Brain Tumor Detection and Segmentation using SAM integrated YOLOv9 Scheme

In recent years, advancements in medical imaging, particularly Magnetic Resonance Imaging (MRI), have provided new avenues for tumor detection. However, MRI-based detection still faces challenges in accurately distinguishing tumor tissues from healthy brain tissue, especially given the presence of noise and variability in tumor appearance. To address these challenges, this work proposes a tumor detection technique that integrates a Segment Anything Model (SAM) with YOLOv9, referred to as SIYO. The proposed technique SIYO is analyzed extensively to use in semantic segmentation to improve the efficiency and accuracy of brain tumor detection and segmentation on MRI images from the BRATS dataset. The proposed methodology integrates innovative techniques such as the information bottleneck principle, reversible functions, Programmable Gradient Information (PGI), and the Generalized Efficient Layer Aggregation Network (GELAN) to address challenges in deep learning models. By leveraging state-of-the-art algorithms and deep learning techniques, the proposed SIYO not only enhances tumor detection but also contributes to better treatment planning and improved patient outcomes, thereby advancing the field of medical image analysis. The overall accuracy of tumor detection using SIYO is found to be 0.94.