Real-Time Multi-Class Alzheimer’s Detection from Structural MRI via Lightweight Deep Transfer Learning
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Alzheimer disease is a progressive and sluggish brain disorder that is manifested by damage to memory, cognitive abilities, and the brain structure. The fact remains that it is hard to diagnose the disease on time as the anatomical changes that occur between normal people and those with very mild, mild, and moderate dementia are insignificant, and they tend to be confused with the typical changes that happen with age. Alzheimer’s Disease methodological techniques of diagnosis, such as neuropsychological tests and the analysis of MRI images by hand, are characterized by low sensitivity, which does not allow them to be used in large-scale screening. In a bid to overcome them, the paper introduces a lightweight architecture of a deep learning stage classifier in four-class Alzheimer classification based on structural MRI data. The T1-weighted two-dimensional slices are standardised and trained on to a class-balanced loss function to reduce data imbalance to train a MobileNetV2-based architecture. The suggested model learns discriminative spatial features automatically and yields stage-wise dementia probabilities accordingly, which can be able to give timely and reliable assistance to clinical decision-making. The system is also suitable to apply in the hospital setting and also remote screening and tele-neurology due to the low computational complexity of the system. The next stage in future work will be to combine attention mechanisms with multimodal imaging modalities and mobile edge computing in order to be able to continuously monitor patients.
Fields of Study & Subjects
Dementia and Cognitive Impairment Research
Medical Image Segmentation Techniques
Artificial Intelligence in Healthcare
Publication details
- Journal / source
- International Conference on Computing, Communication, Control and Cyber-Physical Systems (I5CPS)
- Pages
- 1–6
- Publication date
- 26 February 2026
- Language
- EN