Optimized Neural Network Architectures for Early Detection of Alzheimer\'s disease Using MRI Images

Abstract

Alzheimer\'s disease is a neurodegenerative disorder affecting millions globally, with early detection being critical for effective management. This study investigates the application of optimized neural network architectures for the automated detection of Alzheimer\'s disease from MRI scans. We employ transfer learning with pre-trained models (MobileNetV2, DenseNet121, ResNet50) and apply model optimization techniques to create compact yet accurate diagnostic tools. Throu gh quantization, we achieve significant model size reduction (75-77%) while maintaining high classification accuracy. The optimized MobileNetV2 model achieves 86.25% accuracy at only 22.5% of its original size, while our ensemble approach reaches 91.56% accuracy. These findings demonstrate that optimized neural networks can enable accessible and efficient Alzheimer\'s disease detection on resource-constrained devices, potentially extending diagnostic capabilities to underserved healthcare settings. The optimized models offer a promising path toward developing practical AI-based screening tools that can complement traditional diagnostic methods while being deployable in diverse clinical environments.

Introduction

Alzheimer’s disease (AD) is a common form of dementia characterized by progressive cognitive decline. Early detection is critical but current diagnostic methods (clinical exams, neuropsychological tests, MRI) face challenges like high cost, limited accessibility, and reliance on experts, especially in resource-limited settings. This study explores the use of optimized neural networks with transfer learning to automate AD detection from MRI images.

Three CNN architectures—MobileNetV2, DenseNet121, and ResNet50—were adapted using transfer learning and model optimization techniques such as quantization to reduce model size while maintaining accuracy. The models were trained on an augmented Alzheimer’s MRI dataset and tested in a binary classification setting (Demented vs. Non-Demented).

Results showed MobileNetV2 had the highest baseline accuracy (96.25%), with a slight drop after quantization but significant size reduction (77.5%). DenseNet121 and ResNet50 improved accuracy after quantization. An ensemble model combining predictions from all three networks further increased accuracy to 98.44% (original) and maintained 91.56% after quantization, offering a robust, lightweight diagnostic tool suitable for deployment on resource-constrained devices.

The study highlights the potential of AI to improve accessibility, cost-effectiveness, and remote screening capabilities for AD diagnosis. Future work should focus on multi-class classification, model interpretability, validation on diverse datasets, and clinical integration.

Conclusion

This research demonstrates the potential of optimized neural network architectures for efficient and accurate detection of Alzheimer\'s disease using MRI images. Through the application of transfer learning and quantization techniques, we have shown that it is possible to develop lightweight diagnostic models that maintain high accuracy while requiring significantly fewer computational resources.The quantized MobileNetV2 model achieved 86.25% accuracy at only 22.5% of its original size, while our ensemble approach reached 91.56% accuracy. These results highlight the feasibility of deploying AI-assisted diagnostic tools on resource-constrained devices, potentially extending the reach of early Alzheimer\'s detection to underserved healthcare settings. The combination of transfer learning and quantization addresses the dual challenges of model performance and deployment efficiency. As AI continues to advance in healthcare applications, optimized models like those developed in this research will play an increasingly important role in democratizing access to sophisticated diagnostic tools, particularly in resource-limited environments. Future work should focus on clinical validation, improving model interpretability, and extending the approach to multi-class classification to further enhance the clinical utility of these systems.

References

[1] Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., & Chen, L. C. (2018). \"MobileNetV2: Inverted residuals and linear bottlenecks.\" Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 4510-4520. [2] Huang, G., Liu, Z., Van Der Maaten, L., & Weinberger, K. Q. (2017). \"Densely connected convolutional networks.\" Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 4700-4708. [3] He, K., Zhang, X., Ren, S., & Sun, J. (2016). \"Deep residual learning for image recognition.\" Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 770-778. [4] Jacob, B., Kligys, S., Chen, B., Zhu, M., Tang, M., Howard, A., Adam, H., & Kalenichenko, D. (2018). \"Quantization and training of neural networks for efficient integer-arithmetic-only inference.\" Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2704-2713. [5] Litjens, G., Kooi, T., Bejnordi, B. E., Setio, A. A. A., Ciompi, F., Ghafoorian, M., van der Laak, J., van Ginneken, B., & Sánchez, C. I. (2017). \"A survey on deep learning in medical image analysis.\" Medical Image Analysis, 42, 60-88. [6] Marcus, D. S., Wang, T. H., Parker, J., Csernansky, J. G., Morris, J. C., & Buckner, R. L. (2007). \"Open Access Series of Imaging Studies (OASIS): Cross-sectional MRI data in young, middle-aged, nondemented, and demented older adults.\" Journal of Cognitive Neuroscience, 19(9), 1498-1507. [7] Jo, T., Nho, K., & Saykin, A. J. (2019). \"Deep learning in Alzheimer\'s disease: Diagnostic classification and prognostic prediction using neuroimaging data.\" Frontiers in Aging Neuroscience, 11, 220. [8] Alzheimer\'s Association. (2021). \"2021 Alzheimer\'s disease facts and figures.\" Alzheimer\'s & Dementia, 17(3), 327-406. [9] Howard, A. G., Zhu, M., Chen, B., Kalenichenko, D., Wang, W., Weyand, T., Andreetto, M., & Adam, H. (2017). \"MobileNets: Efficient convolutional neural networks for mobile vision applications.\" arXiv preprint arXiv:1704.04861. [10] Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., Berg, A. C., & Fei-Fei, L. (2015). \"ImageNet large scale visual recognition challenge.\" International Journal of Computer Vision, 115(3), 211-252. [11] Rieke, N., Hancox, J., Li, W., Milletari, F., Roth, H. R., Albarqouni, S., Bakas, S., Galtier, M. N., Landman, B. A., Maier-Hein, K., & Ourselin, S. (2020). \"The future of digital health with federated learning.\" NPJ digital medicine, 3(1), 1-7. [12] Wang, H., Shen, Y., Wang, S., Xiao, T., Deng, L., Wang, X., & Zhao, X. (2020). \"Ensemble of 3D densely connected convolutional network for diagnosis of mild cognitive impairment and Alzheimer\'s disease.\" Neurocomputing, 333, 145-156. [13] M. Shell. (2002) IEEEtran homepage on CTAN. [Online]. Available: http://www.ctan.org/tex-archive/macros/latex/contrib/supported/IEEEtran/ [14] FLEXChip Signal Processor (MC68175/D), Motorola, 1996. [15] “PDCA12-70 data sheet,” Opto Speed SA, Mezzovico, Switzerland. [16] A. Karnik, “Performance of TCP congestion control with rate feedback: TCP/ABR and rate adaptive TCP/IP,” M. Eng. thesis, Indian Institute of Science, Bangalore, India, Jan. 1999. [17] J. Padhye, V. Firoiu, and D. Towsley, “A stochastic model of TCP Reno congestion avoidance and control,” Univ. of Massachusetts, Amherst, MA, CMPSCI Tech. Rep. 99-02, 1999. [18] Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specification, IEEE Std. 802.11, 1997.

Copyright

Copyright © 2025 Tejaswini Sunil Shelke, Dr. M.A. Wakchaure. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.