Vladimir S. Fonov

Background: Neuromelanin-MRI enables in vivo assessment of the substantia nigra (SN) and locus coeruleus (LC) in individuals with Parkinson's disease (PD), yet longitudinal studies rely on cross-sectional processing that may introduce measurement variability and confound estimates of change over time. Objectives: In this paper, a longitudinal neuromelanin-MRI processing framework is presented that is designed and validated to improve measurement stability and reduce processing-related variability across repeated scans. Methods: Imaging and clinical data from the Quebec Parkinson Network were analyzed in 268 participants (199 PD, 69 controls), including a longitudinal subset of 74 participants (49 PD, 25 controls) scanned approximately one year apart. Validation experiments evaluated slice-by-slice intensity normalization for slice dependent intensity variation, bias field correction for LC signal asymmetry, and the effects of longitudinal registration on measurement stability and PD-control discrimination. Results: Slice-by-slice intensity normalization significantly reduced brainstem intensity variability by 3.6%. A systematic leftward signal asymmetry was observed in the LC and persisted following N4 bias field correction, suggesting a scanner-related effect not captured by conventional bias field modeling. Longitudinal registration reduced annualized change variability by 25-36% for SN_CR and 27-34% for LC_CR metrics in controls, indicating improved within-subject measurement stability. Residual variability was also reduced for contrast-based metrics by up to 28%. Longitudinal registration generally produced larger PD-control effect sizes at baseline and follow-up, particularly for SN volume metrics. However, no significant method x group x time interactions were observed, indicating that estimated longitudinal trajectories did not differ significantly between longitudinal and conventional cross-sectional processing. Conclusions: Longitudinal registration reduced technical variability and improved the precision of NM-MRI measurements. Although it did not significantly enhance detection of longitudinal PD-control differences over the follow-up interval examined here, it provides a more robust framework for longitudinal NM-MRI studies and may improve sensitivity to subtler biological effects in future investigations.

Alzheimer’s disease (AD) is characterized by progressive brain changes, including protein aggregation and structural changes. Cerebrospinal fluid (CSF) system abnormalities, such as ventricular dilation, increased choroid plexus volume or positron emission tomography (PET) ligand uptake in the CSF, have also been consistently described. We aimed to examine whether changes in CSF production and clearance might be associated with brain protein aggregation across biological stages of Alzheimer’s disease. We hypothesized an association between brain protein aggregation and changes on the CSF system. We examined 378 individuals from the Translational Biomarkers in Aging and Dementia (TRIAD) cohort with T1-weighted magnetic resonance imaging (MRI), amyloid-PET and tau-PET assessments. We assessed the lateral ventricle and choroid plexus volumes, both corrected for intracranial volume, in the MRI native space. Non-specific ventricular tracer standardized uptake value ratio (SUVR), derived from amyloid- and tau-PET images, was used as an indirect marker of choroid plexus-related clearance activity and served as a metric of CSF dynamics. Linear models tested associations amongst lateral ventricular volume (reflecting CSF space enlargement), choroid plexus volume (reflecting secretory tissue morphology) and ventricular SUVR (reflecting tracer activity within the CSF compartment and serving as an indirect marker of choroid plexus-related clearance function and CSF dynamics) with Aβ and tau aggregations. Analyses were restricted to within-modality associations, relating ventricular radioactivity to cortical pathology for each PET tracer. We found that when considered independently, larger ventricular and choroid plexus volumes were associated with higher neocortical Aβ-PET SUVR, particularly in the precuneus and cingulate cortices. Additionally, lower ventricular radioactivity (derived from amyloid-PET) showed strong negative associations in the dorsal apex of the neocortex. However, when all three ventricular parameters were included in the same model, these effects were mediated by ventricular volume. By contrast, the effect of the ventricular parameters on tau load was mediated by Aβ in the neocortex. Therefore, ventricular enlargement appears to be associated with Aβ load. Distinct from neurodegeneration, changes in ventricular parameters, particularly ventricular volume, are associated with upstream Alzheimer’s disease pathophysiology. While ventricular volume significantly mediated ventricular amyloid clearance, no such effect was observed for tau, suggesting distinct clearance mechanisms for these pathologies in Alzheimer’s disease.

Structured INTRODUCTION: Verbal memory tests like the Rey’s Auditory Learning Test (RAVLT) are widely used in Alzheimer’s disease assessments without consideration of females’ well-known advantages in verbal memory. This study explores if this form of cognitive reserve masks symptoms and delays diagnosis in females. METHODS: We employed a retrospective longitudinal cohort design (ADNI, PREVENT-AD) to model Immediate Recall subcomponents on the RAVLT over time in amyloid negative (Aβ-) cognitively normal participants versus Aβ+ participants that progressed to AD dementia. RESULTS: Our Bayesian model successfully identified the inflection point where cognition deviates from normal onto the AD dementia trajectory. The timing of decline onset differed significantly across sexes, with Aβ+ females maintaining normal cognition longer than Aβ+ males. Among individuals with AD dementia, females experienced a steeper annual decline than males. DISCUSSION: Delayed onset of decline suggests that females preserve cognition despite neuropathology accumulation. This masked period represents a window for earlier detection and timely intervention before irreversible neurodegeneration. Research in Context: Systematic review: The authors reviewed the literature from PubMed and Elsevier to identify verbal memory as a core feature in the Alzheimer’s Disease discourse. Our critical appraisal was complemented with meetings with expert neuropsychologists and neuroinformaticians. Interpretations: We find that females have superior performance on Rey’s Auditory Learning Test (RAVLT) Immediate Recall, females maintain performance for longer than males before being diagnosed with Alzheimer’s Disease, and once diagnosed, they decline faster than males. These findings underscore the importance of conducting Alzheimer’s Disease assessments with considerations for sex differences. Future directions: We aim to continue the investigation of cognitive reserve in females in the context of the Amyloid, Tau and Neurodegeneration (ANT) framework. Considering the early stage of disease is of interest given that our results show the appearance of cognitive symptoms many years before official dementia diagnosis. Highlights: - In cognitively healthy amyloid negative participants, females outperform males on the Rey’s Auditory Learning Test (RAVLT). - In the amyloid positive participants, females maintain normal verbal memory for 2.7 years longer than males. - After divergence from cognitive normality, females experience approximately a 25-50% steeper cognitive decline compared to males. - Our findings call attention to the female verbal memory advantage as an influential form of cognitive reserve that requires diagnostic consideration.

Average anatomical brain templates are routinely used in neuroanatomical and functional studies. Several widely used anatomical models were historically constructed with different algorithms and a limited number of subjects. We performed an experiment to determine the number of subjects required to achieve a stable population average and to estimate variability in individual subjects’ registration. We used a random subset of 2000 subjects from the UK Biobank (between 40 and 60 years of age) to generate a “silver standard” population average and then ran a template generation process with a variable number of subjects from 10 to 320, repeating each draw 50 times in a bootstrapping fashion. We compared two methods which are widely used in the literature to generate population averages (ANIMAL and ANTs). Our results showed that 160 subjects are enough to generate a stable population average, and both methods achieve comparable results, with ANTs having advantage over ANIMAL when a smaller number of subjects are available.

Background: Alzheimer’s disease is a neurodegenerative disease associated with the accumulation of Amyloid‐ß and Tau neurofibrillary tangles following a pattern known as Thal and Braak stages, respectively (Thal 2002; Braak 1995,2011). Recent research (Pascoal 2020) showed the possibility of recapitulating Braak’s histopathological stages in vivo using PET tracer [18F]‐MK‐6240 with manually defined regions of interest. This study analyzes the joint patterns of Amyloid‐ß and Tau accumulation associated with AD in a completely data‐driven fashion.Method: We used T1w MRI, [18F]‐NAV4694, and [18F]MK6240 PET scans from 36 young individuals (47 scans), 192 healthy elderly individuals (280 scans) in the community or outpatients at the McGill University Research Centre for Studies in Aging (97 individuals with mild cognitive impairment (140 scans) and 75 patients with AD (98 scans). T1w MRIs were pre‐processed (non‐uniformity correction, intensity normalization, stereotaxic registration, brain masking, tissue classification) and non‐linearly registered to the ADNI template using ANTs (Avants et al 2011). PET scans were linearly registered to the T1w MRI, and voxel‐wise standardized uptake value ratios (SUVR) were calculated using whole cerebellum grey matter as the reference, and then non‐linearly warped into the space of the ADNI template and downsampled to 2mm3 resolution. All PET scans were concatenated to create a matrix 518×345846. This matrix was factorized into two low‐rank components (Krichene 2018) with the constraint that all matrix elements should be positive, spatial loadings constrained to be between 0 and 1, latent vectors ordered monotonously to capture the additive nature of disease progression, and the loadings aligned with MOCA scores, by minimizing cosine distance. We used spatial components to create pseudo‐probabilistic ROIs associated with the different stages of disease progression.Result: Our method revealed regional patterns of joint tau and amyloid accumulation associated with AD progression, shown in Figure 1. Associated weighted SUVR average values are shown on Figure 2. The first two components, as visible on Figure 2, have almost no relationship with MOCA score and are most likely associated with off‐target bindings for both tracers.Conclusion: This work is a stepping stone towards creation of the Tau & Amyloid AD disease staging system.

Background: Alzheimer’s disease is a neurodegenerative disease associated with the accumulation of Amyloid‐β and Tau neurofibrillary tangles following a pattern known as Thal and Braak stages, respectively (Thal 2002; Braak 1995,2011). Recent research (Pascoal 2020) showed the possibility of recapitulating Braak's histopathological stages in vivo using PET tracer [¹⁸F]‐MK‐6240 with manually defined regions of interest. This study analyzes the joint patterns of Amyloid‐β and Tau accumulation associated with AD in a completely data‐driven fashion.Method: We used T1w MRI, [¹⁸F]‐NAV4694, and [¹⁸F]MK6240 PET scans from 36 young individuals (47 scans), 192 healthy elderly individuals (280 scans) in the community or outpatients at the McGill University Research Centre for Studies in Aging (97 individuals with mild cognitive impairment (140 scans) and 75 patients with AD (98 scans).T1w MRIs were pre‐processed (non‐uniformity correction, intensity normalization, stereotaxic registration, brain masking, tissue classification) and non‐linearly registered to the ADNI template using ANTs ( Avants et al 2011 ).PET scans were linearly registered to the T1w MRI, and voxel‐wise standardized uptake value ratios (SUVR) were calculated using whole cerebellum grey matter as the reference, and then non‐linearly warped into the space of the ADNI template and downsampled to 2mm3 resolution.All PET scans were concatenated to create a matrix 518x345846. This matrix was factorized into two low‐rank components (Krichene 2018) with the constraint that all matrix elements should be positive, spatial loadings constrained to be between 0 and 1, latent vectors ordered monotonously to capture the additive nature of disease progression, and the loadings aligned with MOCA scores, by minimizing cosine distance. We used spatial components to create pseudo‐probabilistic ROIs associated with the different stages of disease progression.Result: Our method revealed regional patterns of joint tau and amyloid accumulation associated with AD progression, shown in Figure 1. Associated weighted SUVR average values are shown on Figure 2. The first two components, as visible on Figure 2, have almost no relationship with MOCA score and are most likely associated with off‐target bindings for both tracers.Conclusion: This work is a stepping stone towards creation of the Tau & Amyloid AD disease staging system.

Developmental and epileptic encephalopathies (DEEs) feature altered brain development, developmental delay and seizures, with seizures exacerbating developmental delay. Here we identify a cohort with biallelic variants in DENND5A, encoding a membrane trafficking protein, and develop animal models with phenotypes like the human syndrome. We demonstrate that DENND5A interacts with Pals1/MUPP1, components of the Crumbs apical polarity complex required for symmetrical division of neural progenitor cells. Human induced pluripotent stem cells lacking DENND5A fail to undergo symmetric cell division with an inherent propensity to differentiate into neurons. These phenotypes result from misalignment of the mitotic spindle in apical neural progenitors. Cells lacking DENND5A orient away from the proliferative apical domain surrounding the ventricles, biasing daughter cells towards a more fate-committed state, ultimately shortening the period of neurogenesis. This study provides a mechanism for DENND5A-related DEE that may be generalizable to other developmental conditions and provides variant-specific clinical information for physicians and families.

Quality control (QC) is an important part of all scientific analyses, including neuroscience. With manual curation considered the gold standard, there remains a lack of available tools that make manual neuroimaging QC accessible, fast, and easy. In this article we present Qrater, a containerized web-based Python application that enables viewing and rating any type of image for QC purposes. Qrater functionalities allow collaboration between various raters on the same dataset which can facilitate completing large QC tasks. Qrater was used to evaluate QC rater performance on three different magnetic resonance (MR) image QC tasks by a group of raters having different amounts of experience. The tasks included QC of raw MR images (10,196 images), QC of linear registration to a standard template (10,196 images), and QC of skull segmentation (6,968 images). We measured the proportion of failed images, average rating time per image, intra- and inter-rater agreement, as well as the comparison against QC using a conventional method. The median time spent rating per image differed significantly between raters (depending on rater experience) in each of the three QC tasks. Evaluating raw MR images was slightly faster using Qrater than an image viewer (expert: 99 vs. 90 images in 63 min; trainee 99 vs 79 images in 98 min). Reviewing the linear registration using Qrater was twice faster for the expert (99 vs. 43 images in 36 min) and three times faster for the trainee (99 vs. 30 images in 37 min). The greatest difference in rating speed resulted from the skull segmentation task where the expert took a full minute to inspect the volume on a slice-by-slice basis compared to just 3 s using Qrater. Rating agreement also depended on the experience of the raters and the task at hand: trained raters’ inter-rater agreements with the expert’s gold standard were moderate for both raw images (Fleiss’ Kappa = 0.44) and linear registration (Fleiss’ Kappa = 0.56); the experts’ inter-rater agreement of the skull segmentation task was excellent (Cohen’s Kappa = 0.83). These results demonstrate that Qrater is a useful asset for QC tasks that rely on manual evaluation of QC images.

Predicting brain age from T1-weighted MRI is a promising marker for understanding brain aging and its associated conditions. While deep learning models have shown success in reducing the Mean Absolute Error (MAE) of predicted brain age, concerns about robust and accurate generalization in new data limit their clinical applicability. The large number of trainable parameters, combined with limited medical imaging training data, contribute to this challenge, often resulting in a generalization gap where there is a significant discrepancy between model performance on training data versus unseen data. In this study, we assess a deep model, SFCN-reg, based on the VGG-16 architecture, and address the generalization gap through comprehensive preprocessing, extensive data augmentation, and model regularization. Using training data from the UK Biobank, we demonstrate substantial improvements in model performance. Specifically, our approach reduces the generalization MAE by 44% (from 5.25 to 2.96 years) in the Alzheimer’s Disease Neuroimaging Initiative dataset and by 22% (from 4.35 to 3.40 years) in the Australian Imaging, Biomarker and Lifestyle dataset. Furthermore, we achieve a 29% reduction in scan-rescan error (from 0.86 to 0.61 years) while enhancing the model’s robustness to registration errors. Feature importance maps highlight anatomical regions used to predict age. These results highlight the critical role of high-quality preprocessing and robust training techniques in improving accuracy and narrowing the generalization gap, both necessary steps towards the clinical use of brain age prediction models. Our study makes valuable contributions to neuroimaging research by offering a potential pathway to improve the clinical applicability of deep learning models.

Background: The hippocampal-to-ventricle ratio (HVR) is a biomarker of medial temporal atrophy, particularly useful in the assessment of neurodegeneration in diseases such as Alzheimer’s disease (AD). To minimize subjectivity and inter-rater variability, an automated, accurate, precise, and reliable segmentation technique for the hippocampus (HC) and surrounding cerebro-spinal fluid (CSF) filled spaces — such as the temporal horns of the lateral ventricles — is essential. Methods: We trained and evaluated three automated methods for the segmentation of both HC and CSF (Multi-Atlas Label Fusion (MALF), Nonlinear Patch-Based Segmentation (NLPB), and a Convolutional Neural Network (CNN)). We then evaluated these methods, including the widely used FreeSurfer technique, using baseline T1w MRIs of 1,641 participants from the AD Neuroimaging Initiative study with various degree of atrophy associated with their cognitive status on the spectrum from cognitively healthy to clinically probable AD. Our gold standard consisted in manual segmentation of HC and CSF from 80 cognitively healthy individuals. We calculated HC volumes and HVR and compared all methods in terms of segmentation reliability, similarity across methods, sensitivity in detecting between-group differences and associations with age, scores of the learning subtest of the Rey Auditory Verbal Learning Test (RAVLT) and the Alzheimer’s Disease Assessment Scale 13 (ADAS13) scores. Results: Cross validation demonstrated that the CNN method yielded more accurate HC and CSF segmentations when compared to MALF and NLPB, demonstrating higher volumetric overlap (Dice Kappa = 0.94) and correlation (rho = 0.99) with the manual labels. It was also the most reliable method in clinical data application, showing minimal failures. Our comparisons yielded high correlations between FreeSurfer, CNN and NLPB volumetric values. HVR yielded higher control:AD effect sizes than HC volumes among all segmentation methods, reinforcing the significance of HVR in clinical distinction. Associations: The positive association with age was significantly stronger for HVR compared to HC volumes on all methods except FreeSurfer. Memory associations with HC volumes or HVR were only significant for individuals with mild cognitive impairment. Finally, the HC volumes and HVR showed comparable negative associations with ADAS13, particularly in the mild cognitive impairment cohort. Conclusion: This study provides an evaluation of automated segmentation methods centered to estimate HVR, emphasizing the superior performance of a CNN-based algorithm. The findings underscore the pivotal role of accurate segmentation in HVR calculations for precise clinical applications, contributing valuable insights into medial temporal lobe atrophy in neurodegenerative disorders, especially AD. Authorship: Sofia Fernandez-Lozano: Conceptualization, Methodology, Software, Investigation, Writing – Original Draft, Visualization. Vladimir Fonov: Software, Data Curation. Dorothee Schoemaker: Resources, Writing – Review & Editing. Jens Pruessner: Resources, Writing – Review & Editing. Olivier Potvin: Resources, Writing – Review & Editing. Simon Duchesne: Resources, Writing – Review & Editing. D. Louis Collins: Conceptualization, Writing – Review & Editing, Supervision.