Revolutionizing MRI Brain Tumor Detection Through GAN Augmentation and Smart Algorithms.
Mrs Rajeshwari1, Venuchandan M G2, Veerendra N M3
1 Artificial Intelligence & Data Science, CIT, Gubbi
2Artificial Intelligence & Data Science, CIT, Gubbi
3Artificial Intelligence & Data Science, CIT, Gubbi
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - The accurate detection and classification of brain tumors in Magnetic Resonance Technology is critical to accurate diagnosis and therapy planning. However, the absence of annotated MR imaging datasets presents a significant challenge for advanced machine learning models, which require a range of data to achieve high accuracy and generalizability. This study provides a novel data augmentation method to enhance MR images for improved tumor identification by combining noise-to-image and image-to-image Generative Adversarial Networks (GANs). Noise-to-image GANs generate synthetic MR images from random noise, expanding the dataset with a range of anatomical variations, whereas image-to-image GANs refine and enhance existing MR images, highlighting important features relevant to tumor detection.
The enlarged set of input data was taken to teach CNN, which was then correlated to a baseline model built solely on the original dataset. The GAN-augmented model beat the baseline cancer sorting and division tasks, as measured by parameters such as accuracy, recall, F1-score, and Dice coefficient. An ablation study verified the independent contributions of noise-to-image and image-to-image GANs, demonstrating that both types of augmentation operate together to improve model performance.
This work illustrates the promise of GAN-based augmentation in medical imaging by giving a realistic solution to data restrictions and improving the robustness of deep learning models in brain tumor detection. This technology provides the way for more accurate and reliable based on artificial intelligence tools for diagnosis that can help medical practitioners make clinical decisions by integrating alternative GAN techniques.
Key Words: GAN-based augmentation, Deep learning, Image augmentation techniques, Generative adversarial networks (GANs)
