRI before and after three-weeks of TE2 administration in 14 women with PO-MDD and 17 euthymic controls. Regions of interest were the hippocampus, amygdala, putamen, thalamus, and caudate nucleus. Multilevel models examined relations between baseline volumes and volumetric changes with symptom trajectories in the PO-MDD group.

RESULTS

In the PO-MDD group, anhedonia (p < 0.004) and VMS (p < 0.001) significantly reduced following TE2 administration. There was a significant Group X Time interaction in the right hippocampus (p < 0.01), driven by volume increases in the control group (p < 0.001). In the PO-MDD group, change in right hippocampal volumes significantly predicted decreases in anhedonia trajectories from baseline to week 2 and week 3 (p's < 0.001) and decreases in VMS across all timepoints (p's < 0.001).

DISCUSSION

Women with PO-MDD, who presented with more severe baseline anhedonia and VMS, experienced greater reductions in anhedonia, VMS, and hippocampal volumes, demonstrating a greater response to E2. Hippocampal volume change may be a candidate for predicting treatment response to E2 for anhedonia and vasomotor symptoms in women with PO-MDD. These findings should be validated with a placebo-controlled trial.

Susceptibility to multitasking in stroke is associated to multiple-demand system damage and leads to lateralized visuospatial deficits

Cognitive impairment after stroke is heterogeneous: there is no strict correspondence between brain damage and magnitude of deficit or recovery. Protective factors such as cognitive or brain reserve have been invoked to explain the mismatch. Here, we consider the opposite point of view: the instances in which this protection is overturned. We leveraged on multitasking to stress the brain's processing limits and unveil deficits that may be missed by standard testing in a sample of 46 patients with unilateral subacute to chronic stroke and no sign of lateralized spatial-attentional disorders at neuropsychological paper-and-pencil tests. Multivariate analyses identified a phenotype of patients with high susceptibility to multitasking, showing stark contralesional spatial awareness deficit only when multitasking. Multivariate brain-behavior mapping based on lesions location and structural disconnections pointed to the Multiple-Demand System, a network of frontal and fronto-parietal areas subserving domain-general processes. Damage in this network may critically interact with domain-specific processes, resulting in subtle and yet invalidating deficits. Indeed, these patients (one-third of the sample) presented worse performance in tests evaluating activities of daily living and domain-general abilities. We conclude that the theoretical construct of susceptibility to multitasking helps understanding what marks the passage to clinically visible deficits after brain damage.

High-resolution mapping of substantia nigra in Parkinson's disease using 7 tesla magnetic resonance imaging

Parkinson's disease causes a progressive loss of dopaminergic neurons and iron accumulation in the substantia nigra pars compacta. Using ultra-high field magnetic resonance imaging (MRI) at 7 tesla in 43 Parkinson's patients and 24 healthy controls, we analyzed the voxel-wise pattern of structural disintegration of dopamine neurons with neuromelanin-sensitive MRI, along with assessing iron accumulation using R2* and quantitative susceptibility mapping (QSM). We also explored correlations between these measures and the severity of residual motor symptoms in the on-medication state and other clinical variables. Differences were most notable in the nigrosomes within the pars compacta, with patients showing reduced neuromelanin signals and increased QSM values. Severity and asymmetry of motor symptoms correlated with higher R2* and QSM values in nigrosome N1. Thus, ultra-high field MRI provides high-resolution maps of various aspects of the underlying neurodegenerative process which reflect individual motor impairment in Parkinson's disease.

Cognitive and brain health in juvenile myoclonic epilepsy: Role of social determinants of health

OBJECTIVE

Juvenile myoclonic epilepsy (JME) is a prevalent genetic generalized epilepsy with linked abnormalities in cognition, behavior, and brain structure. Well recognized is the potential for advancing understanding of the epigenetic contributions to the neurobehavioral complications of JME, but to date there has been no examination of the role of socioeconomic disadvantage in regard to the cognitive and brain health of JME, which is the focus of this investigation.

METHODS

Seventy-seven patients with JME and 44 unrelated controls underwent neuropsychological assessment, structural neuroimaging, and clinical interview to delineate epilepsy history and aspects of family status. The Area Deprivation Index characterized the presence and degree of neighborhood disadvantage, which was examined in relation to cognitive factor scores underlying a comprehensive neuropsychological test battery, academic metrics, integrity of brain structure, and family characteristics.

RESULTS

JME participants resided in neighborhoods associated with significantly more socioeconomic disadvantage, which was associated with significantly poorer performance across all three cognitive factor scores and reading fluency. JME was associated with significant reduction of total subcortical gray matter (GM) but not total cortical gray or white matter volumes. Among controls, participants residing in more advantaged areas exhibited increased volumes of total subcortical GM and diverse subcortical structures as well as areas of increased cortical thickness and volume in frontal/prefrontal regions, findings that were compromised or not evident in JME, raising the possibility of disease-related attenuation of socioeconomic advantage.

SIGNIFICANCE

Socioeconomic disadvantage in JME is associated with adverse effects on cognitive and academic status, whereas socioeconomic advantage in controls is associated with increased brain volumes and thickness, markers of brain health that were largely attenuated or absent in JME. The associations detected here argue for the need to better integrate the social determinants of health with genetic and epigenetic factors in advancing understanding of cognitive and brain health in JME.

MA-SAM: A Multi-Atlas Guided SAM Using Pseudo Mask Prompts Without Manual Annotation for Spine Image Segmentation

Accurate spine segmentation is crucial in clinical diagnosis and treatment of spine diseases. However, due to the complexity of spine anatomical structure, it has remained a challenging task to accurately segment spine images. Recently, the segment anything model (SAM) has achieved superior performance for image segmentation. However, generating high-quality points and boxes is still laborious for high-dimensional medical images. Meanwhile, an accurate mask is difficult to obtain. To address these issues, in this paper, we propose a multi-atlas guided SAM using multiple pseudo mask prompts for spine image segmentation, called MA-SAM. Specifically, we first design a multi-atlas prompt generation sub-network to obtain the anatomical structure prompts. More specifically, we use a network to obtain coarse mask of the input image. Then atlas label maps are registered to the coarse mask. Subsequently, a SAM-based segmentation sub-network is used to segment images. Specifically, we first utilize adapters to fine-tune the image encoder. Meanwhile, we use a prompt encoder to learn the anatomical structure prior knowledge from the multi-atlas prompts. Finally, a mask decoder is used to fuse the image and prompt features to obtain the segmentation results. Moreover, to boost the segmentation performance, different scale features from the prompt encoder are concatenated to the Upsample Block in the mask decoder. We validate our MA-SAM on the two spine segmentation tasks, including spine anatomical structure segmentation with CT images and lumbosacral plexus segmentation with MR images. Experiment results suggest that our method achieves better segmentation performance than SAM with points, boxes, and mask prompts.

Bone-wise rigid registration of femur, tibia, and fibula for the tracking of temporal changes

BACKGROUND

Multiple myeloma (MM) induces temporal alterations in bone structure, such as osteolytic bone lesions, which are challenging to identify through manual image interpretation. The large variation in radiologists' assessments, even at expert centers, further complicates diagnosis. Automatic image analysis methods, including segmentation and registration, can expedite detecting and tracking these bone changes.

PURPOSE

This study presents an automated pipeline for accurately tracking temporal changes in the femurs, tibiae, and fibulae of MM patients using 3D whole-body CT images. The pipeline leverages image segmentation, rigid registration, and temporal subtraction to accelerate disease monitoring and support clinical decision-making.

METHODS

A convolutional neural network (CNN) was trained to segment bones in 3D CT images of 30 MM patients. Nine patients with pre- and post-diagnosis CT scans were used to validate the segmentation and registration process. A two-phase bone-wise rigid registration was applied, followed by temporal subtraction to generate difference images. Segmentation and registration accuracy were assessed using the Dice similarity coefficient (DSC) and mean surface distance (MSD). The proposed method was compared to a non-rigid registration method.

RESULTS

The neural network segmentation resulted in a mean DSC of 0.93 across all bone types and all test data. The registration accuracy measured by the mean DSC across the test data was at least 0.94 for all bone types. The second phase of rigid registration improved the registration fibulae. Metric-wise, the nonrigid method performed better but diminished lesion visibility in difference images.

CONCLUSIONS

An automated pipeline for the longitudinal tracking of bone alterations was presented. Both segmentation and registration demonstrated high accuracy as measured by DSC and MSD. In the difference images produced by temporal subtraction, osteolytic lesions were clearly visible in the femurs. The methodology lays a solid foundation for future improvements, such as inclusion of the axial spine.

Understanding heterogeneity in psychiatric disorders: A method for identifying subtypes and parsing comorbidity

AIM

Most psychiatric and neurodevelopmental disorders are heterogeneous. Neural abnormalities in patients might differ in magnitude and kind, giving rise to distinct subtypes that can be partly overlapping (comorbidity). Identifying disorder-related individual differences is challenging due to the overwhelming presence of disorder-unrelated variation shared with healthy controls. Recently, Contrastive Variational Autoencoders (CVAEs) have been shown to separate disorder-related individual variation from disorder-unrelated variation. However, it is not known if CVAEs can also satisfy the other key desiderata for psychiatric research: capturing disease subtypes and disentangling comorbidity. In this paper, we compare CVAEs to other methods as a function of hyperparameters, such as model size and training data availability. We also introduce a new architecture for modeling comorbid disorders and test a novel training procedure for CVAEs that improves their reproducibility.

METHODS

We use synthetic neuroanatomical MRI data with known ground truth for shared and disorder-specific effects and study the performance of the CVAE and non-contrastive baseline models at detecting disorder-subtypes and disentangling comorbidity in brain images varying along shared and disorder-specific dimensions.

RESULTS

CVAE models consistently outperformed non-contrastive alternatives as measured by correlation with disorder-specific ground truth effects and accuracy of subtype discovery. The CVAE also successfully disentangled neuroanatomical loci of comorbid disorders, due to its novel architecture. Improved training procedure reduced variability in the results by up to 5.5×.

CONCLUSION

The results showcase how the CVAE can be used as an overall framework in precision psychiatry studies, enabling reliable detection of interpretable neuromarkers, discovering disorder subtypes and disentangling comorbidity.

Unraveling the major role of vascular (R2') contributions to R2* signal relaxation at ultra-high-field MRI: A comprehensive analysis with quantitative gradient recalled echo in mouse brain

PURPOSE

Ultra-high-field (UHF) R2* relaxometry is often used for in vivo analysis of biological tissue microstructure without accounting for vascular contributions to R2* signal, that is, the BOLD signal component, and magnetic field inhomogeneities. These effects are especially important at UHF as their contribution to R2* scales linearly with magnetic field. Our study aims to report on the results of separate contributions of R2t* (tissue-specific sub-component) and R2' (vascular BOLD sub-component), corrected for the adverse effects of magnetic field inhomogeneities, to the total R2* signal at in vivo UHF MRI of mouse brain.

METHODS

Four healthy, 8-week-old C57BL/6J mice were imaged in vivo with multi-gradient echo MRI at 9.4 T and analyzed using the quantitative gradient recalled echo (qGRE) approach. A segmentation protocol was established using the Dorr Mouse Brain Atlas and ANTs Syn registration to warp template brain region labels to subject qGRE maps.

RESULTS

By separating R2' contribution from R2* signal, we have established normative R2t* data in mouse brain. Our findings revealed significant contributions of R2' to R2*, with approximately 42% of the R2* signal arising from vascular contributions, thus suggesting the R2t* as a more accurate metric for quantifying tissue microstructural information and its changes in neurodegenerative diseases.

CONCLUSION

qGRE approach allows efficient separation of tissue microstructure-specific (R2t*), vascular BOLD (R2'), and background gradients contributions to the total R2* relaxation at UHF MRI. Due to low concentration of non-heme iron in mouse brain, major contribution to R2t* results from tissue cellular components.

Modular strategies for spatial mapping of diverse cell type data of the mouse brain

Large-scale, international collaborative efforts by members of the BRAIN Initiative Cell Census Network (BICCN) consortium are aggregating the most comprehensive reference database to date for diverse cell type profiling of the mouse brain, which encompasses over 40 different multi-modal profiling techniques from more than 30 research groups. One central challenge for this integrative effort has been the need to map these unique datasets into common reference spaces such that the spatial, structural, and functional information from different cell types can be jointly analyzed. However, significant variation in the acquisition, tissue processing, and imaging techniques across data types makes mapping such diverse data a multifarious problem. Different data types exhibit unique tissue distortion and signal characteristics that precludes a single mapping strategy from being generally applicable across all cell type data. Tailored mapping approaches are often needed to address the unique barriers present in each modality. This work highlights modular atlas mapping strategies developed across separate BICCN studies using the Advanced Normalization Tools Ecosystem (ANTsX) to map spatial transcriptomic (MERFISH) and high-resolution morphology (fMOST) mouse brain data into the Allen Common Coordinate Framework (AllenCCFv3), and developmental (MRI and LSFM) data into the Developmental Common Coordinate Framework (DevCCF). We discuss common mapping strategies that can be shared across modalities and driven by specific challenges from each data type. These mapping strategies include novel open-source contributions that are made publicly available through ANTSX. These include 1) a velocity flow-based approach for continuously mapping developmental trajectories such as that characterizing the DevCCF and 2) an automated framework for determining structural morphology solely through the leveraging of publicly resources. Finally, we provide general guidance to aid investigators to tailor these strategies to address unique data challenges without the need to develop additional specialized software.

Personalized predictions of Glioblastoma infiltration: Mathematical models, Physics-Informed Neural Networks and multimodal scans

Predicting the infiltration of Glioblastoma (GBM) from medical MRI scans is crucial for understanding tumor growth dynamics and designing personalized radiotherapy treatment plans. Mathematical models of GBM growth can complement the data in the prediction of spatial distributions of tumor cells. However, this requires estimating patient-specific parameters of the model from clinical data, which is a challenging inverse problem due to limited temporal data and the limited time between imaging and diagnosis. This work proposes a method that uses Physics-Informed Neural Networks (PINNs) to estimate patient-specific parameters of a reaction-diffusion partial differential equation (PDE) model of GBM growth from a single 3D structural MRI snapshot. PINNs embed both the data and the PDE into a loss function, thus integrating theory and data. Key innovations include the identification and estimation of characteristic non-dimensional parameters, a pre-training step that utilizes the non-dimensional parameters and a fine-tuning step to determine the patient specific parameters. Additionally, the diffuse-domain method is employed to handle the complex brain geometry within the PINN framework. The method is validated on both synthetic and patient datasets, showing promise for personalized GBM treatment through parametric inference within clinically relevant timeframes.

Uncinate Fasciculus Lesion Burden and Anxiety in Multiple Sclerosis

Importance

Multiple sclerosis (MS) is an immune-mediated neurological disorder that affects 2.4 million people worldwide, and up to 60% experience anxiety.

Objective

To investigate whether anxiety in MS is associated with white matter lesion burden in the uncinate fasciculus (UF).

Design, Setting, and Participants

This was a retrospective case-control study of participants aged 18 years or older diagnosed with MS by an MS specialist and identified from the electronic medical record at a single-center academic medical specialty MS clinic in Pennsylvania. Participants received research-quality 3-Tesla magnetic resonance neuroimaging as part of MS clinical care from January 6, 2010, to February 14, 2018. After excluding participants with poor image quality, participants were stratified into 3 groups naturally balanced in age and sex: (1) MS without anxiety, (2) MS with mild anxiety, and (3) MS with severe anxiety. Analyses were performed from June 1 to September 30, 2024.

Exposure

Anxiety diagnosis and anxiolytic medication.

Main Outcomes and Measures

Main outcomes were whether patients with severe anxiety had greater lesion burden in the UF than those without anxiety and whether higher anxiety severity was associated with greater UF lesion burden. Generalized additive models were used, with the burden of lesions (eg, proportion of fascicle impacted) within the UF as the outcome measure and sex, spline of age, and total brain volume as covariates.

Results

Among 372 patients with MS (mean [SD] age, 47.7 [11.4] years; 296 [80%] female), after anxiety phenotype stratification, 99 (27%) had no anxiety (mean [SD] age, 49.4 [11.7] years; 74 [75%] female), 249 (67%) had mild anxiety (mean [SD] age, 47.1 [11.1] years; 203 [82%] female), and 24 (6%) had severe anxiety (mean [SD] age, 47.0 [12.2] years; 19 [79%] female). UF burden was higher in patients with severe anxiety compared with no anxiety (T = 2.01 [P = .047]; Cohen f2, 0.19 [95% CI, 0.08-0.52]). Additionally, higher mean UF burden was associated with higher severity of anxiety (T = 2.09 [P = .04]; Cohen f2, 0.10 [95% CI, 0.05-0.21]).

Conclusions and Relevance

In this case-control study of UF lesion burden and anxiety in MS, overall lesion burden in the UF was associated with the presence and severity of anxiety. Future studies linking white matter lesion burden in the UF with treatment prognosis are warranted.

Anatomically distinct regions in the inferior frontal cortex are modulated by task and reading skill

Inferior frontal cortex (IFC) is a critical region for reading and language. This part of the cortex is highly heterogeneous in its structural and functional organization and shows high variability across individuals. Despite decades of research, the relationship between specific IFC regions and reading skill remains unclear. To shed light on the function of IFC in reading, we aim to (1) characterize the functional landscape of text-selective responses in IFC, while accounting for interindividual variability; and (2) examine how text-selective regions in the IFC relate to reading proficiency. To this end, children with a wide range of reading ability (N=66; age 7-14 years, 34 female, 32 male) completed functional MRI scans while performing two tasks on text and non-text visual stimuli. Importantly, both tasks do not explicitly require reading, and can be performed on all visual stimuli. This design allows us to tease apart stimulus-driven responses from task-driven responses and examine where in IFC task and stimulus interact. We were able to identify three anatomically-distinct, text-selective clusters of activation in IFC, in the inferior frontal sulcus (IFS), and dorsal and ventral precentral gyrus (PrG). ​​These three regions showed a strong task effect that was highly specific to text. Furthermore, text-selectivity in the IFS and dorsal PrG was associated with reading proficiency, such that better readers showed higher selectivity to text. These findings suggest that text-selective regions in the IFC are sensitive to both stimulus and task, and highlight the importance of this region for proficient reading.Significance statement The inferior frontal cortex (IFC) is a critical region for language processing, yet despite decades of research, its relationship with reading skill remains unclear. In a group of children with a wide range of reading skills, we were able to identify three anatomically distinct text-selective clusters of activation in the IFC. ​​These regions showed a strong task effect that was highly selective to text. Text-selectivity was positively correlated with reading proficiency, such that better readers showed higher selectivity to text, even in tasks that did not require reading. These findings suggest that multiple text-selective regions within IFC are sensitive to both stimulus and task, and highlight the critical role of IFC for reading proficiency.

TCDE-Net: An unsupervised dual-encoder network for 3D brain medical image registration

Medical image registration is a critical task in aligning medical images from different time points, modalities, or individuals, essential for accurate diagnosis and treatment planning. Despite significant progress in deep learning-based registration methods, current approaches still face considerable challenges, such as insufficient capture of local details, difficulty in effectively modeling global contextual information, and limited robustness in handling complex deformations. These limitations hinder the precision of high-resolution registration, particularly when dealing with medical images with intricate structures. To address these issues, this paper presents a novel registration network (TCDE-Net), an unsupervised medical image registration method based on a dual-encoder architecture. The dual encoders complement each other in feature extraction, enabling the model to effectively handle large-scale nonlinear deformations and capture intricate local details, thereby enhancing registration accuracy. Additionally, the detail-enhancement attention module aids in restoring fine-grained features, improving the network's capability to address complex deformations such as those at gray-white matter boundaries. Experimental results on the OASIS, IXI, and Hammers-n30r95 3D brain MR dataset demonstrate that this method outperforms commonly used registration techniques across multiple evaluation metrics, achieving superior performance and robustness. Our code is available at https://github.com/muzidongxue/TCDE-Net.

Rapid patient-specific neural networks for intraoperative X-ray to volume registration

The integration of artificial intelligence in image-guided interventions holds transformative potential, promising to extract 3D geometric and quantitative information from conventional 2D imaging modalities during complex procedures. Achieving this requires the rapid and precise alignment of 2D intraoperative images (e.g., X-ray) with 3D preoperative volumes (e.g., CT, MRI). However, current 2D/3D registration methods fail across the broad spectrum of procedures dependent on X-ray guidance: traditional optimization techniques require custom parameter tuning for each subject, whereas neural networks trained on small datasets do not generalize to new patients or require labor-intensive manual annotations, increasing clinical burden and precluding application to new anatomical targets. To address these challenges, we present xvr, a fully automated framework for training patient-specific neural networks for 2D/3D registration. xvr uses physics-based simulation to generate abundant high-quality training data from a patient's own preoperative volumetric imaging, thereby overcoming the inherently limited ability of supervised models to generalize to new patients and procedures. Furthermore, xvr requires only 5 min of training per patient, making it suitable for emergency interventions as well as planned procedures. We perform the largest evaluation of a 2D/3D registration algorithm on real X-ray data to date and find that xvr robustly generalizes across a diverse dataset comprising multiple anatomical structures, imaging modalities, and hospitals. Across surgical tasks, xvr achieves submillimeter-accurate registration at intraoperative speeds, improving upon existing methods by an order of magnitude. xvr is released as open-source software freely available at https://github.com/eigenvivek/xvr.

Generative T2*-weighted images as a substitute for true T2*-weighted images on brain MRI in patients with acute stroke

PURPOSE

The purpose of this study was to validate a deep learning algorithm that generates T2*-weighted images from diffusion-weighted (DW) images and to compare its performance with that of true T2*-weighted images for hemorrhage detection on MRI in patients with acute stroke.

MATERIALS AND METHODS

This single-center, retrospective study included DW- and T2*-weighted images obtained less than 48 hours after symptom onset in consecutive patients admitted for acute stroke. Datasets were divided into training (60 %), validation (20 %), and test (20 %) sets, with stratification by stroke type (hemorrhagic/ischemic). A generative adversarial network was trained to produce generative T2*-weighted images using DW images. Concordance between true T2*-weighted images and generative T2*-weighted images for hemorrhage detection was independently graded by two readers into three categories (parenchymal hematoma, hemorrhagic infarct or no hemorrhage), and discordances were resolved by consensus reading. Sensitivity, specificity and accuracy of generative T2*-weighted images were estimated using true T2*-weighted images as the standard of reference.

RESULTS

A total of 1491 MRI sets from 939 patients (487 women, 452 men) with a median age of 71 years (first quartile, 57; third quartile, 81; range: 21-101) were included. In the test set (n = 300), there were no differences between true T2*-weighted images and generative T2*-weighted images for intraobserver reproducibility (κ = 0.97 [95 % CI: 0.95-0.99] vs. 0.95 [95 % CI: 0.92-0.97]; P = 0.27) and interobserver reproducibility (κ = 0.93 [95 % CI: 0.90-0.97] vs. 0.92 [95 % CI: 0.88-0.96]; P = 0.64). After consensus reading, concordance between true T2*-weighted images and generative T2*-weighted images was excellent (κ = 0.92; 95 % CI: 0.91-0.96). Generative T2*-weighted images achieved 90 % sensitivity (73/81; 95 % CI: 81-96), 97 % specificity (213/219; 95 % CI: 94-99) and 95 % accuracy (286/300; 95 % CI: 92-97) for the diagnosis of any cerebral hemorrhage (hemorrhagic infarct or parenchymal hemorrhage).

CONCLUSION

Generative T2*-weighted images and true T2*-weighted images have non-different diagnostic performances for hemorrhage detection in patients with acute stroke and may be used to shorten MRI protocols.

SORDINO for Silent, Sensitive, Specific, and Artifact-Resisting fMRI in awake behaving mice

Blood-oxygenation-level-dependent (BOLD) functional magnetic resonance imaging (fMRI) has revolutionized our understanding of the brain activity landscape, bridging circuit neuroscience in animal models with noninvasive brain mapping in humans. This immensely utilized technique, however, faces challenges such as acoustic noise, electromagnetic interference, motion artifacts, magnetic-field inhomogeneity, and limitations in sensitivity and specificity. Here, we introduce Steady-state On-the-Ramp Detection of INduction-decay with Oversampling (SORDINO), a transformative fMRI technique that addresses these challenges by maintaining a constant total gradient amplitude while acquiring data during continuously changing gradient direction. When benchmarked against conventional fMRI on a 9.4T system, SORDINO is silent, sensitive, specific, and resistant to motion and susceptibility artifacts. SORDINO offers superior compatibility with multimodal experiments and carries novel contrast mechanisms distinct from BOLD. It also enables brain-wide activity and connectivity mapping in awake, behaving mice, overcoming stress- and motion-related confounds that are among the most challenging barriers in current animal fMRI studies.

Radiation-induced white matter dysfunction in patients with nasopharyngeal carcinoma

Radiation-induced structural abnormalities in white matter (WM) have been reported in patients with nasopharyngeal carcinoma (NPC); however, the alterations in functional domain were insufficiently investigated. A total of 111 NPC patients were included and these patients, based on whether completed radiation therapy (RT) or not, were divided into pre-RT (n = 47) and post-RT (n = 64) groups. Functional connectivity strength (FCS) between WM regions (WW-FCS) and between WM and gray matter (GM) regions (GW-FCS) was used to investigate the radiation-induced changes in WM function. Compared with the pre-RT patients, post-RT NPC patients showed decreased WW-FCS in the left superior cerebellar peduncle, right anterior limb of internal capsule, bilateral posterior thalamic radiation, and left tapetum. Compared with the pre-RT patients, post-RT NPC patients showed decreased GW-FCS in the left caudate, bilateral visual cortex, and the right ventral prefrontal cortex. In the post-RT group, the GW-FCS in left visual cortex was negatively correlated with radiation dosage for the brain stem (r = -0.35, p = 0.039), and for the left temporal lobe (r = -0.46, p = 0.0058). The GW-FCS in right visual cortex was negatively correlated with radiation dosage for the left temporal lobe (r = -0.38, p = 0.025). Our findings of decreased WW-FCS and GW-FCS in such brain regions (such as visual cortex, posterior thalamic radiation, and anterior limb of internal capsule, as well as superior cerebellar peduncle) suggest potential functional impairments in visual and motor systems.

Biological subtyping of autism via cross-species fMRI

It is frequently assumed that the phenotypic heterogeneity in autism spectrum disorder reflects underlying pathobiological variation. However, direct evidence in support of this hypothesis is lacking. Here, we leverage cross-species functional neuroimaging to examine whether variability in brain functional connectivity reflects distinct biological mechanisms. We find that fMRI connectivity alterations in 20 distinct mouse models of autism (n=549 individual mice) can be clustered into two prominent hypo- and hyperconnectivity subtypes. We show that these connectivity profiles are linked to distinct signaling pathways, with hypoconnectivity being associated with synaptic dysfunction, and hyperconnectivity reflecting transcriptional and immune-related alterations. Extending these findings to humans, we identify analogous hypo- and hyperconnectivity subtypes in a large, multicenter resting state fMRI dataset of n=940 autistic and n=1036 neurotypical individuals. Remarkably, hypo- and hyperconnectivity autism subtypes are replicable across independent cohorts (accounting for 25.1% of all autism data), exhibit distinct functional network architecture, are behaviorally dissociable, and recapitulate synaptic and immune mechanisms identified in corresponding mouse subtypes. Our cross-species investigation, thus, decodes the heterogeneity of fMRI connectivity in autism into distinct pathway-specific etiologies, offering a new empirical framework for targeted subtyping of autism.

Reproducible Brain Charts: An open data resource for mapping brain development and its associations with mental health

Major mental disorders are increasingly understood as disorders of brain development. Large and heterogeneous samples are required to define generalizable links between brain development and psychopathology. To this end, we introduce the Reproducible Brain Charts (RBC), an open data resource that integrates data from 5 large studies of brain development in youth from three continents (N=6,346; 45% Female). Confirmatory bifactor models were used to create harmonized psychiatric phenotypes that capture major dimensions of psychopathology. Following rigorous quality assurance, neuroimaging data were carefully curated and processed using consistent pipelines in a reproducible manner with DataLad, the Configurable Pipeline for the Analysis of Connectomes (C-PAC), and FreeSurfer. Initial analyses of RBC data emphasize the benefit of careful quality assurance and data harmonization in delineating developmental effects and associations with psychopathology. Critically, all RBC data - including harmonized psychiatric phenotypes, unprocessed images, and fully processed imaging derivatives - are openly shared without a data use agreement via the International Neuroimaging Data-sharing Initiative. Together, RBC facilitates large-scale, reproducible, and generalizable research in developmental and psychiatric neuroscience.

Deep learning-based Intraoperative MRI reconstruction

BACKGROUND

We retrospectively evaluated the quality of deep learning (DL) reconstructions of on-scanner accelerated intraoperative MRI (iMRI) during respective brain tumor surgery.

METHODS

Accelerated iMRI was performed using dual surface coils positioned around the area of resection. A DL model was trained on the fastMRI neuro dataset to mimic the data from the iMRI protocol. The evaluation was performed on imaging material from 40 patients imaged from Nov 1, 2021, to June 1, 2023, who underwent iMRI during tumor resection surgery. A comparative analysis was conducted between the conventional compressed sense (CS) method and the trained DL reconstruction method. Blinded evaluation of multiple image quality metrics was performed by two neuroradiologists and one neurosurgeon using a 1-to-5 Likert scale (1, nondiagnostic; 2, poor; 3, acceptable; 4, good; and 5, excellent), and the favored reconstruction variant.

RESULTS

The DL reconstruction was strongly favored or favored over the CS reconstruction for 33/40, 39/40, and 8/40 of cases for readers 1, 2, and 3, respectively. For the evaluation metrics, the DL reconstructions had a higher score than their respective CS counterparts for 72%, 72%, and 14% of the cases for readers 1, 2, and 3, respectively. Still, the DL reconstructions exhibited shortcomings such as a striping artifact and reduced signal.

CONCLUSION

DL shows promise in allowing for high-quality reconstructions of iMRI. The neuroradiologists noted an improvement in the perceived spatial resolution, signal-to-noise ratio, diagnostic confidence, diagnostic conspicuity, and spatial resolution compared to CS, while the neurosurgeon preferred the CS reconstructions across all metrics.

RELEVANCE STATEMENT

DL shows promise to allow for high-quality reconstructions of iMRI, however, due to the challenging setting of iMRI, further optimization is needed.

KEY POINTS

iMRI is a surgical tool with a challenging image setting. DL allowed for high-quality reconstructions of iMRI. Additional optimization is needed due to the challenging intraoperative setting.

Fc-engineered large molecules targeting blood-brain barrier transferrin receptor and CD98hc have distinct central nervous system and peripheral biodistribution

Blood brain barrier-crossing molecules targeting transferrin receptor (TfR) and CD98 heavy chain (CD98hc) are widely reported to promote enhanced brain delivery of therapeutics. Here, we provide a comprehensive and unbiased biodistribution characterization of TfR and CD98hc antibody transport vehicles (ATVTfR and ATVCD98hc) compared to control IgG. Mouse whole-body tissue clearing reveals distinct organ localization for each molecule. In the brain, ATVTfR and ATVCD98hc achieve enhanced exposure and parenchymal distribution even when brain exposures are matched between ATV and control IgG in bulk tissue. Using a combination of cell sorting and single-cell RNAseq, we reveal that control IgG is nearly absent from parenchymal cells and is distributed primarily to brain perivascular and leptomeningeal cells. In contrast, ATVTfR and ATVCD98hc exhibit broad and unique parenchymal cell-type distribution. Finally, we profile in detail brain region-specific biodistribution of ATVTfR in cynomolgus monkey brain and spinal cord. Taken together, this in-depth multiscale characterization will guide platform selection for therapeutic targets of interest.

Moving Beyond CT Body Composition Analysis: Using Style Transfer for Bringing CT-Based Fully-Automated Body Composition Analysis to T2-Weighted MRI Sequences

OBJECTIVES

Deep learning for body composition analysis (BCA) is gaining traction in clinical research, offering rapid and automated ways to measure body features like muscle or fat volume. However, most current methods prioritize computed tomography (CT) over magnetic resonance imaging (MRI). This study presents a deep learning approach for automatic BCA using MR T2-weighted sequences.

METHODS

Initial BCA segmentations (10 body regions and 4 body parts) were generated by mapping CT segmentations from body and organ analysis (BOA) model to synthetic MR images created using an in-house trained CycleGAN. In total, 30 synthetic data pairs were used to train an initial nnU-Net V2 in 3D, and this preliminary model was then applied to segment 120 real T2-weighted MRI sequences from 120 patients (46% female) with a median age of 56 (interquartile range, 17.75), generating early segmentation proposals. These proposals were refined by human annotators, and nnU-Net V2 2D and 3D models were trained using 5-fold cross-validation on this optimized dataset of real MR images. Performance was evaluated using Sørensen-Dice, Surface Dice, and Hausdorff Distance metrics including 95% confidence intervals for cross-validation and ensemble models.

RESULTS

The 3D ensemble segmentation model achieved the highest Dice scores for the body region classes: bone 0.926 (95% confidence interval [CI], 0.914-0.937), muscle 0.968 (95% CI, 0.961-0.975), subcutaneous fat 0.98 (95% CI, 0.971-0.986), nervous system 0.973 (95% CI, 0.965-0.98), thoracic cavity 0.978 (95% CI, 0.969-0.984), abdominal cavity 0.989 (95% CI, 0.986-0.991), mediastinum 0.92 (95% CI, 0.901-0.936), pericardium 0.945 (95% CI, 0.924-0.96), brain 0.966 (95% CI, 0.927-0.989), and glands 0.905 (95% CI, 0.886-0.921). Furthermore, body part 2D ensemble model reached the highest Dice scores for all labels: arms 0.952 (95% CI, 0.937-0.965), head + neck 0.965 (95% CI, 0.953-0.976), legs 0.978 (95% CI, 0.968-0.988), and torso 0.99 (95% CI, 0.988-0.991). The overall average Dice across body parts (2D = 0.971, 3D = 0.969, P = ns) and body regions (2D = 0.935, 3D = 0.955, P < 0.001) ensemble models indicates stable performance across all classes.

CONCLUSIONS

The presented approach facilitates efficient and automated extraction of BCA parameters from T2-weighted MRI sequences, providing precise and detailed body composition information across various regions and body parts.

Dense convolution-based attention network for Alzheimer's disease classification

Recently, deep learning-based medical image classification models have made substantial advancements. However, many existing models prioritize performance at the cost of efficiency, limiting their practicality in clinical use. Traditional Convolutional Neural Network (CNN)-based methods, Transformer-based methods, and hybrid approaches combining these two struggle to balance performance and model complexity. To achieve efficient predictions with a low parameter count, we propose DenseAttentionNetwork (DANet), a lightweight model for Alzheimer's disease detection in 3D MRI images. DANet leverages dense connections and a linear attention mechanism to enhance feature extraction and capture long-range dependencies. Its architecture integrates convolutional layers for localized feature extraction with linear attention for global context, enabling efficient multi-scale feature reuse across the network. By replacing traditional self-attention with a parameter-efficient linear attention mechanism, DANet overcomes some limitations of standard self-attention. Extensive experiments across multi-institutional datasets demonstrate that DANet achieves the best performance in area under the receiver operating characteristic curve (AUC), which underscores the model's robustness and effectiveness in capturing relevant features for Alzheimer's disease detection while also attaining a strong accuracy structure with fewer parameters. Visualizations based on activation maps further verify the model's ability to highlight AD-relevant regions in 3D MRI images, providing clinically interpretable insights into disease progression.

Left-Right Brain-Wide Asymmetry of Neuroanatomy in the Mouse Brain

Left-right asymmetry of the human brain is widespread through its anatomy and function. However, limited microscopic understanding of it exists, particularly for anatomical asymmetry where there are few well-established animal models. In humans, most brain regions show subtle, population-average regional asymmetries in thickness or surface area, alongside a macro-scale twisting called the cerebral petalia in which the right hemisphere protrudes past the left. Here, we ask whether neuroanatomical asymmetries can be observed in mice, leveraging 6 mouse neuroimaging cohorts from 5 different research groups (∼3,500 animals). We found an anterior-posterior pattern of volume asymmetry with anterior regions larger on the right and posterior regions larger on the left. This pattern appears driven by similar trends in surface area and positional asymmetries, with the results together indicating a small brain-wide twisting pattern, similar to the human cerebral petalia. Furthermore, the results show no apparent relationship to known functional asymmetries in mice, emphasizing the complexity of the structure-function relationship in brain asymmetry. Our results recapitulate and extend previous patterns of asymmetry from two published studies as well as capture well-established, bilateral male-female differences in the mouse brain as a positive control. By establishing a signature of anatomical brain asymmetry in mice, we aim to provide a foundation for future studies to probe the mechanistic underpinnings of brain asymmetry seen in humans - a feature of the brain with extremely limited understanding.

Predicting Post-Operative Side Effects in VIM MRgFUS Based on THalamus Optimized Multi Atlas Segmentation (THOMAS) on White-Matter-Nulled MRI: A Retrospective Study

BACKGROUND AND PURPOSE

Precise and individualized targeting of the ventral intermediate thalamic nucleus for the MR-guided focused ultrasound is crucial for enhancing treatment efficacy and avoiding undesirable side effects. In this study, we tested the hypothesis that the spatial relationships between Thalamus Optimized Multi Atlas Segmentation derived segmentations and the post-focused ultrasound lesion can predict post-operative side effects in patients treated with MR-guided focused ultrasound.

MATERIALS AND METHODS

We retrospectively analyzed 30 patients (essential tremor, n = 26; tremor-dominant Parkinson's disease, n = 4) who underwent unilateral ventral intermediate thalamic nucleus focused ultrasound treatment. We created ROIs of coordinate-based indirect treatment target, focused ultrasound-induced lesion, and thalamus and ventral intermediate thalamic nucleus segmentations. We extracted imaging features including 1) focused ultrasound-induced lesion volumes, 2) overlap between lesions and thalamus and ventral intermediate thalamic nucleus segmentations, 3) distance between lesions and ventral intermediate thalamic nucleus segmentation and 4) distance between lesions and the indirect standard target. These imaging features were compared between patients with and without post-operative gait/balance side effects using Wilcoxon rank-sum test. Multivariate prediction models of side effects based on the imaging features were evaluated using the receiver operating characteristic analyses.

RESULTS

Patients with self-reported gait/balance side effects had a significantly larger extent of focused ultrasound-induced edema, a smaller fraction of the lesion within the ventral intermediate thalamic nucleus segmentation, a larger fraction of the off-target lesion outside the thalamus segmentation, a more inferior centroid of the lesion from the ventral intermediate thalamic nucleus segmentation, and a larger distance between the centroid of the lesion and ventral intermediate thalamic nucleus segmentation (p < 0.05). Similar results were found for exam-based side effects. Multivariate regression models based on the imaging features achieved areas under the curve of 0.99 (95% CI: 0.88 to 1.00) and 0.96 (95% CI: 0.73 to 1.00) for predicting self-reported and exam-based side effects, respectively.

CONCLUSIONS

Thalamus Optimized Multi Atlas Segmentation-based patient-specific segmentation of the ventral intermediate thalamic nucleus can predict post-operative side effects, which has implications for aiding the direct targeting of MR-guided focused ultrasound and reducing side effects.

Calibration-free whole-brain CEST imaging at 7T with parallel transmit pulse design for saturation homogeneity utilizing universal pulses (PUSHUP)

PURPOSE

Chemical exchange saturation transfer (CEST) measurements at ultra-high field (UHF) suffer from strong saturation inhomogeneity. Retrospective correction of this inhomogeneity is possible to some extent, but requires a time-consuming repetition of the measurement. Here, we propose a calibration-free parallel transmit (pTx)-based saturation scheme that homogenizes the saturation over the imaging volume, which we call PUlse design for Saturation Homogeneity utilizing Universal Pulses (PUSHUP).

THEORY

Magnetization transfer effects depend on the saturationB 1 rms $$ {\mathrm{B}}_1^{\mathrm{rms}} $$ . PUSHUP homogenizes the saturationB 1 rms $$ {\mathrm{B}}_1^{\mathrm{rms}} $$ by using multiple saturation pulses with alternatingB 1 $$ {\mathrm{B}}_1 $$ -shims. Using a database ofB 1 $$ {\mathrm{B}}_1 $$ maps, universal pulses are calculated that remove the necessity of time-consuming, subject-based pulse calculation during the measurement.

METHODS

PUSHUP was combined with a whole-brain three-dimensional-echo planar imaging (3D-EPI) readout. Two PUSHUP saturation modules were calculated by either applying whole-brain or cerebellum masks to the database maps. The saturation homogeneity and the group mean CEST amplitudes were calculated for differentB 1 $$ {\mathrm{B}}_1 $$ -correction methods and were compared to circular polarized (CP) saturation in five healthy volunteers using an eight-channel transmit coil at 7 Tesla.

RESULTS

In contrast to CP saturation, where accurate CEST maps were impossible to obtain in the cerebellum, even with extensiveB 1 $$ {\mathrm{B}}_1 $$ -correction, PUSHUP CEST maps were artifact-free throughout the whole brain. A 1-point retrospectiveB 1 $$ {\mathrm{B}}_1 $$ -correction, that does not need repeated measurements, sufficiently removed the effect of residual saturation inhomogeneity.

CONCLUSION

The presented method allows for homogeneous whole-brain CEST imaging at 7 Tesla without the need of a repetition-basedB 1 $$ {\mathrm{B}}_1 $$ -correction or online pulse calculation. With the fast 3D-EPI readout, whole-brain CEST imaging with 45 saturation offsets is possible at 1.6 mm resolution in under 4 min.

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.

Artificial Intelligence Apps for Medical Image Analysis using pyCERR and Cancer Genomics Cloud

This work introduces a user-friendly, cloud-based software framework for conducting Artificial Intelligence (AI) analyses of medical images. The framework allows users to deploy AI-based workflows by customizing software and hardware dependencies. The components of our software framework include the Python-native Computational Environment for Radiological Research (pyCERR) platform for radiological image processing, Cancer Genomics Cloud (CGC) for accessing hardware resources and user management utilities for accessing images from data repositories and installing AI models and their dependencies. GNU-GPL copyright pyCERR was ported to Python from MATLAB-based CERR to enable researchers to organize, access, and transform metadata from high dimensional, multi-modal datasets to build cloud-compatible workflows for AI modeling in radiation therapy and medical image analysis. pyCERR provides an extensible data structure to accommodate metadata from commonly used medical imaging file formats and a viewer to allow for multi-modal visualization. Analysis modules are provided to facilitate cloud-compatible AI-based workflows for image segmentation, radiomics, DCE MRI analysis, radiotherapy dose-volume histogram-based features, and normal tissue complication and tumor control models for radiotherapy. Image processing utilities are provided to help train and infer convolutional neural network-based models for image segmentation, registration and transformation. The framework allows for round-trip analysis of imaging data, enabling users to apply AI models to their images on CGC and retrieve and review results on their local machine without requiring local installation of specialized software or GPU hardware. The deployed AI models can be accessed using APIs provided by CGC, enabling their use in a variety of programming languages. In summary, the presented framework facilitates end-to-end radiological image analysis and reproducible research, including pulling data from sources, training or inferring from an AI model, utilities for data management, visualization, and simplified access to image metadata.

Small or absent Visual Word Form Area is a trait of dyslexia

Understanding the balance between plastic and persistent traits in the dyslexic brain is critical for developing effective interventions. This longitudinal intervention study examines the Visual Word Form Area (VWFA) in dyslexic and typical readers, exploring how this key component of the brain's reading circuitry changes with learning. We found that dyslexic readers show significant differences in VWFA presence, size, and tuning properties compared to typical readers. While reading intervention improved reading skills and increased VWFA size, disparities persisted, suggesting that VWFA abnormalities are an enduring trait of dyslexia. Notably, we found that even with sufficient intervention to close the reading skill gap, dyslexic readers are still expected to have smaller VWFAs. Our results reveal intervention-driven long-term neural and behavioral changes, while also elucidating stable differences in the functional architecture of the dyslexic brain. This provides new insights into the potential and limitations of learning-induced plasticity in the human visual cortex.

Brain volumes, cognitive, and adaptive skills in school-age children with Down syndrome

BACKGROUND

Down syndrome (DS) is the most common congenital neurodevelopmental disorder, present in about 1 in every 700 live births. Despite its prevalence, literature exploring the neurobiology underlying DS and how this neurobiology is related to behavior is limited. This study fills this gap by examining cortical volumes and behavioral correlates in school-age children with DS.

METHODS

School-age children (mean = 9.7 years ± 1.1) underwent comprehensive assessments, including cognitive and adaptive assessments, as well as an MRI scan without the use of sedation. Children with DS (n = 35) were compared to available samples of typically developing (TD; n = 80) and ASD children (n = 29). ANOVAs were conducted to compare groups on cognitive and adaptive assessments. ANCOVAs (covarying for age, sex, and total cerebral volume; TCV) compared cortical brain volumes between groups. Correlations between behavioral metrics and cortical and cerebellar volumes (separately for gray (GM) and white matter (WM)) were conducted separately by group.

RESULTS

As expected, children with DS had significantly lower cognitive skills compared to ASD and TD children. Daily Living adaptive skills were comparable between ASD children and children with DS, and both groups scored lower than TD children. Children with DS exhibited a smaller TCV compared to ASD and TD children. Additionally, when controlling for TCV, age, and sex, children with DS had significantly smaller total GM and tissue volumes. Cerebellum volumes were significantly correlated with Daily Living adaptive behaviors in the DS group only.

CONCLUSIONS

Despite children with DS exhibiting lower cognitive skills and smaller brain volume overall than children with ASD, their deficits in Socialization and Daily Living adaptive skills are comparable. Differences in lobar volumes (e.g., Right Frontal GM/WM, Left Frontal WM, and Left and Right Temporal WM) were observed above and beyond overall differences in total volume. The correlation between cerebellum volumes and Daily Living adaptive behaviors in the DS group provides a novel area to explore in future research.

eLemur: A cellular-resolution 3D atlas of the mouse lemur brain

The gray mouse lemur (Microcebus murinus), one of the smallest living primates, emerges as a promising model organism for neuroscience research. This is due to its genetic similarity to humans, its evolutionary position between rodents and humans, and its primate-like features encapsulated within a rodent-sized brain. Despite its potential, the absence of a comprehensive reference brain atlas impedes the progress of research endeavors in this species, particularly at the microscopic level. Existing references have largely been confined to the macroscopic scale, lacking detailed anatomical information. Here, we present eLemur, a unique resource, comprising a repository of high-resolution brain-wide images immunostained with multiple cell type and structural markers, elucidating the cyto- and chemoarchitecture of the mouse lemur brain. Additionally, it encompasses a segmented two-dimensional reference and 3D anatomical brain atlas delineated into cortical, subcortical, and other vital regions. Furthermore, eLemur includes a comprehensive 3D cell atlas, providing densities and spatial distributions of non-neuronal and neuronal cells across the mouse lemur brain. Accessible via a web-based viewer (https://eeum-brain.com/#/lemurdatasets), the eLemur resource streamlines data sharing and integration, fostering the exploration of different hypotheses and experimental designs using the mouse lemur as a model organism. Moreover, in conjunction with the growing 3D datasets for rodents, nonhuman primates, and humans, our eLemur 3D digital framework enhances the potential for comparative analysis and translation research, facilitating the integration of extensive rodent study data into human studies.

Rapid submillimeter QSM and R2* mapping using interleaved multishot 3D-EPI at 7 and 3 Tesla

PURPOSE

To explore the high signal-to-noise ratio (SNR) efficiency of interleaved multishot 3D-EPI with standard image reconstruction for fast and robust high-resolution whole-brain quantitative susceptibility (QSM) andR 2 ∗ $$ {R}_2^{\ast } $$ mapping at 7 and 3T.

METHODS

Single- and multi-TE segmented 3D-EPI is combined with conventional CAIPIRINHA undersampling for up to 72-fold effective gradient echo (GRE) imaging acceleration. Across multiple averages, scan parameters are varied (e.g., dual-polarity frequency-encoding) to additionally correct forB 0 $$ {\mathrm{B}}_0 $$ -induced artifacts, geometric distortions and motion retrospectively. A comparison to established GRE protocols is made. Resolutions range from 1.4 mm isotropic (1 multi-TE average in 36 s) up to 0.4 mm isotropic (2 single-TE averages in approximately 6 min) with whole-head coverage.

RESULTS

Only 1-4 averages are needed for sufficient SNR with 3D-EPI, depending on resolution and field strength. Fast scanning and small voxels together with retrospective corrections result in substantially reduced image artifacts, which improves susceptibility andR 2 ∗ $$ {R}_2^{\ast } $$ mapping. Additionally, much finer details are obtained in susceptibility-weighted image projections through significantly reduced partial voluming.

CONCLUSION

Using interleaved multishot 3D-EPI, single-TE and multi-TE data can readily be acquired 10 times faster than with conventional, accelerated GRE imaging. Even 0.4 mm isotropic whole-head QSM within 6 min becomes feasible at 7T. At 3T, motion-robust 0.8 mm isotropic whole-brain QSM andR 2 ∗ $$ {R}_2^{\ast } $$ mapping with no apparent distortion in less than 7 min becomes clinically feasible. Stronger gradient systems may allow for even higher effective acceleration rates through larger EPI factors while maintaining optimal contrast.

Molecular signatures of cortical expansion in the human foetal brain

The third trimester of human gestation is characterised by rapid increases in brain volume and cortical surface area. Recent studies have revealed a remarkable molecular diversity across the prenatal cortex but little is known about how this diversity translates into the differential rates of cortical expansion observed during gestation. We present a digital resource, μBrain, to facilitate knowledge translation between molecular and anatomical descriptions of the prenatal brain. Using μBrain, we evaluate the molecular signatures of preferentially-expanded cortical regions, quantified in utero using magnetic resonance imaging. Our findings demonstrate a spatial coupling between areal differences in the timing of neurogenesis and rates of neocortical expansion during gestation. We identify genes, upregulated from mid-gestation, that are highly expressed in rapidly expanding neocortex and implicated in genetic disorders with cognitive sequelae. The μBrain atlas provides a tool to comprehensively map early brain development across domains, model systems and resolution scales.

Improving MR axon radius estimation in human white matter using spiral acquisition and field monitoring

PURPOSE

To compare MR axon radius estimation in human white matter using a multiband spiral sequence combined with field monitoring to the current state-of-the-art echo-planar imaging (EPI)-based approach.

METHODS

A custom multiband spiral sequence was used for diffusion-weighted imaging at ultra-highb $$ b $$ -values. Field monitoring and higher order image reconstruction were employed to greatly reduce artifacts in spiral images. Diffusion weighting parameters were chosen to match a state-of-the art EPI-based axon radius mapping protocol. The spiral approach was compared to the EPI approach by comparing the image signal-to-noise ratio (SNR) and performing a test-retest study to assess the respective variability and repeatability of axon radius mapping. Effective axon radius estimates were compared over white matter voxels and along the left corticospinal tract.

RESULTS

Increased SNR and reduced artifacts in spiral images led to reduced variability in resulting axon radius maps, especially in low-SNR regions. Test-retest variability was reduced by a factor of approximately 1.5 using the spiral approach. Reduced repeatability due to significant bias was found for some subjects in both spiral and EPI approaches, and attributed to scanner instability, pointing to a previously unknown limitation of the state-of-the-art approach.

CONCLUSION

Combining spiral readouts with field monitoring improved mapping of the effective axon radius compared to the conventional EPI approach.

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.

ATOMMIC: An Advanced Toolbox for Multitask Medical Imaging Consistency to facilitate Artificial Intelligence applications from acquisition to analysis in Magnetic Resonance Imaging

BACKGROUND AND OBJECTIVES

Artificial intelligence (AI) is revolutionizing Magnetic Resonance Imaging (MRI) along the acquisition and processing chain. Advanced AI frameworks have been applied in various successive tasks, such as image reconstruction, quantitative parameter map estimation, and image segmentation. However, existing frameworks are often designed to perform tasks independently of each other or are focused on specific models or single datasets, limiting generalization. This work introduces the Advanced Toolbox for Multitask Medical Imaging Consistency (ATOMMIC), a novel open-source toolbox that streamlines AI applications for accelerated MRI reconstruction and analysis. ATOMMIC implements several tasks using deep learning (DL) models and enables MultiTask Learning (MTL) to perform related tasks in an integrated manner, targeting generalization in the MRI domain.

METHODS

We conducted a comprehensive literature review and analyzed 12,479 GitHub repositories to assess the current landscape of AI frameworks for MRI. Subsequently, we demonstrate how ATOMMIC standardizes workflows and improves data interoperability, enabling effective benchmarking of various DL models across MRI tasks and datasets. To showcase ATOMMIC's capabilities, we evaluated twenty-five DL models on eight publicly available datasets, focusing on accelerated MRI reconstruction, segmentation, quantitative parameter map estimation, and joint accelerated MRI reconstruction and segmentation using MTL.

RESULTS

ATOMMIC's high-performance training and testing capabilities, utilizing multiple GPUs and mixed precision support, enable efficient benchmarking of multiple models across various tasks. The framework's modular architecture implements each task through a collection of data loaders, models, loss functions, evaluation metrics, and pre-processing transformations, facilitating seamless integration of new tasks, datasets, and models. Our findings demonstrate that ATOMMIC supports MTL for multiple MRI tasks with harmonized complex-valued and real-valued data support while maintaining active development and documentation. Task-specific evaluations demonstrate that physics-based models outperform other approaches in reconstructing highly accelerated acquisitions. These high-quality reconstruction models also show superior accuracy in estimating quantitative parameter maps. Furthermore, when combining high-performing reconstruction models with robust segmentation networks through MTL, performance is improved in both tasks.

CONCLUSIONS

ATOMMIC advances MRI reconstruction and analysis by leveraging MTL and ensuring consistency across tasks, models, and datasets. This comprehensive framework serves as a versatile platform for researchers to use existing AI methods and develop new approaches in medical imaging.

Prediction, prognosis and monitoring of neurodegeneration at biobank-scale via machine learning and imaging

Background

Alzheimer's disease and related dementias (ADRD) and Parkinson's disease (PD) are the most common neurodegenerative conditions. These central nervous system disorders impact both the structure and function of the brain and may lead to imaging changes that precede symptoms. Patients with ADRD or PD have long asymptomatic phases that exhibit significant heterogeneity. Hence, quantitative measures that can provide early disease indicators are necessary to improve patient stratification, clinical care, and clinical trial design. This work uses machine learning techniques to derive such a quantitative marker from T1-weighted (T1w) brain Magnetic resonance imaging (MRI).

Methods

In this retrospective study, we developed machine learning (ML) based disease-specific scores based on T1w brain MRI utilizing Parkinson's Disease Progression Marker Initiative (PPMI) and Alzheimer's Disease Neuroimaging Initiative (ADNI) cohorts. We evaluated the potential of ML-based scores for early diagnosis, prognosis, and monitoring of ADRD and PD in an independent large-scale population-based longitudinal cohort, UK Biobank.

Findings

1,826 dementia images from 731 participants, 3,161 healthy control images from 925 participants from the ADNI cohort, 684 PD images from 319 participants, and 232 healthy control images from 145 participants from the PPMI cohort were used to train machine learning models. The classification performance is 0.94 [95% CI: 0.93-0.96] area under the ROC Curve (AUC) for ADRD detection and 0.63 [95% CI: 0.57-0.71] for PD detection using 790 extracted structural brain features. The most predictive regions include the hippocampus and temporal brain regions in ADRD and the substantia nigra in PD. The normalized ML model's probabilistic output (ADRD and PD imaging scores) was evaluated on 42,835 participants with imaging data from the UK Biobank. There are 66 cases for ADRD and 40 PD cases whose T1 brain MRI is available during pre-diagnostic phases. For diagnosis occurrence events within 5 years, the integrated survival model achieves a time-dependent AUC of 0.86 [95% CI: 0.80-0.92] for dementia and 0.89 [95% CI: 0.85-0.94] for PD. ADRD imaging score is strongly associated with dementia-free survival (hazard ratio (HR) 1.76 [95% CI: 1.50-2.05] per S.D. of imaging score), and PD imaging score shows association with PD-free survival (hazard ratio 2.33 [95% CI: 1.55-3.50]) in our integrated model. HR and prevalence increased stepwise over imaging score quartiles for PD, demonstrating heterogeneity. As a proxy for diagnosis, we validated AD/PD polygenic risk scores of 42,835 subjects against the imaging scores, showing a highly significant association after adjusting for covariates. In both the PPMI and ADNI cohorts, the scores are associated with clinical assessments, including the Mini-Mental State Examination (MMSE), Alzheimer's Disease Assessment Scale-cognitive subscale (ADAS-Cog), and pathological markers, which include amyloid and tau. Finally, imaging scores are associated with polygenic risk scores for multiple diseases. Our results suggest that we can use imaging scores to assess the genetic architecture of such disorders in the future.

Interpretation

Our study demonstrates the use of quantitative markers generated using machine learning techniques for ADRD and PD. We show that disease probability scores obtained from brain structural features are useful for early detection, prognosis prediction, and monitoring disease progression. To facilitate community engagement and external tests of model utility, an interactive app to explore summary level data from this study and dive into external data can be found here https://ndds-brainimaging-ml.streamlit.app. As far as we know, this is the first publicly available cloud-based MRI prediction application.

Funding

US National Institute on Aging, and US National Institutes of Health.

Developmental mouse brain common coordinate framework

3D brain atlases are key resources to understand the brain's spatial organization and promote interoperability across different studies. However, unlike the adult mouse brain, the lack of developing mouse brain 3D reference atlases hinders advancements in understanding brain development. Here, we present a 3D developmental common coordinate framework (DevCCF) spanning embryonic day (E)11.5, E13.5, E15.5, E18.5, and postnatal day (P)4, P14, and P56, featuring undistorted morphologically averaged atlas templates created from magnetic resonance imaging and co-registered high-resolution light sheet fluorescence microscopy templates. The DevCCF with 3D anatomical segmentations can be downloaded or explored via an interactive 3D web-visualizer. As a use case, we utilize the DevCCF to unveil GABAergic neuron emergence in embryonic brains. Moreover, we map the Allen CCFv3 and spatial transcriptome cell-type data to our stereotaxic P56 atlas. In summary, the DevCCF is an openly accessible resource for multi-study data integration to advance our understanding of brain development.

Association of Anxiety with Uncinate Fasciculus Lesion Burden in Multiple Sclerosis

Importance

Multiple sclerosis (MS) is an immune-mediated neurological disorder that affects 2.4 million people world-wide, and up to 60% experience anxiety.

Objective

We investigated how anxiety in MS is associated with white matter lesion burden in the uncinate fasciculus (UF).

Design

Retrospective case-control study of participants who received research-quality 3-tesla (3T) neuroimaging as part of MS clinical care from 2010-2018. Analyses were performed from June 1st to September 30th, 2024.

Setting

Single-center academic medical specialty MS clinic.

Participants

Participants were identified from the electronic medical record. All participants were diagnosed by an MS specialist and completed research-quality MRI at 3T. After excluding participants with poor image quality, 372 were stratified into three groups which were balanced for age and sex: 1) MS without anxiety (MS+noA, n=99); 2) MS with mild anxiety (MS+mildA, n=249); and 3) MS with severe anxiety (MS+severeA, n=24).

Exposure

Anxiety diagnosis and anxiolytic medication.

Main Outcome and Measure

We first evaluated whether MS+severeA patients had greater lesion burden in the UF than MS+noA. Next, we examined whether increasing anxiety severity was associated with greater UF lesion burden. Generalized additive models were employed, with the burden of lesions (e.g. proportion of fascicle impacted) within the UF as the outcome measure and sex and spline of age as covariates.

Results

UF burden was higher in MS+severeA as compared to MS+noA (T=2.02, P=0.045, Cohen's f 2=0.19). A dose-response effect was also found, where higher mean UF burden was associated with higher anxiety severity (T=2.08, P=0.038, Cohen's f 2=0.10).

Conclusions and Relevance

We demonstrate that overall lesion burden in UF was associated with the presence and severity of anxiety in patients with MS. Future studies linking white matter lesion burden in UF with treatment prognosis are warranted.

Investigating Mood and Cognition in People with Multiple Sclerosis: A Prospective Study Protocol

Multiple sclerosis (MS) is an immune-mediated neurological disorder that affects one million people in the United States. Up to 50% of people with MS experience depression, yet the mechanisms of depression in MS remain under-investigated. Studies of medically healthy participants with depression have described associations between white matter variability and depressive symptoms, but frequently exclude participants with medical comorbidities and thus cannot be extrapolated to people with intracranial diseases. White matter lesions are a key pathologic feature of MS and could disrupt pathways involved in depression symptoms. The purpose of this study is to investigate the impact of brain network disruption on depression using MS as a model. We will obtain structured clinical and cognitive assessments from two hundred fifty participants with MS and prospectively evaluate white matter lesion burden as a predictor of depressive symptoms. Ethics approval was obtained from The University of Pennsylvania Institutional Review Board (Protocol #853883). The results of this study will be presented at scientific meetings and conferences and published in peer-reviewed journals.

Medial prefrontal neuroplasticity during extended-release naltrexone treatment of opioid use disorder - a longitudinal structural magnetic resonance imaging study

Opioid use disorder (OUD) has been linked to macroscopic structural alterations in the brain. The monthly injectable, extended-release formulation of μ-opioid antagonist naltrexone (XR-NTX) is highly effective in reducing opioid craving and preventing opioid relapse. Here, we investigated the neuroanatomical effects of XR-NTX by examining changes in cortical thickness during treatment for OUD. Forty-seven OUD patients underwent structural magnetic resonance imaging and subjectively rated their opioid craving ≤1 day before (pre-treatment) and 11 ± 3 days after (on-treatment) the first XR-NTX injection. A sample of fifty-six non-OUD individuals completed a single imaging session and served as the comparison group. A publicly available [¹¹C]carfentanil positron emission tomography dataset was used to assess the relationship between changes in cortical thickness and μ-opioid receptor (MOR) binding potential across brain regions. We found that the thickness of the medial prefrontal and anterior cingulate cortices (mPFC/aCC; regions with high MOR binding potential) was comparable between the non-OUD individuals and the OUD patients at pre-treatment. However, among the OUD patients, mPFC/aCC thickness significantly decreased from pre-treatment to on-treatment. A greater reduction in mPFC/aCC thickness was associated with a greater reduction in opioid craving. Taken together, our study suggests XR-NTX-induced cortical thickness reduction in the mPFC/aCC regions in OUD patients. The reduction in thickness does not appear to indicate a restoration to the non-OUD level but rather reflects XR-NTX's distinct therapeutic impact on an MOR-rich brain structure. Our findings highlight the neuroplastic effects of XR-NTX that may inform the development of novel OUD interventions.

Improved Dementia Prediction in Cerebral Small Vessel Disease Using Deep Learning-Derived Diffusion Scalar Maps From T1

BACKGROUND

Cerebral small vessel disease is the most common pathology underlying vascular dementia. In small vessel disease, diffusion tensor imaging is more sensitive to white matter damage and better predicts dementia risk than conventional magnetic resonance imaging sequences, such as T1 and fluid attenuation inversion recovery, but diffusion tensor imaging takes longer to acquire and is not routinely available in clinical practice. As diffusion tensor imaging-derived scalar maps-fractional anisotropy (FA) and mean diffusivity (MD)-are frequently used in clinical settings, one solution is to synthesize FA/MD from T1 images.

METHODS

We developed a deep learning model to synthesize FA/MD from T1. The training data set consisted of 4998 participants with the highest white matter hyperintensity volumes in the UK Biobank. Four external validations data sets with small vessel disease were included: SCANS (St George's Cognition and Neuroimaging in Stroke; n=120), RUN DMC (Radboud University Nijmegen Diffusion Tensor and Magnetic Resonance Imaging Cohort; n=502), PRESERVE (Blood Pressure in Established Cerebral Small Vessel Disease; n=105), and NETWORKS (n=26), along with 1000 normal controls from the UK Biobank.

RESULTS

The synthetic maps resembled ground-truth maps (structural similarity index >0.89 for MD maps and >0.80 for FA maps across all external validation data sets except for SCANS). The prediction accuracy of dementia using whole-brain median MD from the synthetic maps is comparable to the ground truth (SCANS ground-truth c-index, 0.822 and synthetic, 0.821; RUN DMC ground truth, 0.816 and synthetic, 0.812) and better than white matter hyperintensity volume (SCANS, 0.534; RUN DMC, 0.710).

CONCLUSIONS

We have developed a fast and generalizable method to synthesize FA/MD maps from T1 to improve the prediction accuracy of dementia in small vessel disease when diffusion tensor imaging data have not been acquired.

HaN-Seg: The head and neck organ-at-risk CT and MR segmentation challenge

BACKGROUND AND PURPOSE

To promote the development of auto-segmentation methods for head and neck (HaN) radiation treatment (RT) planning that exploit the information of computed tomography (CT) and magnetic resonance (MR) imaging modalities, we organized HaN-Seg: The Head and Neck Organ-at-Risk CT and MR Segmentation Challenge.

MATERIALS AND METHODS

The challenge task was to automatically segment 30 organs-at-risk (OARs) of the HaN region in 14 withheld test cases given the availability of 42 publicly available training cases. Each case consisted of one contrast-enhanced CT and one T1-weighted MR image of the HaN region of the same patient, with up to 30 corresponding reference OAR delineation masks. The performance was evaluated in terms of the Dice similarity coefficient (DSC) and 95-percentile Hausdorff distance (HD95), and statistical ranking was applied for each metric by pairwise comparison of the submitted methods using the Wilcoxon signed-rank test.

RESULTS

While 23 teams registered for the challenge, only seven submitted their methods for the final phase. The top-performing team achieved a DSC of 76.9 % and a HD95 of 3.5 mm. All participating teams utilized architectures based on U-Net, with the winning team leveraging rigid MR to CT registration combined with network entry-level concatenation of both modalities.

CONCLUSION

This challenge simulated a real-world clinical scenario by providing non-registered MR and CT images with varying fields-of-view and voxel sizes. Remarkably, the top-performing teams achieved segmentation performance surpassing the inter-observer agreement on the same dataset. These results set a benchmark for future research on this publicly available dataset and on paired multi-modal image segmentation in general.

Uncovering the hidden effects of repetitive subconcussive head impact exposure: A mega-analytic approach characterizing seasonal brain microstructural changes in contact and collision sports athletes

Repetitive subconcussive head impacts (RSHI) are believed to induce sub-clinical brain injuries, potentially resulting in cumulative, long-term brain alterations. This study explores patterns of longitudinal brain white matter changes across sports with RSHI-exposure. A systematic literature search identified 22 datasets with longitudinal diffusion magnetic resonance imaging data. Four datasets were centrally pooled to perform uniform quality control and data preprocessing. A total of 131 non-concussed active athletes (American football, rugby, ice hockey; mean age: 20.06 ± 2.06 years) with baseline and post-season data were included. Nonparametric permutation inference (one-sample t tests, one-sided) was applied to analyze the difference maps of multiple diffusion parameters. The analyses revealed widespread lateralized patterns of sports-season-related increases and decreases in mean diffusivity (MD), radial diffusivity (RD), and axial diffusivity (AD) across spatially distinct white matter regions. Increases were shown across one MD-cluster (3195 voxels; mean change: 2.34%), one AD-cluster (5740 voxels; mean change: 1.75%), and three RD-clusters (817 total voxels; mean change: 3.11 to 4.70%). Decreases were shown across two MD-clusters (1637 total voxels; mean change: -1.43 to -1.48%), two RD-clusters (1240 total voxels; mean change: -1.92 to -1.93%), and one AD-cluster (724 voxels; mean change: -1.28%). The resulting pattern implies the presence of strain-induced injuries in central and brainstem regions, with comparatively milder physical exercise-induced effects across frontal and superior regions of the left hemisphere, which need further investigation. This article highlights key considerations that need to be addressed in future work to enhance our understanding of the nature of observed white matter changes, improve the comparability of findings across studies, and promote data pooling initiatives to allow more detailed investigations (e.g., exploring sex- and sport-specific effects).

Morphological simulation tests the limits on phenotype discovery in 3D image analysis

In the past few decades, advances in 3D imaging have created new opportunities for reverse genetic screens. Rapidly growing datasets of 3D images of genetic knockouts require high-throughput, automated computational approaches for identifying and characterizing new phenotypes. However, exploratory, discovery-oriented image analysis pipelines used to discover these phenotypes can be difficult to validate because, by their nature, the expected outcome is not known a priori . Introducing known morphological variation through simulation can help distinguish between real phenotypic differences and random variation; elucidate the effects of sample size; and test the sensitivity and reproducibility of morphometric analyses. Here we present a novel approach for 3D morphological simulation that uses open-source, open-access tools available in 3D Slicer, SlicerMorph, and Advanced Normalization Tools in R (ANTsR). While we focus on diffusible-iodine contrast-enhanced micro-CT (diceCT) images, this approach can be used on any volumetric image. We then use our simulated datasets to test whether tensor-based morphometry (TBM) can recover our introduced differences; to test how effect size and sample size affect detectability; and to determine the reproducibility of our results. In our approach to morphological simulation, we first generate a simulated deformation based on a reference image and then propagate this deformation to subjects using inverse transforms obtained from the registration of subjects to the reference. This produces a new dataset with a shifted population mean while retaining individual variability because each sample deforms more or less based on how different or similar it is from the reference. TBM is a widely-used technique that statistically compares local volume differences associated with local deformations. Our results showed that TBM recovered our introduced morphological differences, but that detectability was dependent on the effect size, the sample size, and the region of interest (ROI) included in the analysis. Detectability of subtle phenotypes can be improved both by increasing the sample size and by limiting analyses to specific body regions. However, it is not always feasible to increase sample sizes in screens of essential genes. Therefore, methodical use of ROIs is a promising way to increase the power of TBM to detect subtle phenotypes. Generating known morphological variation through simulation has broad applicability in developmental, evolutionary, and biomedical morphometrics and is a useful way to distinguish between a failure to detect morphological difference and a true lack of morphological difference. Morphological simulation can also be applied to AI-based supervised learning to augment datasets and overcome dataset limitations.