Bibliographic Information: Akindele, R. G., Adebayo, S., Kanda, P. S., & Yu, M. (2024). AlzhiNet: Traversing from 2DCNN to 3DCNN, Towards Early Detection and Diagnosis of Alzheimer’s Disease. IEEE Transactions on Biomedical Engineering, 60(12).
Research Objective: This study aims to develop a more accurate method for the early detection and diagnosis of Alzheimer's disease (AD) using a hybrid deep learning framework that leverages the strengths of both 2D and 3D convolutional neural networks (CNNs).
Methodology: The researchers designed a novel hybrid model called AlzhiNet, which integrates a 2D CNN based on ResNet-18 with a 3D CNN. The 2D module extracts detailed features from individual MRI slices, while the 3D module captures spatial and volumetric information from 3D volumes constructed from augmented 2D slices. A custom loss function, combining cross-entropy loss and mean squared error loss, optimizes the model's performance. The framework was trained and evaluated on two datasets: the Kaggle Alzheimer's Dataset and the MIRIAD Dataset.
Key Findings: AlzhiNet demonstrated superior performance compared to standalone 2D and 3D CNN models, achieving accuracies of 98.9% and 99.99% on the Kaggle and MIRIAD datasets, respectively. The study also highlighted the importance of the custom loss function and the depth of 3D volumes in enhancing classification accuracy. Furthermore, AlzhiNet exhibited greater robustness to various image perturbations, including noise and distortions, compared to a standalone ResNet-18 model.
Main Conclusions: The integration of 2D and 3D CNNs, coupled with a custom loss function and volumetric data augmentation, significantly improves the accuracy and robustness of AD classification. The AlzhiNet framework shows promise for advancing early AD diagnosis and treatment planning.
Significance: This research contributes to the growing field of AI-powered medical image analysis, specifically for AD diagnosis. The development of a robust and accurate model like AlzhiNet could lead to earlier interventions and improved patient outcomes.
Limitations and Future Research: The study acknowledges the need for further validation of AlzhiNet on larger and more diverse datasets. Future research could explore the application of this framework to other neurodegenerative diseases and investigate its potential for predicting AD progression.
Til et andet sprog
fra kildeindhold
arxiv.org
Vigtigste indsigter udtrukket fra
by Romoke Grace... kl. arxiv.org 10-04-2024
https://arxiv.org/pdf/2410.02714.pdfDybere Forespørgsler