Loading article...
Loading article...
Stream/Area:
-
Subject:
-
Paper ID:
IJSREM63507
Volume:
10 |
5 |
May 2026
Pages:
1-8
Publication Frequency:
Monthly (12 issues per year)
License:International Journal of Scientific Research in Engineering and Management follows the Creative Commons Attribution 4.0 (CC BY 4.0) License, allowing free use and distribution with proper attribution.
Centre for Artificial Intelligence
Madhav Institute Of Technology And Science
The first point about the need for early detection of Alzheimer’s Disease (AD) relates directly to research in brain imaging. Early detection allows doctors to develop an effective treatment plan for each patient. This paper also states that as the disease progresses there will be a close relationship between the individual’s disease stages and their degree of cognitive decline. The paper introduces a new methodology called “Multi-Task Learning” (MTL), which is designed to identify a subject’s disease stage(s) and their cognitive decline levels simultaneously. The MTL methodology uses MRI data from subjects’ brains. This work created an MTL methodology based upon a pre-trained ResNet50 neural network encoder paired with SE attention blocks. These attention blocks enhance channel response across all feature maps, making the resulting features more specific to the pathological anatomy. A classification head was trained to predict four diagnostic stages of AD (Non-Demented, Very Mild Demented, Mild Demented and Moderate Demented). The second head was a regression head that estimated a subject’s clinical cognitive decline on a 0 through 3 scale. Both heads were trained together using a combination of cross entropy and mean squared error (MSE) objective functions balanced by a weighting factor λ. When tested on independent validation sets, the MTL methodology achieved 98% accuracy when classifying diagnostic stages and had a clinical cognitive decline estimation error of less than 0.11 on a 0–3 scale. These results demonstrate that other single-task models cannot achieve this level of performance. Additionally, gradient-based class activation mapping (Grad-CAM) was utilized to visualize where the model focused its attention and activation patterns aligned with those expected based on anatomically defined regions known to be impacted by AD-related pathology. In conclusion, the proposed model demonstrates that multi-task learning is a viable option for utilizing clinical neuro-imaging to improve staging performance and develop improved prognostic tools.
KEYWORDS
This study demonstrates that a single ResNet50-SE network with a dual-task loss can simultaneously perform AD staging and severity prediction from MRI slices, achieving 98.02% classification accuracy and an MAE of 0.1011 on the severity scale. The consistent performance gap over single-task baselines across all evaluated metrics confirms that multi-task learning provides genuine complementary benefits for neuro-degenerative disease analysis. Future work should address cross-dataset generalisation via domain adaptation, incorporate longitudinal MRI for trajectory modelling, and explore multi-modal inputs (PET, CSF biomarkers) to further enhance clinical utility.
[1] Y. Tian et al., “MRI-based Multi-task Decoupling Learning for Alzheimer’s Disease Detection and MMSE Score Prediction: A Multi-site Validation,” arXiv.org, 2022. [2] Z. Huo et al., “Multistage Diagnosis of Alzheimer’s Disease Based on Slice Attention Network,” in Proc. Int. Conf. Artificial Neural Networks, 2022. [3] Y. Tian et al., “MRI-based Multi-task Decoupling Learning for Alzheimer’s Disease Detection and MMSE Score Prediction: A Multi-site Validation,” Artificial Intelligence in Medicine, vol. 128, 2022. [4] S. Yu-feng et al., “M3AD: Multi-task Multi-gate Mixture of Experts for Alzheimer’s Disease Diagnosis with Conversion Pattern Modeling,” arXiv.org, 2025. [5] N. Amini et al., “Diagnosis of Alzheimer’s Disease Severity with fMRI Images Using Robust Multitask Feature Extraction Method and CNN,” Computational and Mathematical Methods in Medicine, vol. 2021, 2021. [6] P. Joysar et al., “Alzheimer’s Disease Early Prediction: A Synergistic Fusion of Deep Learning Models,” in Proc. IEEE Int. Conf. AI in Machine Vision, 2025. [7] B. Sahu et al., “Identification and Classification of Alzheimer’s Disease Based on Advanced Deep Learning Techniques,” in Proc. IEEE Int. Conf. AI and Evolutionary Technology, 2025. [8] Y. Zhang et al., “FPRes-Net: Feature Pyramid-Based Residual Network for Alzheimer’s Disease Diagnosis,” in Proc. IEEE Int. Conf. Systems, Man, and Cybernetics, 2024. [9] M. M. Rahman et al., “NeuroNet-AD: A Multimodal Deep Learning Framework for Multiclass Alzheimer’s Disease Diagnosis,” Bioengineering, vol. 12, no. 10, 2025. [10] S. Nasir et al., “Alzheimer’s MRI Classification Using Deep and Meta-Learning Models,” arXiv.org, 2024.
Publication Date
5/19/2026
Peer Review Type
Double-Blind Peer Review
Review Rounds: 2
Documents
Article Views
0
Total views since publication
PDF Downloads
2
Total downloads since publication
Geographic Reach
5
Digital Preservation: This article is permanently archived on Zenodo and complies with all data preservation standards across major indexing platforms.
Plagiarism Screening
This manuscript has been checked for plagiarism using TurnitinPlagiarismChecker software.
Conflict of Interest
The authors declare no conflict of interest related to this research work.
Funding Information
No funding information available for this research.
Related research in the same subject area
Loading similar articles...