What question did this study set out to answer?

To evaluate whether applying deep learning-based image enhancement to diagnostic-quality structural MRI improves downstream classification performance and training data efficiency in Alzheimer's disease.

August 15, 2026Journal of Magnetic Resonance Imaging

Deep Learning‐Based Enhancement of Already Diagnostic‐Quality MRI for Alzheimer's Disease Classification: Effects on Model Performance and Training Data Requirements

View Full Paper

Ask AI

Bookmark

Authors

ZZZheng ZhangDassault Systèmes (United States)ZZZechen ZhouJiangsu University LXLei XiangATUM (United States)

Discussion

Member takes

Implication

Retrospective study demonstrates that deep learning-based MRI enhancement improves diagnostic classification in Alzheimer's disease cohorts, highlighting reduced training data requirements.

Key Points

To evaluate whether applying deep learning-based image enhancement to diagnostic-quality structural MRI improves downstream classification performance and training data efficiency in Alzheimer's disease.
Retrospective evaluation analyzed 2293 1.5 T 3D T1-weighted MRI scans from ADNI (1605 training, 229 validation, 459 internal test) and an external test set of 270 scans from NACC.
Scans were processed with SubtleHD deep learning enhancement, and ResNet34 and DenseNet121 architectures were trained to classify cognitively normal, mild cognitive impairment, and Alzheimer's disease using standard versus enhanced inputs.
SubtleHD enhancement significantly increased ResNet34 accuracy from 85.2% to 88.7% and macro-AUC from 0.951 to 0.968 (both p < 0.05), while DenseNet121 reached 92.2% accuracy (p = 0.18) and 0.982 macro-AUC (p = 0.29).
Classifiers trained on only 70% of the enhanced training dataset matched the diagnostic performance of models trained on 100% of the standard-of-care dataset (accuracy: 85.9%, macro-AUC: 0.942).
On external NACC validation, models trained on enhanced images achieved significantly higher accuracy (63.0% vs. 49.2%) and macro-AUC (0.819 vs. 0.679; both p < 0.05) compared to standard-trained models.

Cite This Study

Zhang et al. (2026) studied this question.

synapsesocial.com/papers/6a8019bb75c2e31742c85d2d https://doi.org/10.1002/jmri.70507

View Full Paper

Ask AI

Bookmark

Implication

Retrospective study demonstrates that deep learning-based MRI enhancement improves diagnostic classification in Alzheimer's disease cohorts, highlighting reduced training data requirements.

Deep Learning‐Based Enhancement of Already Diagnostic‐Quality MRI for Alzheimer's Disease Classification: Effects on Model Performance and Training Data Requirements

Discussion

Member takes

Implication

Key Points

Cite This Study

Deep Learning‐Based Enhancement of Already Diagnostic‐Quality MRI for Alzheimer's Disease Classification: Effects on Model Performance and Training Data Requirements

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider