Lightweight GAN-VAE hybrid model for real-time brain tumor detection on edge devices
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
Detecting brain tumors is a vital aspect of medical imaging, as timely and precise diagnosis can greatly enhance patient prognosis and treatment outcomes. Traditional deep learning models, while effective, often require substantial computational resources, limiting their deployment in real-time and resource-constrained environments. This study introduces an efficient hybrid framework that integrates Variational Autoencoders (VAEs) with Generative Adversarial Networks (GANs) for brain tumor detection, specifically optimized for deployment on edge devices like the Raspberry Pi. The proposed framework utilizes the latent feature learning strength of VAEs alongside the image generation capability of GANs, combined with a lightweight classifier to enable fast and efficient inference. Experimental evaluations show that the model attains a classification accuracy of 93.2% and a Dice score of 0.89, while maintaining a compact size of 4.2 MB and an inference time of less than one second per image. In comparison with existing state-of-the-art approaches, the VAE–GAN hybrid achieves an optimal balance between accuracy and deployability. Its effective generalization through synthetic data and suitability for low-power hardware highlight its potential for real-time clinical use in resource-constrained environments.
Fields of Study & Subjects
Brain Tumor Detection and Classification
Medical Image Segmentation Techniques
AI in cancer detection
Publication details
- Journal / source
- International Conference on Optical and Wireless Technologies (OWT)
- Pages
- 33–33
- Publication date
- 30 March 2026
- Language
- EN