MEDICAL IMAGE ANOMALY DETECTION USING DEEP LEARNING: A HYBRID CNN-VAE APPROACH
DOI:
https://doi.org/10.34218/IJCET_15_02_025Keywords:
Medical Imaging, Anomaly Detection, Deep Learning, CNN, VAE, Unsupervised Learning, Image Reconstruction, Medical DiagnosticsAbstract
The rapid evolution of deep learning has significantly enhanced medical image analysis, enabling more accurate detection of anomalies such as tumors, lesions, and rare pathologies. This research proposes a hybrid architecture combining Convolutional Neural Networks (CNNs) and Variational Autoencoders (VAEs) to detect anomalies in medical images. By leveraging the spatial feature extraction capabilities of CNNs and the generative modeling power of VAEs, the model excels in unsupervised anomaly detection tasks. Evaluated on datasets such as Brain MRI and Chest X-ray, our hybrid model demonstrates superior accuracy and robustness compared to baseline models. This work contributes an interpretable, scalable solution to support radiologists in early disease diagnosis.
References
Singh, R., Kalusivalingam, A. K., & Bose, M. (2022). Enhancing anomaly detection in histopathological images using convolutional neural networks and variational autoencoders. European Advanced AI Letters. http://www.eaaij.com/index.php/eaaij/article/view/52
Hassanaly, M., Perumal, K., Balachandran, P., & Kakade, A. (2022). Unsupervised brain PET anomaly detection using pseudo-healthy image synthesis. Medical Image Analysis, 80, 102469. https://doi.org/10.1016/j.media.2022.102469
Kascenas, A., Zhang, M., & Rundo, L. (2022). Brain anomaly detection using variational autoencoders: A comparative study. Journal of Biomedical Imaging and Bioengineering, 9(1), 15–25. [No public link available; please check institutional access]
Yang, L., Zong, X., Li, Q., & Liu, J. (2023). Deep learning methods for dynamic anomaly detection in CT angiography time-series. Computerized Medical Imaging and Graphics, 101, 102155. https://doi.org/10.1016/j.compmedimag.2022.102155
Sutradhar, A. (2023). Automated detection of diabetic retinopathy using CNNs with image augmentation. Biomedical Signal Processing and Control, 80, 104282. https://doi.org/10.1016/j.bspc.2023.104282
Zhang, J., Liu, C., & Zhang, H. (2023). Enhanced medical image reconstruction using CNN-decoder-based variational autoencoders. Journal of Intelligent Systems, 32(5), 523–540. https://doi.org/10.1515/jisys-2023-0052
Lu, Y., Cheng, H., & Wang, S. (2023). A heterogeneous CNN-VAE model for cross-modal brain MRI anomaly detection. IEEE Transactions on Neural Networks and Learning Systems. https://doi.org/10.1109/TNNLS.2023.3245367
Duong, T., Phan, H., & Nguyen, M. (2022). A study of VAE design strategies for unsupervised anomaly detection in medical imaging. Pattern Recognition Letters, 155, 10–18. https://doi.org/10.1016/j.patrec.2021.12.001
Gao, X., Chen, R., & Lin, Y. (2022). A VQ-VAE and SVDD combined framework for localizing lung CT anomalies. Artificial Intelligence in Medicine, 125, 102144. https://doi.org/10.1016/j.artmed.2022.102144
Fan, Z., Huang, J., & Wang, P. (2023). VAE-GAN for multi-modal time-series anomaly detection in healthcare monitoring. IEEE Access, 11, 11201–11213. https://doi.org/10.1109/ACCESS.2023.3245432
Kumari, S., Anand, R., & Srivastava, M. (2023). Explainable hybrid CNN-VAE model for medical image anomaly detection using Grad-CAM visualization. Journal of Digital Imaging, 36(3), 589–601. https://doi.org/10.1007/s10278-023-00731-z
Downloads
Published
Issue
Section
License
Copyright (c) 2024 Priya Balasubramanian (Author)
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
