Deep Learning in neuroimaging for neurodegenerative diseases: State-of-the art, Challenges, and Opportunities
Bibliographic record
Abstract
Neuroimaging is commonly used to diagnose neurodegenerative diseases (NDDs), providing crucial insights into brain changes before clinical symptoms manifest. Deep learning (DL) for neuroimaging can improve early diagnosis and disease monitoring. Clinical implementation of DL faces challenges in accurately representing real-world data. Recent models, particularly those focused on diagnostic categorization, have achieved high accuracy, but their applicability to patients is limited. Conflicting inferences have been reported, with findings from small cohorts generalizing conclusions without considering inter-scanner, intra- and inter-site variations. A theoretically feasible method involves gathering a comprehensive dataset that encompasses all patient demographics, but this presents practical challenges including harmonization, data incompleteness, class imbalance, and substantial costs. Existing research has also mostly focused on common NDDs like Alzheimer's Disease (AD) and Parkinson's Disease (PD). This contribution expands the literature by looking at a wider range of NDDs, exploring the latest advancements in applying deep learning algorithms to neuroimaging analysis for the diagnosis and monitoring of NDDs, including AD, Frontotemporal Dementia (FTD), Lewy Body Dementia, PD, Huntington's Disease, Amyotrophic Lateral Sclerosis, and Multiple Sclerosis. We emphasize how these approaches are handling spatial/temporal information available in brain volume imaging data. We conclude by discussing the challenges associated with the use of voxel-based, patch-based, ROI-based, and slice-based approaches in brain volume imaging. These challenges are further compounded by issues such as inter-site and inter-scanner variability, class imbalances in medical datasets, and the scarcity of accurately annotated data, all of which impact the performance and generalizability of deep learning models.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.006 | 0.009 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.002 |
| Science and technology studies | 0.000 | 0.002 |
| Scholarly communication | 0.003 | 0.005 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.002 | 0.004 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".