Hybrid deep learning and meta learning for Alzheimer disease classification via multimodal radiomic and visual features
- 1Software Engineering Department, College of Engineering, University of Business and Technology (UBT)
- 2Department of Computer Science, Faculty of Computing and Information Technology, King Abdulaziz University
Abstract
This paper presents a novel hybrid multi-modal meta-learning framework that synergistically integrates five distinct meta-learning paradigms for enhanced cognitive disorder classification utilizing radiomic features as meta-knowledge to guide (contrastive language-image pre-training) CLIP feature learning. The proposed approach combines transfer meta-learning, metric learning, gradient-based meta-learning, attention-based meta-learning, and regularization-based meta-learning to achieve superior diagnostic performance through intelligent cross-modal feature fusion. Experimental validation on a dataset of 5228 samples across three cognitive states Alzheimer’s disease, cognitive normal, and mild cognitive impairment (MCI) demonstrates exceptional performance with 92% test accuracy, representing substantial improvements of 28–36% over baseline single-modality and concatenation methods. The framework achieves clinical-grade discriminative capability with AUC values exceeding 0.97 across all classes, validating its potential for real-world medical applications. The framework’s ability to distinguish between subtle cognitive states, particularly MCI identification 0.95 precision, provides valuable capability for detecting patients at risk for Alzheimer’s disease progression. This early detection capability is crucial for timely intervention and treatment planning. The achieved performance levels represent clinically significant diagnostic capability suitable for healthcare deployment. It had a reliable requirement for clinical decision support systems, while balanced performance across cognitive disorder categories enables effective early intervention strategies. This methodology demonstrates the effectiveness of meta learning techniques applied to radiometric features extracted from medical imaging data.
Related
- Multimodal patch-based deep learning for early Alzheimer’s disease classification using MRI and PET
- An uncertainty-aware multimodal deep learning framework for Alzheimer’s disease classification using MRI and electronic health records
- Boundary-aware multi-scale residual learning with hybrid radiomic–deep fusion for lung tumor segmentation and classification
- Lobar analysis of classical and deep learning radiomic features in chronic airflow limitation
- CT radiomic features detect mineral, collagen, and water contributions to cortical bone texture
BibTeX
@article{bayahya2026hybrid,
title = {{Hybrid deep learning and meta learning for Alzheimer disease classification via multimodal radiomic and visual features}},
author = {Areej Y. Bayahya and Fares Jammal},
journal = {Scientific Reports},
year = {2026},
month = jul,
doi = {10.1038/s41598-026-64370-6},
url = {https://doi.org/10.1038/s41598-026-64370-6},
note = {CC BY},
}
RIS
TY - JOUR
TI - Hybrid deep learning and meta learning for Alzheimer disease classification via multimodal radiomic and visual features
AU - Areej Y. Bayahya
AU - Fares Jammal
PY - 2026
DA - 2026/07/01
JO - Scientific Reports
DO - 10.1038/s41598-026-64370-6
UR - https://doi.org/10.1038/s41598-026-64370-6
AB - This paper presents a novel hybrid multi-modal meta-learning framework that synergistically integrates five distinct meta-learning paradigms for enhanced cognitive disorder classification utilizing radiomic features as meta-knowledge to guide (contrastive language-image pre-training) CLIP feature learning. The proposed approach combines transfer meta-learning, metric learning, gradient-based meta-learning, attention-based meta-learning, and regularization-based meta-learning to achieve superior diagnostic performance through intelligent cross-modal feature fusion. Experimental validation on a dataset of 5228 samples across three cognitive states Alzheimer’s disease, cognitive normal, and mild cognitive impairment (MCI) demonstrates exceptional performance with 92% test accuracy, representing substantial improvements of 28–36% over baseline single-modality and concatenation methods. The framework achieves clinical-grade discriminative capability with AUC values exceeding 0.97 across all classes, validating its potential for real-world medical applications. The framework’s ability to distinguish between subtle cognitive states, particularly MCI identification 0.95 precision, provides valuable capability for detecting patients at risk for Alzheimer’s disease progression. This early detection capability is crucial for timely intervention and treatment planning. The achieved performance levels represent clinically significant diagnostic capability suitable for healthcare deployment. It had a reliable requirement for clinical decision support systems, while balanced performance across cognitive disorder categories enables effective early intervention strategies. This methodology demonstrates the effectiveness of meta learning techniques applied to radiometric features extracted from medical imaging data.
SN - 2045-2322
ER -