Assessing the Performance of Artificial Intelligence Models: Insights from the American Society of Functional Neuroradiology Artificial Intelligence Competition

BACKGROUND AND PURPOSE

Artificial intelligence (AI) models in radiology are frequently developed and validated using datasets from a single institution and are rarely tested on independent, external datasets, raising questions about their generalizability and applicability in clinical practice. The American Society of Functional Neuroradiology (ASFNR) organized a multi-center AI competition to evaluate the proficiency of developed models in identifying various pathologies on NCCT, assessing age-based normality and estimating medical urgency.

MATERIALS AND METHODS

In total, 1201 anonymized, full-head NCCT clinical scans from five institutions were pooled to form the dataset. The dataset encompassed normal studies as well as pathologies including acute ischemic stroke, intracranial hemorrhage, traumatic brain injury, and mass effect (detection of these-task 1). NCCTs were also assessed to determine if findings were consistent with expected brain changes for the patient's age (task 2: age-based normality assessment) and to identify any abnormalities requiring immediate medical attention (task 3: evaluation of findings for urgent intervention). Five neuroradiologists labeled each NCCT, with consensus interpretations serving as the ground truth. The competition was announced online, inviting academic institutions and companies. Independent central analysis assessed each model's performance. Accuracy, sensitivity, specificity, positive and negative predictive values, and receiver operating characteristic (ROC) curves were generated for each AI model, along with the area under the ROC curve (AUROC).

RESULTS

1177 studies were processed by four teams. The median age of patients was 62, with an interquartile range of 33. 19 teams from various academic institutions registered for the competition. Of these, four teams submitted their final results. No commercial entities participated in the competition. For task 1, AUROCs ranged from 0.49 to 0.59. For task 2, two teams completed the task with AUROC values of 0.57 and 0.52. For task 3, teams had little to no agreement with the ground truth.

CONCLUSIONS

To assess the performance of AI models in real-world clinical scenarios, we analyzed their performance in the ASFNR AI Competition. The first ASFNR Competition underscored the gap between expectation and reality; the models largely fell short in their assessments. As the integration of AI tools into clinical workflows increases, neuroradiologists must carefully recognize the capabilities, constraints, and consistency of these technologies. Before institutions adopt these algorithms, thorough validation is essential to ensure acceptable levels of performance in clinical settings.ABBREVIATIONS: AI = artificial intelligence; ASFNR = American Society of Functional Neuroradiology; AUROC = area under the receiver operating characteristic curve; DICOM = Digital Imaging and Communications in Medicine; GEE = generalized estimation equation; IQR = interquartile range; NPV = negative predictive value; PPV = positive predictive value; ROC = receiver operating characteristic; TBI = traumatic brain injury.

Recommendations for quantitative cerebral perfusion MRI using multi-timepoint arterial spin labeling: Acquisition, quantification, and clinical applications

Accurate assessment of cerebral perfusion is vital for understanding the hemodynamic processes involved in various neurological disorders and guiding clinical decision-making. This guidelines article provides a comprehensive overview of quantitative perfusion imaging of the brain using multi-timepoint arterial spin labeling (ASL), along with recommendations for its acquisition and quantification. A major benefit of acquiring ASL data with multiple label durations and/or post-labeling delays (PLDs) is being able to account for the effect of variable arterial transit time (ATT) on quantitative perfusion values and additionally visualize the spatial pattern of ATT itself, providing valuable clinical insights. Although multi-timepoint data can be acquired in the same scan time as single-PLD data with comparable perfusion measurement precision, its acquisition and postprocessing presents challenges beyond single-PLD ASL, impeding widespread adoption. Building upon the 2015 ASL consensus article, this work highlights the protocol distinctions specific to multi-timepoint ASL and provides robust recommendations for acquiring high-quality data. Additionally, we propose an extended quantification model based on the 2015 consensus model and discuss relevant postprocessing options to enhance the analysis of multi-timepoint ASL data. Furthermore, we review the potential clinical applications where multi-timepoint ASL is expected to offer significant benefits. This article is part of a series published by the International Society for Magnetic Resonance in Medicine (ISMRM) Perfusion Study Group, aiming to guide and inspire the advancement and utilization of ASL beyond the scope of the 2015 consensus article.

A Clinical and Imaging Fused Deep Learning Model Matches Expert Clinician Prediction of 90-Day Stroke Outcomes

BACKGROUND AND PURPOSE

Predicting long-term clinical outcome in acute ischemic stroke is beneficial for prognosis, clinical trial design, resource management, and patient expectations. This study used a deep learning-based predictive model (DLPD) to predict 90-day mRS outcomes and compared its predictions with those made by physicians.

MATERIALS AND METHODS

A previously developed DLPD that incorporated DWI and clinical data from the acute period was used to predict 90-day mRS outcomes in 80 consecutive patients with acute ischemic stroke from a single-center registry. We assessed the predictions of the model alongside those of 5 physicians (2 stroke neurologists and 3 neuroradiologists provided with the same imaging and clinical information). The primary analysis was the agreement between the ordinal mRS predictions of the model or physician and the ground truth using the Gwet Agreement Coefficient. We also evaluated the ability to identify unfavorable outcomes (mRS >2) using the area under the curve, sensitivity, and specificity. Noninferiority analyses were undertaken using limits of 0.1 for the Gwet Agreement Coefficient and 0.05 for the area under the curve analysis. The accuracy of prediction was also assessed using the mean absolute error for prediction, percentage of predictions ±1 categories away from the ground truth (±1 accuracy [ACC]), and percentage of exact predictions (ACC).

RESULTS

To predict the specific mRS score, the DLPD yielded a Gwet Agreement Coefficient score of 0.79 (95% CI, 0.71-0.86), surpassing the physicians' score of 0.76 (95% CI, 0.67-0.84), and was noninferior to the readers (P < .001). For identifying unfavorable outcome, the model achieved an area under the curve of 0.81 (95% CI, 0.72-0.89), again noninferior to the readers' area under the curve of 0.79 (95% CI, 0.69-0.87) (P < .005). The mean absolute error, ±1ACC, and ACC were 0.89, 81%, and 36% for the DLPD.

CONCLUSIONS

A deep learning method using acute clinical and imaging data for long-term functional outcome prediction in patients with acute ischemic stroke, the DLPD, was noninferior to that of clinical readers.

Turning brain MRI into diagnostic PET: 15O-water PET CBF synthesis from multi-contrast MRI via attention-based encoder-decoder networks

Accurate quantification of cerebral blood flow (CBF) is essential for the diagnosis and assessment of a wide range of neurological diseases. Positron emission tomography (PET) with radiolabeled water (15O-water) is the gold-standard for the measurement of CBF in humans, however, it is not widely available due to its prohibitive costs and the use of short-lived radiopharmaceutical tracers that require onsite cyclotron production. Magnetic resonance imaging (MRI), in contrast, is more accessible and does not involve ionizing radiation. This study presents a convolutional encoder-decoder network with attention mechanisms to predict the gold-standard 15O-water PET CBF from multi-contrast MRI scans, thus eliminating the need for radioactive tracers. The model was trained and validated using 5-fold cross-validation in a group of 126 subjects consisting of healthy controls and cerebrovascular disease patients, all of whom underwent simultaneous 15O-water PET/MRI. The results demonstrate that the model can successfully synthesize high-quality PET CBF measurements (with an average SSIM of 0.924 and PSNR of 38.8 dB) and is more accurate compared to concurrent and previous PET synthesis methods. We also demonstrate the clinical significance of the proposed algorithm by evaluating the agreement for identifying the vascular territories with impaired CBF. Such methods may enable more widespread and accurate CBF evaluation in larger cohorts who cannot undergo PET imaging due to radiation concerns, lack of access, or logistic challenges.

Short- and Long-Term MRI Assessed Hemodynamic Changes in Pediatric Moyamoya Patients After Revascularization

BACKGROUND

Cerebrovascular reserve (CVR) reflects the capacity of cerebral blood flow (CBF) to change following a vasodilation challenge. Decreased CVR is associated with a higher stroke risk in patients with cerebrovascular diseases. While revascularization can improve CVR and reduce this risk in adult patients with vasculopathy such as those with Moyamoya disease, its impact on hemodynamics in pediatric patients remains to be elucidated. Arterial spin labeling (ASL) is a quantitative MRI technique that can measure CBF, CVR, and arterial transit time (ATT) non-invasively.

PURPOSE

To investigate the short- and long-term changes in hemodynamics after bypass surgeries in patients with Moyamoya disease.

STUDY TYPE

Longitudinal.

POPULATION

Forty-six patients (11 months-18 years, 28 females) with Moyamoya disease.

FIELD STRENGTH/SEQUENCE

3-T, single- and multi-delay ASL, T1-weighted, T2-FLAIR, 3D MRA.

ASSESSMENT

Imaging was performed 2 weeks before and 1 week and 6 months after surgical intervention. Acetazolamide was employed to induce vasodilation during the imaging procedure. CBF and ATT were measured by fitting the ASL data to the general kinetic model. CVR was computed as the percentage change in CBF. The mean CBF, ATT, and CVR values were measured in the regions affected by vasculopathy.

STATISTICAL TESTS

Pre- and post-revascularization CVR, CBF, and ATT were compared for different regions of the brain. P-values <0.05 were considered statistically significant.

RESULTS

ASL-derived CBF in flow territories affected by vasculopathy significantly increased after bypass by 41 ± 31% within a week. At 6 months, CBF significantly increased by 51 ± 34%, CVR increased by 68 ± 33%, and ATT was significantly reduced by 6.6 ± 2.9%.

DATA CONCLUSION

There may be short- and long-term improvement in the hemodynamic parameters of pediatric Moyamoya patients after bypass surgery.

EVIDENCE LEVEL

4 TECHNICAL EFFICACY: Stage 2.

Predicting FDG-PET Images From Multi-Contrast MRI Using Deep Learning in Patients With Brain Neoplasms

BACKGROUND

18 F-fluorodeoxyglucose (FDG) positron emission tomography (PET) is valuable for determining presence of viable tumor, but is limited by geographical restrictions, radiation exposure, and high cost.

PURPOSE

To generate diagnostic-quality PET equivalent imaging for patients with brain neoplasms by deep learning with multi-contrast MRI.

STUDY TYPE

Retrospective.

SUBJECTS

Patients (59 studies from 51 subjects; age 56 ± 13 years; 29 males) who underwent 18 F-FDG PET and MRI for determining recurrent brain tumor.

FIELD STRENGTH/SEQUENCE

3T; 3D GRE T1, 3D GRE T1c, 3D FSE T2-FLAIR, and 3D FSE ASL, 18 F-FDG PET imaging.

ASSESSMENT

Convolutional neural networks were trained using four MRIs as inputs and acquired FDG PET images as output. The agreement between the acquired and synthesized PET was evaluated by quality metrics and Bland-Altman plots for standardized uptake value ratio. Three physicians scored image quality on a 5-point scale, with score ≥3 as high-quality. They assessed the lesions on a 5-point scale, which was binarized to analyze diagnostic consistency of the synthesized PET compared to the acquired PET.

STATISTICAL TESTS

The agreement in ratings between the acquired and synthesized PET were tested with Gwet's AC and exact Bowker test of symmetry. Agreement of the readers was assessed by Gwet's AC. P = 0.05 was used as the cutoff for statistical significance.

RESULTS

The synthesized PET visually resembled the acquired PET and showed significant improvement in quality metrics (+21.7% on PSNR, +22.2% on SSIM, -31.8% on RSME) compared with ASL. A total of 49.7% of the synthesized PET were considered as high-quality compared to 73.4% of the acquired PET which was statistically significant, but with distinct variability between readers. For the positive/negative lesion assessment, the synthesized PET had an accuracy of 87% but had a tendency to overcall.

CONCLUSION

The proposed deep learning model has the potential of synthesizing diagnostic quality FDG PET images without the use of radiotracers.

EVIDENCE LEVEL

3 TECHNICAL EFFICACY: Stage 2.

Exploring the performance and explainability of fine-tuned BERT models for neuroradiology protocol assignment

BACKGROUND

Deep learning has demonstrated significant advancements across various domains. However, its implementation in specialized areas, such as medical settings, remains approached with caution. In these high-stake environments, understanding the model's decision-making process is critical. This study assesses the performance of different pretrained Bidirectional Encoder Representations from Transformers (BERT) models and delves into understanding its decision-making within the context of medical image protocol assignment.

METHODS

Four different pre-trained BERT models (BERT, BioBERT, ClinicalBERT, RoBERTa) were fine-tuned for the medical image protocol classification task. Word importance was measured by attributing the classification output to every word using a gradient-based method. Subsequently, a trained radiologist reviewed the resulting word importance scores to assess the model's decision-making process relative to human reasoning.

RESULTS

The BERT model came close to human performance on our test set. The BERT model successfully identified relevant words indicative of the target protocol. Analysis of important words in misclassifications revealed potential systematic errors in the model.

CONCLUSIONS

The BERT model shows promise in medical image protocol assignment by reaching near human level performance and identifying key words effectively. The detection of systematic errors paves the way for further refinements to enhance its safety and utility in clinical settings.

Recommended Resting-State fMRI Acquisition and Preprocessing Steps for Preoperative Mapping of Language and Motor and Visual Areas in Adult and Pediatric Patients with Brain Tumors and Epilepsy

Resting-state (rs) fMRI has been shown to be useful for preoperative mapping of functional areas in patients with brain tumors and epilepsy. However, its lack of standardization limits its widespread use and hinders multicenter collaboration. The American Society of Functional Neuroradiology, American Society of Pediatric Neuroradiology, and the American Society of Neuroradiology Functional and Diffusion MR Imaging Study Group recommend specific rs-fMRI acquisition approaches and preprocessing steps that will further support rs-fMRI for future clinical use. A task force with expertise in fMRI from multiple institutions provided recommendations on the rs-fMRI steps needed for mapping of language, motor, and visual areas in adult and pediatric patients with brain tumor and epilepsy. These were based on an extensive literature review and expert consensus.Following rs-fMRI acquisition parameters are recommended: minimum 6-minute acquisition time; scan with eyes open with fixation; obtain rs-fMRI before both task-based fMRI and contrast administration; temporal resolution of ≤2 seconds; scanner field strength of 3T or higher. The following rs-fMRI preprocessing steps and parameters are recommended: motion correction (seed-based correlation analysis [SBC], independent component analysis [ICA]); despiking (SBC); volume censoring (SBC, ICA); nuisance regression of CSF and white matter signals (SBC); head motion regression (SBC, ICA); bandpass filtering (SBC, ICA); and spatial smoothing with a kernel size that is twice the effective voxel size (SBC, ICA).The consensus recommendations put forth for rs-fMRI acquisition and preprocessing steps will aid in standardization of practice and guide rs-fMRI program development across institutions. Standardized rs-fMRI protocols and processing pipelines are essential for multicenter trials and to implement rs-fMRI as part of standard clinical practice.

Early-Frame [18F]Florbetaben PET/MRI for Cerebral Blood Flow Quantification in Patients with Cognitive Impairment: Comparison to an [15O]Water Gold Standard

Cerebral blood flow (CBF) may be estimated from early-frame PET imaging of lipophilic tracers, such as amyloid agents, enabling measurement of this important biomarker in participants with dementia and memory decline. Although previous methods could map relative CBF, quantitative measurement in absolute units (mL/100 g/min) remained challenging and has not been evaluated against the gold standard method of [15O]water PET. The purpose of this study was to develop and validate a minimally invasive quantitative CBF imaging method combining early [18F]florbetaben (eFBB) with phase-contrast MRI using simultaneous PET/MRI. Methods: Twenty participants (11 men and 9 women; 8 cognitively normal, 9 with mild cognitive impairment, and 3 with dementia; 10 β-amyloid negative and 10 β-amyloid positive; 69 ± 9 y old) underwent [15O]water PET, phase-contract MRI, and eFBB imaging in a single session on a 3-T PET/MRI scanner. Quantitative CBF images were created from the first 2 min of brain activity after [18F]florbetaben injection combined with phase-contrast MRI measurement of total brain blood flow. These maps were compared with [15O]water CBF using concordance correlation (CC) and Bland-Altman statistics for gray matter, white matter, and individual regions derived from the automated anatomic labeling (AAL) atlas. Results: The 2 methods showed similar results in gray matter ([15O]water, 55.2 ± 14.7 mL/100 g/min; eFBB, 55.9 ± 14.2 mL/100 g/min; difference, 0.7 ± 2.4 mL/100 g/min; P = 0.2) and white matter ([15O]water, 21.4 ± 5.6 mL/100 g/min; eFBB, 21.2 ± 5.3 mL/100 g/min; difference, -0.2 ± 1.0 mL/100 g/min; P = 0.4). The intrasubject CC for AAL-derived regions was high (0.91 ± 0.04). Intersubject CC in different AAL-derived regions was similarly high, ranging from 0.86 for midfrontal regions to 0.98 for temporal regions. There were no significant differences in performance between the methods in the amyloid-positive and amyloid-negative groups as well as participants with different cognitive statuses. Conclusion: We conclude that eFBB PET/MRI can provide robust CBF measurements, highlighting the capability of simultaneous PET/MRI to provide measurements of both CBF and amyloid burden in a single imaging session in participants with memory disorders.

Measuring Quantitative Cerebral Blood Flow in Healthy Children: A Systematic Review of Neuroimaging Techniques

Cerebral blood flow (CBF) is an important hemodynamic parameter to evaluate brain health. It can be obtained quantitatively using medical imaging modalities such as magnetic resonance imaging and positron emission tomography (PET). Although CBF in adults has been widely studied and linked with cerebrovascular and neurodegenerative diseases, CBF data in healthy children are sparse due to the challenges in pediatric neuroimaging. An understanding of the factors affecting pediatric CBF and its normal range is crucial to determine the optimal CBF measuring techniques in pediatric neuroradiology. This review focuses on pediatric CBF studies using neuroimaging techniques in 32 articles including 2668 normal subjects ranging from birth to 18 years old. A systematic literature search was conducted in PubMed, Embase, and Scopus and reported following the preferred reporting items for systematic reviews and meta-analyses (PRISMA). We identified factors (such as age, gender, mood, sedation, and fitness) that have significant effects on pediatric CBF quantification. We also investigated factors influencing the CBF measurements in infants. Based on this review, we recommend best practices to improve CBF measurements in pediatric neuroimaging. LEVEL OF EVIDENCE: 1 TECHNICAL EFFICACY: Stage 2.

USE-Evaluator: Performance metrics for medical image segmentation models supervised by uncertain, small or empty reference annotations in neuroimaging

Performance metrics for medical image segmentation models are used to measure the agreement between the reference annotation and the predicted segmentation. Usually, overlap metrics, such as the Dice, are used as a metric to evaluate the performance of these models in order for results to be comparable. However, there is a mismatch between the distributions of cases and the difficulty level of segmentation tasks in public data sets compared to clinical practice. Common metrics used to assess performance fail to capture the impact of this mismatch, particularly when dealing with datasets in clinical settings that involve challenging segmentation tasks, pathologies with low signal, and reference annotations that are uncertain, small, or empty. Limitations of common metrics may result in ineffective machine learning research in designing and optimizing models. To effectively evaluate the clinical value of such models, it is essential to consider factors such as the uncertainty associated with reference annotations, the ability to accurately measure performance regardless of the size of the reference annotation volume, and the classification of cases where reference annotations are empty. We study how uncertain, small, and empty reference annotations influence the value of metrics on a stroke in-house data set regardless of the model. We examine metrics behavior on the predictions of a standard deep learning framework in order to identify suitable metrics in such a setting. We compare our results to the BRATS 2019 and Spinal Cord public data sets. We show how uncertain, small, or empty reference annotations require a rethinking of the evaluation. The evaluation code was released to encourage further analysis of this topic https://github.com/SophieOstmeier/UncertainSmallEmpty.git.

Ethical Considerations and Fairness in the Use of Artificial Intelligence for Neuroradiology

In this review, concepts of algorithmic bias and fairness are defined qualitatively and mathematically. Illustrative examples are given of what can go wrong when unintended bias or unfairness in algorithmic development occurs. The importance of explainability, accountability, and transparency with respect to artificial intelligence algorithm development and clinical deployment is discussed. These are grounded in the concept of "primum no nocere" (first, do no harm). Steps to mitigate unfairness and bias in task definition, data collection, model definition, training, testing, deployment, and feedback are provided. Discussions on the implementation of fairness criteria that maximize benefit and minimize unfairness and harm to neuroradiology patients will be provided, including suggestions for neuroradiologists to consider as artificial intelligence algorithms gain acceptance into neuroradiology practice and become incorporated into routine clinical workflow.

Revascularization improves vascular hemodynamics - a study assessing cerebrovascular reserve and transit time in Moyamoya patients using MRI

Cerebrovascular reserve (CVR) reflects the capacity of cerebral blood flow (CBF) to change. Decreased CVR implies poor hemodynamics and is linked to a higher risk for stroke. Revascularization has been shown to improve CBF in patients with vasculopathy such as Moyamoya disease. Dynamic susceptibility contrast (DSC) can measure transit time to evaluate patients suspected of stroke. Arterial spin labeling (ASL) is a non-invasive technique for CBF, CVR, and arterial transit time (ATT) measurements. Here, we investigate the change in hemodynamics 4-12 months after extracranial-to-intracranial direct bypass in 52 Moyamoya patients using ASL with single and multiple post-labeling delays (PLD). Images were collected using ASL and DSC with acetazolamide. CVR, CBF, ATT, and time-to-maximum (Tmax) were measured in different flow territories. Results showed that hemodynamics improved significantly in regions affected by arterial occlusions after revascularization. CVR increased by 16 ± 11% (p < 0.01) and 25 ± 13% (p < 0.01) for single- and multi-PLD ASL, respectively. Transit time measured by multi-PLD ASL and post-vasodilation DSC reduced by 13 ± 7% (p < 0.01) and 9 ± 5% (p < 0.01), respectively. For all regions, ATT correlated significantly with Tmax (R2  =  0.59, p < 0.01). Thus, revascularization improved CVR and decreased transit times. Multi-PLD ASL can serve as an effective and non-invasive modality to examine vascular hemodynamics in Moyamoya patients.

Reducing Gadolinium Contrast With Artificial Intelligence

Gadolinium contrast is an important agent in magnetic resonance imaging (MRI), particularly in neuroimaging where it can help identify blood-brain barrier breakdown from an inflammatory, infectious, or neoplastic process. However, gadolinium contrast has several drawbacks, including nephrogenic systemic fibrosis, gadolinium deposition in the brain and bones, and allergic-like reactions. As computer hardware and technology continues to evolve, machine learning has become a possible solution for eliminating or reducing the dose of gadolinium contrast. This review summarizes the clinical uses of gadolinium contrast, the risks of gadolinium contrast, and state-of-the-art machine learning methods that have been applied to reduce or eliminate gadolinium contrast administration, as well as their current limitations, with a focus on neuroimaging applications. EVIDENCE LEVEL: 3 TECHNICAL EFFICACY: Stage 1.

Artificial Intelligence to Reduce or Eliminate the Need for Gadolinium-Based Contrast Agents in Brain and Cardiac MRI: A Literature Review

ABSTRACT

Brain and cardiac MRIs are fundamental noninvasive imaging tools, which can provide important clinical information and can be performed without or with gadolinium-based contrast agents (GBCAs), depending on the clinical indication. It is currently a topic of debate whether it would be feasible to extract information such as standard gadolinium-enhanced MRI while injecting either less or no GBCAs. Artificial intelligence (AI) is a great source of innovation in medical imaging and has been explored as a method to synthesize virtual contrast MR images, potentially yielding similar diagnostic performance without the need to administer GBCAs. If possible, there would be significant benefits, including reduction of costs, acquisition time, and environmental impact with respect to conventional contrast-enhanced MRI examinations. Given its promise, we believe additional research is needed to increase the evidence to make these AI solutions feasible, reliable, and robust enough to be integrated into the clinical framework. Here, we review recent AI studies aimed at reducing or replacing gadolinium in brain and cardiac imaging while maintaining diagnostic image quality.

International EANM-SNMMI-ISMRM consensus recommendation for PET/MRI in oncology

PREAMBLE

The Society of Nuclear Medicine and Molecular Imaging (SNMMI) is an international scientific and professional organization founded in 1954 to promote the science, technology, and practical application of nuclear medicine. The European Association of Nuclear Medicine (EANM) is a professional non-profit medical association that facilitates communication worldwide between individuals pursuing clinical and research excellence in nuclear medicine. The EANM was founded in 1985. The merged International Society for Magnetic Resonance in Medicine (ISMRM) is an international, nonprofit, scientific association whose purpose is to promote communication, research, development, and applications in the field of magnetic resonance in medicine and biology and other related topics and to develop and provide channels and facilities for continuing education in the field.The ISMRM was founded in 1994 through the merger of the Society of Magnetic Resonance in Medicine and the Society of Magnetic Resonance Imaging. SNMMI, ISMRM, and EANM members are physicians, technologists, and scientists specializing in the research and practice of nuclear medicine and/or magnetic resonance imaging. The SNMMI, ISMRM, and EANM will periodically define new guidelines for nuclear medicine practice to help advance the science of nuclear medicine and/or magnetic resonance imaging and to improve the quality of service to patients throughout the world. Existing practice guidelines will be reviewed for revision or renewal, as appropriate, on their fifth anniversary or sooner, if indicated. Each practice guideline, representing a policy statement by the SNMMI/EANM/ISMRM, has undergone a thorough consensus process in which it has been subjected to extensive review. The SNMMI, ISMRM, and EANM recognize that the safe and effective use of diagnostic nuclear medicine imaging and magnetic resonance imaging requires specific training, skills, and techniques, as described in each document. Reproduction or modification of the published practice guideline by those entities not providing these services is not authorized. These guidelines are an educational tool designed to assist practitioners in providing appropriate care for patients. They are not inflexible rules or requirements of practice and are not intended, nor should they be used, to establish a legal standard of care. For these reasons and those set forth below, the SNMMI, the ISMRM, and the EANM caution against the use of these guidelines in litigation in which the clinical decisions of a practitioner are called into question. The ultimate judgment regarding the propriety of any specific procedure or course of action must be made by the physician or medical physicist in light of all the circumstances presented. Thus, there is no implication that an approach differing from the guidelines, standing alone, is below the standard of care. To the contrary, a conscientious practitioner may responsibly adopt a course of action different from that set forth in the guidelines when, in the reasonable judgment of the practitioner, such course of action is indicated by the condition of the patient, limitations of available resources, or advances in knowledge or technology subsequent to publication of the guidelines. The practice of medicine includes both the art and the science of the prevention, diagnosis, alleviation, and treatment of disease. The variety and complexity of human conditions make it impossible to always reach the most appropriate diagnosis or to predict with certainty a particular response to treatment. Therefore, it should be recognized that adherence to these guidelines will not ensure an accurate diagnosis or a successful outcome. All that should be expected is that the practitioner will follow a reasonable course of action based on current knowledge, available resources, and the needs of the patient to deliver effective and safe medical care. The sole purpose of these guidelines is to assist practitioners in achieving this objective.

LSOR: Longitudinally-Consistent Self-Organized Representation Learning

Interpretability is a key issue when applying deep learning models to longitudinal brain MRIs. One way to address this issue is by visualizing the high-dimensional latent spaces generated by deep learning via self-organizing maps (SOM). SOM separates the latent space into clusters and then maps the cluster centers to a discrete (typically 2D) grid preserving the high-dimensional relationship between clusters. However, learning SOM in a high-dimensional latent space tends to be unstable, especially in a self-supervision setting. Furthermore, the learned SOM grid does not necessarily capture clinically interesting information, such as brain age. To resolve these issues, we propose the first self-supervised SOM approach that derives a high-dimensional, interpretable representation stratified by brain age solely based on longitudinal brain MRIs (i.e., without demographic or cognitive information). Called Longitudinally-consistent Self-Organized Representation learning (LSOR), the method is stable during training as it relies on soft clustering (vs. the hard cluster assignments used by existing SOM). Furthermore, our approach generates a latent space stratified according to brain age by aligning trajectories inferred from longitudinal MRIs to the reference vector associated with the corresponding SOM cluster. When applied to longitudinal MRIs of the Alzheimer's Disease Neuroimaging Initiative (ADNI, N = 632 ), LSOR generates an interpretable latent space and achieves comparable or higher accuracy than the state-of-the-art representations with respect to the downstream tasks of classification (static vs. progressive mild cognitive impairment) and regression (determining ADAS-Cog score of all subjects). The code is available at https://github.com/ouyangjiahong/longitudinal-som-single-modality.

Non-inferiority of deep learning ischemic stroke segmentation on non-contrast CT within 16-hours compared to expert neuroradiologists

We determined if a convolutional neural network (CNN) deep learning model can accurately segment acute ischemic changes on non-contrast CT compared to neuroradiologists. Non-contrast CT (NCCT) examinations from 232 acute ischemic stroke patients who were enrolled in the DEFUSE 3 trial were included in this study. Three experienced neuroradiologists independently segmented hypodensity that reflected the ischemic core on each scan. The neuroradiologist with the most experience (expert A) served as the ground truth for deep learning model training. Two additional neuroradiologists' (experts B and C) segmentations were used for data testing. The 232 studies were randomly split into training and test sets. The training set was further randomly divided into 5 folds with training and validation sets. A 3-dimensional CNN architecture was trained and optimized to predict the segmentations of expert A from NCCT. The performance of the model was assessed using a set of volume, overlap, and distance metrics using non-inferiority thresholds of 20%, 3 ml, and 3 mm, respectively. The optimized model trained on expert A was compared to test experts B and C. We used a one-sided Wilcoxon signed-rank test to test for the non-inferiority of the model-expert compared to the inter-expert agreement. The final model performance for the ischemic core segmentation task reached a performance of 0.46 ± 0.09 Surface Dice at Tolerance 5mm and 0.47 ± 0.13 Dice when trained on expert A. Compared to the two test neuroradiologists the model-expert agreement was non-inferior to the inter-expert agreement, [Formula: see text]. The before, CNN accurately delineates the hypodense ischemic core on NCCT in acute ischemic stroke patients with an accuracy comparable to neuroradiologists.

Longitudinal Alterations of Cerebral Blood Flow in High-Contact Sports

OBJECTIVE

Repetitive head trauma is common in high-contact sports. Cerebral blood flow (CBF) can measure changes in brain perfusion that could indicate injury. Longitudinal studies with a control group are necessary to account for interindividual and developmental effects. We investigated whether exposure to head impacts causes longitudinal CBF changes.

METHODS

We prospectively studied 63 American football (high-contact cohort) and 34 volleyball (low-contact controls) male collegiate athletes, tracking CBF using 3D pseudocontinuous arterial spin labeling magnetic resonance imaging for up to 4 years. Regional relative CBF (rCBF, normalized to cerebellar CBF) was computed after co-registering to T1-weighted images. A linear mixed effects model assessed the relationship of rCBF to sport, time, and their interaction. Within football players, we modeled rCBF against position-based head impact risk and baseline Standardized Concussion Assessment Tool score. Additionally, we evaluated early (1-5 days) and delayed (3-6 months) post-concussion rCBF changes (in-study concussion).

RESULTS

Supratentorial gray matter rCBF declined in football compared with volleyball (sport-time interaction p = 0.012), with a strong effect in the parietal lobe (p = 0.002). Football players with higher position-based impact-risk had lower occipital rCBF over time (interaction p = 0.005), whereas players with lower baseline Standardized Concussion Assessment Tool score (worse performance) had relatively decreased rCBF in the cingulate-insula over time (interaction effect p = 0.007). Both cohorts showed a left-right rCBF asymmetry that decreased over time. Football players with an in-study concussion showed an early increase in occipital lobe rCBF (p = 0.0166).

INTERPRETATION

These results suggest head impacts may result in an early increase in rCBF, but cumulatively a long-term decrease in rCBF. ANN NEUROL 2023;94:457-469.

Generative Adversarial Network-Enhanced Ultra-Low-Dose [18F]-PI-2620 τ PET/MRI in Aging and Neurodegenerative Populations

BACKGROUND AND PURPOSE

With the utility of hybrid τ PET/MR imaging in the screening, diagnosis, and follow-up of individuals with neurodegenerative diseases, we investigated whether deep learning techniques can be used in enhancing ultra-low-dose [18F]-PI-2620 τ PET/MR images to produce diagnostic-quality images.

MATERIALS AND METHODS

Forty-four healthy aging participants and patients with neurodegenerative diseases were recruited for this study, and [18F]-PI-2620 τ PET/MR data were simultaneously acquired. A generative adversarial network was trained to enhance ultra-low-dose τ images, which were reconstructed from a random sampling of 1/20 (approximately 5% of original count level) of the original full-dose data. MR images were also used as additional input channels. Region-based analyses as well as a reader study were conducted to assess the image quality of the enhanced images compared with their full-dose counterparts.

RESULTS

The enhanced ultra-low-dose τ images showed apparent noise reduction compared with the ultra-low-dose images. The regional standard uptake value ratios showed that while, in general, there is an underestimation for both image types, especially in regions with higher uptake, when focusing on the healthy-but-amyloid-positive population (with relatively lower τ uptake), this bias was reduced in the enhanced ultra-low-dose images. The radiotracer uptake patterns in the enhanced images were read accurately compared with their full-dose counterparts.

CONCLUSIONS

The clinical readings of deep learning-enhanced ultra-low-dose τ PET images were consistent with those performed with full-dose imaging, suggesting the possibility of reducing the dose and enabling more frequent examinations for dementia monitoring.

Functional Outcome Prediction in Acute Ischemic Stroke Using a Fused Imaging and Clinical Deep Learning Model

BACKGROUND

Predicting long-term clinical outcome based on the early acute ischemic stroke information is valuable for prognostication, resource management, clinical trials, and patient expectations. Current methods require subjective decisions about which imaging features to assess and may require time-consuming postprocessing. This study's goal was to predict ordinal 90-day modified Rankin Scale (mRS) score in acute ischemic stroke patients by fusing a Deep Learning model of diffusion-weighted imaging images and clinical information from the acute period.

METHODS

A total of 640 acute ischemic stroke patients who underwent magnetic resonance imaging within 1 to 7 days poststroke and had 90-day mRS follow-up data were randomly divided into 70% (n=448) for model training, 15% (n=96) for validation, and 15% (n=96) for internal testing. Additionally, external testing on a cohort from Lausanne University Hospital (n=280) was performed to further evaluate model generalization. Accuracy for ordinal mRS, accuracy within ±1 mRS category, mean absolute prediction error, and determination of unfavorable outcome (mRS score >2) were evaluated for clinical only, imaging only, and 2 fused clinical-imaging models.

RESULTS

The fused models demonstrated superior performance in predicting ordinal mRS score and unfavorable outcome in both internal and external test cohorts when compared with the clinical and imaging models. For the internal test cohort, the top fused model had the highest area under the curve of 0.92 for unfavorable outcome prediction and the lowest mean absolute error (0.96 [95% CI, 0.77-1.16]), with the highest proportion of mRS score predictions within ±1 category (79% [95% CI, 71%-88%]). On the external Lausanne University Hospital cohort, the best fused model had an area under the curve of 0.90 for unfavorable outcome prediction and outperformed other models with an mean absolute error of 0.90 (95% CI, 0.79-1.01), and the highest percentage of mRS score predictions within ±1 category (83% [95% CI, 78%-87%]).

CONCLUSIONS

A Deep Learning-based imaging model fused with clinical variables can be used to predict 90-day stroke outcome with reduced subjectivity and user burden.

One Model to Synthesize Them All: Multi-Contrast Multi-Scale Transformer for Missing Data Imputation

Multi-contrast magnetic resonance imaging (MRI) is widely used in clinical practice as each contrast provides complementary information. However, the availability of each imaging contrast may vary amongst patients, which poses challenges to radiologists and automated image analysis algorithms. A general approach for tackling this problem is missing data imputation, which aims to synthesize the missing contrasts from existing ones. While several convolutional neural networks (CNN) based algorithms have been proposed, they suffer from the fundamental limitations of CNN models, such as the requirement for fixed numbers of input and output channels, the inability to capture long-range dependencies, and the lack of interpretability. In this work, we formulate missing data imputation as a sequence-to-sequence learning problem and propose a multi-contrast multi-scale Transformer (MMT), which can take any subset of input contrasts and synthesize those that are missing. MMT consists of a multi-scale Transformer encoder that builds hierarchical representations of inputs combined with a multi-scale Transformer decoder that generates the outputs in a coarse-to-fine fashion. The proposed multi-contrast Swin Transformer blocks can efficiently capture intra- and inter-contrast dependencies for accurate image synthesis. Moreover, MMT is inherently interpretable as it allows us to understand the importance of each input contrast in different regions by analyzing the in-built attention maps of Transformer blocks in the decoder. Extensive experiments on two large-scale multi-contrast MRI datasets demonstrate that MMT outperforms the state-of-the-art methods quantitatively and qualitatively.

Deep Learning-Generated Synthetic MR Imaging STIR Spine Images Are Superior in Image Quality and Diagnostically Equivalent to Conventional STIR: A Multicenter, Multireader Trial

BACKGROUND AND PURPOSE

Deep learning image reconstruction allows faster MR imaging acquisitions while matching or exceeding the standard of care and can create synthetic images from existing data sets. This multicenter, multireader spine study evaluated the performance of synthetically created STIR compared with acquired STIR.

MATERIALS AND METHODS

From a multicenter, multiscanner data base of 328 clinical cases, a nonreader neuroradiologist randomly selected 110 spine MR imaging studies in 93 patients (sagittal T1, T2, and STIR) and classified them into 5 categories of disease and healthy. A DICOM-based deep learning application generated a synthetically created STIR series from the sagittal T1 and T2 images. Five radiologists (3 neuroradiologists, 1 musculoskeletal radiologist, and 1 general radiologist) rated the STIR quality and classified disease pathology (study 1, n = 80). They then assessed the presence or absence of findings typically evaluated with STIR in patients with trauma (study 2, n = 30). The readers evaluated studies with either acquired STIR or synthetically created STIR in a blinded and randomized fashion with a 1-month washout period. The interchangeability of acquired STIR and synthetically created STIR was assessed using a noninferiority threshold of 10%.

RESULTS

For classification, there was a decrease in interreader agreement expected by randomly introducing synthetically created STIR of 3.23%. For trauma, there was an overall increase in interreader agreement by +1.9%. The lower bound of confidence for both exceeded the noninferiority threshold, indicating interchangeability of synthetically created STIR with acquired STIR. Both the Wilcoxon signed-rank and t tests showed higher image-quality scores for synthetically created STIR over acquired STIR (P < .0001).

CONCLUSIONS

Synthetically created STIR spine MR images were diagnostically interchangeable with acquired STIR, while providing significantly higher image quality, suggesting routine clinical practice potential.

Effect of vitamin D supplementation on cerebral blood flow in male patients with adrenoleukodystrophy

One-third of boys with X-linked adrenoleukodystrophy (ALD) develop inflammatory demyelinating lesions, typically at the splenium. These lesions share similarities with multiple sclerosis, including cerebral hypoperfusion and links to vitamin D insufficiency. We hypothesized that increasing vitamin D levels would increase cerebral blood flow (CBF) in ALD boys. We conducted an exploratory analysis of vitamin D supplementation and CBF using all available data from participants enrolled in a recent single-arm interventional study of vitamin D supplementation in boys with ALD. We measured whole brain and splenium CBF using arterial spin labeling (ASL) from three study time points (baseline, 6 months, and 12 months). We used linear generalized estimating equations to evaluate CBF changes between time points and to test for an association between CBF and vitamin D. ASL data were available for 16 participants, aged 2-22 years. Mean vitamin D levels increased by 72.7% (p < .001) after 6 months and 88.6% (p < .01) after 12 months. Relative to baseline measures, mean CBF of the whole brain (6 months: +2.5%, p = .57; 12 months: +6.1%, p = .18) and splenium (6 months: +1.2%, p = .80; 12 months: +7.4%, p = .058) were not significantly changed. Vitamin D levels were positively correlated with CBF in the splenium (slope = .59, p < .001). In this exploratory analysis, we observed a correlation between vitamin D levels and splenial CBF in ALD boys. We confirm the feasibility of measuring CBF in this brain region and population, but further work is needed to establish a causal role for vitamin D in modulating CBF.

Role of Brain Imaging in the Prediction of Intracerebral Hemorrhage Following Endovascular Therapy for Acute Stroke

Currently most acute ischemic stroke patients presenting with a large vessel occlusion are treated with endovascular therapy (EVT), which results in high rates of successful recanalization. Despite this success, more than half of EVT-treated patients are significantly disabled 3 months later partly due to the occurrence of post-EVT intracerebral hemorrhage. Predicting post-EVT intracerebral hemorrhage is important for individualizing treatment strategies in clinical practice (eg, safe initiation of early antithrombotic therapies), as well as in selecting the optimal candidates for clinical trials that aim to reduce this deleterious outcome. Emerging data suggest that brain and vascular imaging biomarkers may be particularly relevant since they provide insights into the ongoing acute stroke pathophysiology. In this review/perspective, we summarize the accumulating literature on the role of cerebrovascular imaging biomarkers in predicting post-EVT-associated intracerebral hemorrhage. We focus on imaging acquired before EVT, during the EVT procedure, and in the early post-EVT time frames when new therapeutic therapies could be tested. Accounting for the complex pathophysiology of post-EVT-associated intracerebral hemorrhage, this review may provide some guidance for future prospective observational or therapeutic studies.

Critical Appraisal of Artificial Intelligence-Enabled Imaging Tools Using the Levels of Evidence System

Clinical adoption of an artificial intelligence-enabled imaging tool requires critical appraisal of its life cycle from development to implementation by using a systematic, standardized, and objective approach that can verify both its technical and clinical efficacy. Toward this concerted effort, the ASFNR/ASNR Artificial Intelligence Workshop Technology Working Group is proposing a hierarchal evaluation system based on the quality, type, and amount of scientific evidence that the artificial intelligence-enabled tool can demonstrate for each component of its life cycle. The current proposal is modeled after the levels of evidence in medicine, with the uppermost level of the hierarchy showing the strongest evidence for potential impact on patient care and health care outcomes. The intended goal of establishing an evidence-based evaluation system is to encourage transparency, foster an understanding of the creation of artificial intelligence tools and the artificial intelligence decision-making process, and to report the relevant data on the efficacy of artificial intelligence tools that are developed. The proposed system is an essential step in working toward a more formalized, clinically validated, and regulated framework for the safe and effective deployment of artificial intelligence imaging applications that will be used in clinical practice.

Current state and guidance on arterial spin labeling perfusion MRI in clinical neuroimaging

This article focuses on clinical applications of arterial spin labeling (ASL) and is part of a wider effort from the International Society for Magnetic Resonance in Medicine (ISMRM) Perfusion Study Group to update and expand on the recommendations provided in the 2015 ASL consensus paper. Although the 2015 consensus paper provided general guidelines for clinical applications of ASL MRI, there was a lack of guidance on disease-specific parameters. Since that time, the clinical availability and clinical demand for ASL MRI has increased. This position paper provides guidance on using ASL in specific clinical scenarios, including acute ischemic stroke and steno-occlusive disease, arteriovenous malformations and fistulas, brain tumors, neurodegenerative disease, seizures/epilepsy, and pediatric neuroradiology applications, focusing on disease-specific considerations for sequence optimization and interpretation. We present several neuroradiological applications in which ASL provides unique information essential for making the diagnosis. This guidance is intended for anyone interested in using ASL in a routine clinical setting (i.e., on a single-subject basis rather than in cohort studies) building on the previous ASL consensus review.

Predicting Hypoperfusion Lesion and Target Mismatch in Stroke from Diffusion-weighted MRI Using Deep Learning

Background Perfusion imaging is important to identify a target mismatch in stroke but requires contrast agents and postprocessing software. Purpose To use a deep learning model to predict the hypoperfusion lesion in stroke and identify patients with a target mismatch profile from diffusion-weighted imaging (DWI) and clinical information alone, using perfusion MRI as the reference standard. Materials and Methods Imaging data sets of patients with acute ischemic stroke with baseline perfusion MRI and DWI were retrospectively reviewed from multicenter data available from 2008 to 2019 (Imaging Collaterals in Acute Stroke, Diffusion and Perfusion Imaging Evaluation for Understanding Stroke Evolution 2, and University of California, Los Angeles stroke registry). For perfusion MRI, rapid processing of perfusion and diffusion software automatically segmented the hypoperfusion lesion (time to maximum, ≥6 seconds) and ischemic core (apparent diffusion coefficient [ADC], ≤620 × 10-6 mm2/sec). A three-dimensional U-Net deep learning model was trained using baseline DWI, ADC, National Institutes of Health Stroke Scale score, and stroke symptom sidedness as inputs, with the union of hypoperfusion and ischemic core segmentation serving as the ground truth. Model performance was evaluated using the Dice score coefficient (DSC). Target mismatch classification based on the model was compared with that of the clinical-DWI mismatch approach defined by the DAWN trial by using the McNemar test. Results Overall, 413 patients (mean age, 67 years ± 15 [SD]; 207 men) were included for model development and primary analysis using fivefold cross-validation (247, 83, and 83 patients in the training, validation, and test sets, respectively, for each fold). The model predicted the hypoperfusion lesion with a median DSC of 0.61 (IQR, 0.45-0.71). The model identified patients with target mismatch with a sensitivity of 90% (254 of 283; 95% CI: 86, 93) and specificity of 77% (100 of 130; 95% CI: 69, 83) compared with the clinical-DWI mismatch sensitivity of 50% (140 of 281; 95% CI: 44, 56) and specificity of 89% (116 of 130; 95% CI: 83, 94) (P < .001 for all). Conclusion A three-dimensional U-Net deep learning model predicted the hypoperfusion lesion from diffusion-weighted imaging (DWI) and clinical information and identified patients with a target mismatch profile with higher sensitivity than the clinical-DWI mismatch approach. ClinicalTrials.gov registration nos. NCT02225730, NCT01349946, NCT02586415 © RSNA, 2022 Supplemental material is available for this article. See also the editorial by Kallmes and Rabinstein in this issue.

Amyloid-Related Imaging Abnormalities: An Update

Amyloid-related imaging abnormalities (ARIA) is a term introduced in 2010 to encompass a spectrum of MRI findings observed in patients receiving investigational anti-amyloid beta (Aβ) immunotherapies for Alzheimer disease (AD). The entity can be broadly categorized into ARIA characterized by edema and effusion (ARIA-E) and ARIA characterized by microhemorrhages and superficial siderosis (ARIA-H). ARIA typically occurs early in the treatment course and has a higher incidence in patients who are apolipoprotein E ε4 allele carriers. ARIA-E has an additional dose dependence, with higher incidence in patients receiving higher doses of anti-Aβ immuno-therapies. ARIA is often asymptomatic and self-resolving. The recognition of ARIA has implications for patient selection and monitoring for Aβ immunotherapies, and its development can potentially lead to a pause or discontinuation of therapy. The FDA's first approval of an Aβ-targeting monoclonal antibody for AD treatment in 2021 will lead to such therapy's expanded use beyond the clinical trial setting and to radiologists more commonly encountering ARIA in clinical practice. This review explores the theorized pathophysiologic mechanisms for ARIA, describes the MRI findings and grading schemes for ARIA-E and AREA-H, and summarizes relevant Aβ immunotherapies. Through such knowledge, radiologists can optimally impact the management of patients receiving targeted AD therapies.

Abstract WMP53: Revascularization Improves Cerebrovascular Reserve In Moyamoya Patients

Introduction: Cerebrovascular reserve (CVR) reflects the change in cerebral blood flow (CBF) in response to a vasoactive stimulus and can be measured using a brain stress test whereby CBF is augmented using vasodilators. Studies have demonstrated that impaired CVR was associated with a higher risk of acute stroke and that bypass can reduce the risk of strokes. Arterial spin labeling is a quantitative MRI technique that enables non-invasive CBF and CVR measurement. Studies have shown the effectiveness of ASL in identifying CVR impairment in Moyamoya patients. Here, we compare the CBF and CVR of Moyamoya patients before and after bypass using single and multiple delay ASL techniques.

Methods: MRI data were acquired from 24 Moyamoya patients (16-54 years, 3 males) using a 3T MRI system. All patients had unilateral or bilateral vessel occlusion at ACA, MCA and/or PCA. The scanning parameters of the single post-labeling delay pseudo-continuous ASL (single-PLD PCASL), delay pseudo-continuous ASL (multi-PLD PCASL, 5 PLDs) 2. ASL data were acquired simultaneously at baseline and 15 minutes after the injection of the vasodilator acetazolamide (15 mg/kg with a maximum 1 g). CBF was computed using the general kinetic model. CVR was computed as the percentage of CBF change compared with baseline CBF. Flow territories (right and left ACA, MCA, and PCA) were defined based on the Harvard-Oxford cortical and subcortical structural atlases. Mean CVR within the territories affected by occlusion and normal territories was computed for each subject. T-tests were performed to compare the mean CVR between the affected and unaffected territories.

Results: For this cohort, our quantitative analysis revealed that CVR of the affected regions increased significantly by 12% and 21% measured by single and multiple delay PCASL respectively, implying improved CBF and lower risk for strokes.

Discussion: ASL was effective to detect improved CBF and CVR in Moyamoya patients after bypass surgeries.

2.5D and 3D segmentation of brain metastases with deep learning on multinational MRI data

Introduction

Management of patients with brain metastases is often based on manual lesion detection and segmentation by an expert reader. This is a time- and labor-intensive process, and to that end, this work proposes an end-to-end deep learning segmentation network for a varying number of available MRI available sequences.

Methods

We adapt and evaluate a 2.5D and a 3D convolution neural network trained and tested on a retrospective multinational study from two independent centers, in addition, nnU-Net was adapted as a comparative benchmark. Segmentation and detection performance was evaluated by: (1) the dice similarity coefficient, (2) a per-metastases and the average detection sensitivity, and (3) the number of false positives.

Results

The 2.5D and 3D models achieved similar results, albeit the 2.5D model had better detection rate, whereas the 3D model had fewer false positive predictions, and nnU-Net had fewest false positives, but with the lowest detection rate. On MRI data from center 1, the 2.5D, 3D, and nnU-Net detected 79%, 71%, and 65% of all metastases; had an average per patient sensitivity of 0.88, 0.84, and 0.76; and had on average 6.2, 3.2, and 1.7 false positive predictions per patient, respectively. For center 2, the 2.5D, 3D, and nnU-Net detected 88%, 86%, and 78% of all metastases; had an average per patient sensitivity of 0.92, 0.91, and 0.85; and had on average 1.0, 0.4, and 0.1 false positive predictions per patient, respectively.

Discussion/Conclusion

Our results show that deep learning can yield highly accurate segmentations of brain metastases with few false positives in multinational data, but the accuracy degrades for metastases with an area smaller than 0.4 cm².

Predicting final ischemic stroke lesions from initial diffusion-weighted images using a deep neural network

BACKGROUND

For prognosis of stroke, measurement of the diffusion-perfusion mismatch is a common practice for estimating tissue at risk of infarction in the absence of timely reperfusion. However, perfusion-weighted imaging (PWI) adds time and expense to the acute stroke imaging workup. We explored whether a deep convolutional neural network (DCNN) model trained with diffusion-weighted imaging obtained at admission could predict final infarct volume and location in acute stroke patients.

METHODS

In 445 patients, we trained and validated an attention-gated (AG) DCNN to predict final infarcts as delineated on follow-up studies obtained 3 to 7 days after stroke. The input channels consisted of MR diffusion-weighted imaging (DWI), apparent diffusion coefficients (ADC) maps, and thresholded ADC maps with values less than 620 × 10^-6 mm²/s, while the output was a voxel-by-voxel probability map of tissue infarction. We evaluated performance of the model using the area under the receiver-operator characteristic curve (AUC), the Dice similarity coefficient (DSC), absolute lesion volume error, and the concordance correlation coefficient (ρ_c) of the predicted and true infarct volumes.

RESULTS

The model obtained a median AUC of 0.91 (IQR: 0.84-0.96). After thresholding at an infarction probability of 0.5, the median sensitivity and specificity were 0.60 (IQR: 0.16-0.84) and 0.97 (IQR: 0.93-0.99), respectively, while the median DSC and absolute volume error were 0.50 (IQR: 0.17-0.66) and 27 ml (IQR: 7-60 ml), respectively. The model's predicted lesion volumes showed high correlation with ground truth volumes (ρ_c = 0.73, p < 0.01).

CONCLUSION

An AG-DCNN using diffusion information alone upon admission was able to predict infarct volumes at 3-7 days after stroke onset with comparable accuracy to models that consider both DWI and PWI. This may enable treatment decisions to be made with shorter stroke imaging protocols.

Disentangling Normal Aging From Severity of Disease via Weak Supervision on Longitudinal MRI

The continuous progression of neurological diseases are often categorized into conditions according to their severity. To relate the severity to changes in brain morphometry, there is a growing interest in replacing these categories with a continuous severity scale that longitudinal MRIs are mapped onto via deep learning algorithms. However, existing methods based on supervised learning require large numbers of samples and those that do not, such as self-supervised models, fail to clearly separate the disease effect from normal aging. Here, we propose to explicitly disentangle those two factors via weak-supervision. In other words, training is based on longitudinal MRIs being labelled either normal or diseased so that the training data can be augmented with samples from disease categories that are not of primary interest to the analysis. We do so by encouraging trajectories of controls to be fully encoded by the direction associated with brain aging. Furthermore, an orthogonal direction linked to disease severity captures the residual component from normal aging in the diseased cohort. Hence, the proposed method quantifies disease severity and its progression speed in individuals without knowing their condition. We apply the proposed method on data from the Alzheimer's Disease Neuroimaging Initiative (ADNI, N =632 ). We then show that the model properly disentangled normal aging from the severity of cognitive impairment by plotting the resulting disentangled factors of each subject and generating simulated MRIs for a given chronological age and condition. Moreover, our representation obtains higher balanced accuracy when used for two downstream classification tasks compared to other pre-training approaches. The code for our weak-supervised approach is available at https://github.com/ouyangjiahong/longitudinal-direction-disentangle.

Velocity-selective arterial spin labeling perfusion MRI: A review of the state of the art and recommendations for clinical implementation

This review article provides an overview of the current status of velocity-selective arterial spin labeling (VSASL) perfusion MRI and is part of a wider effort arising from the International Society for Magnetic Resonance in Medicine (ISMRM) Perfusion Study Group. Since publication of the 2015 consensus paper on arterial spin labeling (ASL) for cerebral perfusion imaging, important advancements have been made in the field. The ASL community has, therefore, decided to provide an extended perspective on various aspects of technical development and application. Because VSASL has the potential to become a principal ASL method because of its unique advantages over traditional approaches, an in-depth discussion was warranted. VSASL labels blood based on its velocity and creates a magnetic bolus immediately proximal to the microvasculature within the imaging volume. VSASL is, therefore, insensitive to transit delay effects, in contrast to spatially selective pulsed and (pseudo-) continuous ASL approaches. Recent technical developments have improved the robustness and the labeling efficiency of VSASL, making it a potentially more favorable ASL approach in a wide range of applications where transit delay effects are of concern. In this review article, we (1) describe the concepts and theoretical basis of VSASL; (2) describe different variants of VSASL and their implementation; (3) provide recommended parameters and practices for clinical adoption; (4) describe challenges in developing and implementing VSASL; and (5) describe its current applications. As VSASL continues to undergo rapid development, the focus of this review is to summarize the fundamental concepts of VSASL, describe existing VSASL techniques and applications, and provide recommendations to help the clinical community adopt VSASL.

Using arterial spin labeling to measure cerebrovascular reactivity in Moyamoya disease: Insights from simultaneous PET/MRI

Cerebrovascular reactivity (CVR) reflects the CBF change to meet different physiological demands. The reference CVR technique is PET imaging with vasodilators but is inaccessible to most patients. DSC can measure transit time to evaluate patients suspected of stroke, but the use of gadolinium may cause side-effects. Arterial spin labeling (ASL) is a non-invasive MRI technique for CBF measurements. Here, we investigate the effectiveness of ASL with single and multiple post labeling delays (PLD) to replace PET and DSC for CVR and transit time mapping in 26 Moyamoya patients. Images were collected using simultaneous PET/MRI with acetazolamide. CVR, CBF, arterial transit time (ATT), and time-to-maximum (Tmax) were measured in different flow territories. Results showed that CVR was lower in occluded regions than normal regions (by 68 ± 12%, 52 ± 5%, and 56 ± 9%, for PET, single- and multi-PLD PCASL, respectively, all p < 0.05). Multi-PLD PCASL correlated slightly higher with PET (CCC = 0.36 and 0.32 in affected and unaffected territories respectively). Vasodilation caused ATT to reduce by 4.5 ± 3.1% (p < 0.01) in occluded regions. ATT correlated significantly with Tmax (R² > 0.35, p < 0.01). Therefore, multi-PLD ASL is recommended for CVR studies due to its high agreement with the reference PET technique and the capability of measuring transit time.

Unsupervised Deep Learning for Stroke Lesion Segmentation on Follow-up CT Based on Generative Adversarial Networks

BACKGROUND AND PURPOSE

Supervised deep learning is the state-of-the-art method for stroke lesion segmentation on NCCT. Supervised methods require manual lesion annotations for model development, while unsupervised deep learning methods such as generative adversarial networks do not. The aim of this study was to develop and evaluate a generative adversarial network to segment infarct and hemorrhagic stroke lesions on follow-up NCCT scans.

MATERIALS AND METHODS

Training data consisted of 820 patients with baseline and follow-up NCCT from 3 Dutch acute ischemic stroke trials. A generative adversarial network was optimized to transform a follow-up scan with a lesion to a generated baseline scan without a lesion by generating a difference map that was subtracted from the follow-up scan. The generated difference map was used to automatically extract lesion segmentations. Segmentation of primary hemorrhagic lesions, hemorrhagic transformation of ischemic stroke, and 24-hour and 1-week follow-up infarct lesions were evaluated relative to expert annotations with the Dice similarity coefficient, Bland-Altman analysis, and intraclass correlation coefficient.

RESULTS

The median Dice similarity coefficient was 0.31 (interquartile range, 0.08-0.59) and 0.59 (interquartile range, 0.29-0.74) for the 24-hour and 1-week infarct lesions, respectively. A much lower Dice similarity coefficient was measured for hemorrhagic transformation (median, 0.02; interquartile range, 0-0.14) and primary hemorrhage lesions (median, 0.08; interquartile range, 0.01-0.35). Predicted lesion volume and the intraclass correlation coefficient were good for the 24-hour (bias, 3 mL; limits of agreement, -64-59 mL; intraclass correlation coefficient, 0.83; 95% CI, 0.78-0.88) and excellent for the 1-week (bias, -4 m; limits of agreement,-66-58 mL; intraclass correlation coefficient, 0.90; 95% CI, 0.83-0.93) follow-up infarct lesions.

CONCLUSIONS

An unsupervised generative adversarial network can be used to obtain automated infarct lesion segmentations with a moderate Dice similarity coefficient and good volumetric correspondence.

Self-supervised learning of neighborhood embedding for longitudinal MRI

In recent years, several deep learning models recommend first to represent Magnetic Resonance Imaging (MRI) as latent features before performing a downstream task of interest (such as classification or regression). The performance of the downstream task generally improves when these latent representations are explicitly associated with factors of interest. For example, we derived such a representation for capturing brain aging by applying self-supervised learning to longitudinal MRIs and then used the resulting encoding to automatically identify diseases accelerating the aging of the brain. We now propose a refinement of this representation by replacing the linear modeling of brain aging with one that is consistent in local neighborhoods in the latent space. Called Longitudinal Neighborhood Embedding (LNE), we derive an encoding so that neighborhoods are age-consistent (i.e., brain MRIs of different subjects with similar brain ages are in close proximity of each other) and progression-consistent, i.e., the latent space is defined by a smooth trajectory field where each trajectory captures changes in brain ages between a pair of MRIs extracted from a longitudinal sequence. To make the problem computationally tractable, we further propose a strategy for mini-batch sampling so that the resulting local neighborhoods accurately approximate the ones that would be defined based on the whole cohort. We evaluate LNE on three different downstream tasks: (1) to predict chronological age from T1-w MRI of 274 healthy subjects participating in a study at SRI International; (2) to distinguish Normal Control (NC) from Alzheimer's Disease (AD) and stable Mild Cognitive Impairment (sMCI) from progressive Mild Cognitive Impairment (pMCI) based on T1-w MRI of 632 participants of the Alzheimer's Disease Neuroimaging Initiative (ADNI); and (3) to distinguish no-to-low from moderate-to-heavy alcohol drinkers based on fractional anisotropy derived from diffusion tensor MRIs of 764 adolescents recruited by the National Consortium on Alcohol and NeuroDevelopment in Adolescence (NCANDA). Across the three data sets, the visualization of the smooth trajectory vector fields and superior accuracy on downstream tasks demonstrate the strength of the proposed method over existing self-supervised methods in extracting information related to brain aging, which could help study the impact of substance use and neurodegenerative disorders. The code is available at https://github.com/ouyangjiahong/longitudinal-neighbourhood-embedding.

Kidney Disease, Hypertension Treatment, and Cerebral Perfusion and Structure

RATIONALE & OBJECTIVE

The safety of intensive blood pressure (BP) targets is controversial for persons with chronic kidney disease (CKD). We studied the effects of hypertension treatment on cerebral perfusion and structure in individuals with and without CKD.

STUDY DESIGN

Neuroimaging substudy of a randomized trial.

SETTING & PARTICIPANTS

A subset of participants in the Systolic Blood Pressure Intervention Trial (SPRINT) who underwent brain magnetic resonance imaging studies. Presence of baseline CKD was assessed by estimated glomerular filtration rate (eGFR) and urinary albumin-creatinine ratio (UACR).

INTERVENTION

Participants were randomly assigned to intensive (systolic BP <120 mm Hg) versus standard (systolic BP <140 mm Hg) BP lowering.

OUTCOMES

The magnetic resonance imaging outcome measures were the 4-year change in global cerebral blood flow (CBF), white matter lesion (WML) volume, and total brain volume (TBV).

RESULTS

A total of 716 randomized participants with a mean age of 68 years were enrolled; follow-up imaging occurred after a median 3.9 years. Among participants with eGFR <60 mL/min/1.73 m2 (n = 234), the effects of intensive versus standard BP treatment on change in global CBF, WMLs, and TBV were 3.38 (95% CI, 0.32 to 6.44) mL/100 g/min, -0.06 (95% CI, -0.16 to 0.04) cm3 (inverse hyperbolic sine-transformed), and -3.8 (95% CI, -8.3 to 0.7) cm3, respectively. Among participants with UACR >30 mg/g (n = 151), the effects of intensive versus standard BP treatment on change in global CBF, WMLs, and TBV were 1.91 (95% CI, -3.01 to 6.82) mL/100 g/min, 0.003 (95% CI, -0.13 to 0.13) cm3 (inverse hyperbolic sine-transformed), and -7.0 (95% CI, -13.3 to -0.3) cm3, respectively. The overall treatment effects on CBF and TBV were not modified by baseline eGFR or UACR; however, the effect on WMLs was attenuated in participants with albuminuria (P = 0.04 for interaction).

LIMITATIONS

Measurement variability due to multisite design.

CONCLUSIONS

Among adults with hypertension who have primarily early kidney disease, intensive versus standard BP treatment did not appear to have a detrimental effect on brain perfusion or structure. The findings support the safety of intensive BP treatment targets on brain health in persons with early kidney disease.

FUNDING

SPRINT was funded by the National Institutes of Health (including the National Heart, Lung, and Blood Institute; the National Institute of Diabetes and Digestive and Kidney Diseases; the National Institute on Aging; and the National Institute of Neurological Disorders and Stroke), and this substudy was funded by the National Institutes of Diabetes and Digestive and Kidney Diseases.

TRIAL REGISTRATION

SPRINT was registered at ClinicalTrials.gov with study number NCT01206062.

Reproducibility of cerebrovascular reactivity measurements: A systematic review of neuroimaging techniques. J Cereb Blood Flow Metab 2022;42:700-717. [PMID: 34806918 PMCID: PMC9254040 DOI: 10.1177/0271678x211056702]

Cerebrovascular reactivity (CVR), the capacity of the brain to increase cerebral blood flow (CBF) to meet changes in physiological demand, is an important biomarker to evaluate brain health. Typically, this brain "stress test" is performed by using a medical imaging modality to measure the CBF change between two states: at baseline and after vasodilation. However, since there are many imaging modalities and many ways to augment CBF, a wide range of CVR values have been reported. An understanding of CVR reproducibility is critical to determine the most reliable methods to measure CVR as a clinical biomarker. This review focuses on CVR reproducibility studies using neuroimaging techniques in 32 articles comprising 427 total subjects. The literature search was performed in PubMed, Embase, and Scopus. The review was conducted using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA). We identified 5 factors of the experimental subjects (such as sex, blood characteristics, and smoking) and 9 factors of the measuring technique (such as the imaging modality, the type of the vasodilator, and the quantification method) that have strong effects on CVR reproducibility. Based on this review, we recommend several best practices to improve the reproducibility of CVR quantification in neuroimaging studies.

Automated detection of arterial landmarks and vascular occlusions in patients with acute stroke receiving digital subtraction angiography using deep learning

BACKGROUND

Digital subtraction angiography (DSA) is the gold-standard method of assessing arterial blood flow and blockages prior to endovascular thrombectomy.

OBJECTIVE

To detect anatomical features and arterial occlusions with DSA using artificial intelligence techniques.

METHODS

We included 82 patients with acute ischemic stroke who underwent DSA imaging and whose carotid terminus was visible in at least one run. Two neurointerventionalists labeled the carotid location (when visible) and vascular occlusions on 382 total individual DSA runs. For detecting the carotid terminus, positive and negative image patches (either containing or not containing the internal carotid artery terminus) were extracted in a 1:1 ratio. Two convolutional neural network architectures (ResNet-50 pretrained on ImageNet and ResNet-50 trained from scratch) were evaluated. Area under the curve (AUC) of the receiver operating characteristic and pixel distance from the ground truth were calculated. The same training and analysis methods were used for detecting arterial occlusions.

RESULTS

The ResNet-50 trained from scratch most accurately detected the carotid terminus (AUC 0.998 (95% CI 0.997 to 0.999), p<0.00001) and arterial occlusions (AUC 0.973 (95% CI 0.971 to 0.975), p<0.0001). Average pixel distances from ground truth for carotid terminus and occlusion localization were 63±45 and 98±84, corresponding to approximately 1.26±0.90 cm and 1.96±1.68 cm for a standard angiographic field-of-view.

CONCLUSION

These results may serve as an unbiased standard for clinical stroke trials, as optimal standardization would be useful for core laboratories in endovascular thrombectomy studies, and also expedite decision-making during DSA-based procedures.