Examining the Clinical Utility of Deep Learning Models for Perivascular Space Segmentation in Alzheimer’s Disease

Topic: Brain imaging, connectivity and assessment

Open-source deep learning models promise scalable quantification of MRI-visible perivascular spaces (PVS), an emerging marker of glymphatic dysfunction in Alzheimer’s disease (AD). Whether these models remain valid when applied to heterogeneous AD cohorts and preserve ground-truth derived biological associations has not been established. We benchmarked five open-source nnU-Net-based PVS segmentation models (nnU-Net T1, nnU-Net T1+FLAIR, mcPVS-Net, MedNet-PVS, ADNI-PVS-Net) against manual labels in 60 multi-site ADNI-3 participants balanced across cognitively unimpaired, mild cognitive impairment, and AD dementia, and in 12 publicly available scans from the VAscular Lesions object-level and segmentatiOn (VALDO) challenge. We evaluated overall, voxel, object, and lesion-wise Dice; stratified performance by region, diagnosis, scanner, PVS severity, and white matter hyperintensity (WMH) burden; and tested whether model-derived PVS burden reproduced ground-truth derived associations with AD-related biomarkers. Despite high internal performance, all models showed a specialist-generalist trade-off: the AD-trained ADNI-PVS-Net achieved highest internal performance (overall Dice = 0.58±0.14) but degraded most externally (Dice ≤ 0.11), while MedNet-PVS and mcPVS-Net generalized externally (Dice 0.27-0.30) but systematically misestimated PVS burden. Models were relatively robust across scanner and diagnosis but inconsistent across PVS severity, and some misclassified up to 32% of WMH as PVS. Notably, only AD-trained models reproduced the expected PVS-amyloid biomarker relationship; other models returned a null result, despite comparable or superior Dice scores. These findings show that internal segmentation accuracy and external generalizability alone do not guarantee downstream inferential validity. Careful model selection and cohort-specific validation against manual labels should precede clinical inference in AD.

@article{tang2026-a1525cf3,
  author  = {Tang, S. and Younes, K. and Penukonda, E. and Hausle, I. and Thropp, P. and Lupo, J. M. and Tosun, D. and {the Alzheimer’s Disease Neuroimaging Initiative}},
  title   = {{Examining the Clinical Utility of Deep Learning Models for Perivascular Space Segmentation in Alzheimer’s Disease}},
  journal = {medRxiv},
  year    = {2026},
  doi     = {10.64898/2026.09.27.26363725},
  url     = {https://doi.org/10.64898/2026.09.27.26363725}
}

Key findings

• We evaluated overall, voxel, object, and lesion-wise Dice; stratified performance by region, diagnosis, scanner, PVS severity, and white matter hyperintensity (WMH) burden; and tested whether model-derived PVS burden reproduced ground-truth derived associations with AD-related biomarkers.
• Despite high internal performance, all models showed a specialist-generalist trade-off: the AD-trained ADNI-PVS-Net achieved highest internal performance (overall Dice = 0.58±0.14) but degraded most externally (Dice ≤ 0.11), while MedNet-PVS and mcPVS-Net generalized externally (Dice 0.27-0.30) but systematically misestimated PVS burden.
• Notably, only AD-trained models reproduced the expected PVS-amyloid biomarker relationship; other models returned a null result, despite comparable or superior Dice scores.
• These findings show that internal segmentation accuracy and external generalizability alone do not guarantee downstream inferential validity.

New to this subject? Start with our plain-English guide: Digital health →

Study details

• Evidence strengthModerate

Read the original

• DOI: 10.64898/2026.09.27.26363725
• Original source →

Free copy unknown: no free-copy check has an answer for this record yet. That does not mean it is paywalled.

Related studies in Digital health tools

• Perivascular space dilation is associated with vascular amyloid-β accumulation in the overlying cortex Acta neuropathologica · 2021-12-20 · usefulness score 43/100
• Multimodal deep learning models for early detection of Alzheimer's disease stage Scientific reports · 2021-02-05 · usefulness score 43/100
• Deep ensemble learning of sparse regression models for brain disease diagnosis Medical image analysis · 2017-01-24 · usefulness score 41/100
• MedBackdoorGuard: Detecting and Mitigating Backdoor Attacks in Deep Learning Models for MRI-Based Alzheimer's Disease and Mild Cognitive Impairment Diagnosis Preprints.org · 2026-09-20 · usefulness score 16/100
• Leveraging explainable artificial intelligence with ensemble of deep learning model for dementia prediction to enhance clinical decision support systems Scientific Reports · 2025-05-13 · usefulness score 44/100
• Transformer-based deep neural network language models for Alzheimer's disease risk assessment from targeted speech BMC medical informatics and decision making · 2021-03-09 · usefulness score 47/100

Also tagged “convolutional”

205 studies in the library carry the topic tag “convolutional”; these are this study's neighbours in a fixed, alphabetical-by-hash tour of them, not the closest in content. All 205 tagged studies.

• Multi-Modality Cascaded Convolutional Neural Networks for Alzheimer's Disease Diagnosis Neuroinformatics · 2018-10-01 · usefulness score 50/100 · Digital health tools
• Disease progression timeline estimation for Alzheimer's disease using discriminative event based modeling NeuroImage · 2018-11-22 · usefulness score 45/100 · Digital health tools

Similar studies

• Automated quantification of cerebral microbleeds for ARIA-H monitoring in Aging and Alzheimer’s Disease: A multicenter deep learning validation 2026 · Treatments
• Automated classification of Alzheimer's disease and mild cognitive impairment using a single MRI and deep neural networks 2019 · Digital health tools
• Automatic lesion detection and segmentation in 18F-flutemetamol positron emission tomography images using deep learning 2022 · Brain scans and imaging
• AT(N)-Anchored Multimodal Deep Learning for Alzheimer’s Staging: A Taxonomy and Continuous Burden Index for Low-Resource Settings 2026 · Tau biomarkers
• Interpretable classification of Alzheimer's disease pathologies with a convolutional neural network pipeline 2019 · Digital health tools
• Early detection of dementia through retinal imaging and trustworthy AI 2024 · Brain scans and imaging

More on Digital health tools

• A review that covers this ground Clinical prediction models using artificial intelligence approaches in dementia Aging Clinical and Experimental Research · 2025-07-25 · usefulness score 59/100
• Newest in this subject Speaker-Wise Dementia Screening fromNaturalistic Speech with Lexical Entropy, Semantic Drift, and Multimodal Fusion Research Square · 2026-10-21 · Preprint — not yet peer-reviewed · usefulness score 16/100
• Newest in this subject Integrating Multisource Target Evidence with Machine-Learning Potency Prediction for Alzheimer’s Disease Drug Discovery bioRxiv · 2026-10-20 · Preprint — not yet peer-reviewed · usefulness score 16/100

See all studies in Digital health tools →