Improving Accuracy in Medical Imaging: A YOLO-NAS and SAM Hybrid Approach for Enhanced Medical Image Segmentation

In today’s healthcare system, medical imaging plays a critical role in disease monitoring, diagnosis, and treatment planning. Our ability to see anatomical structures and identify anomalies has been completely transformed by advanced imaging modalities including computed tomography (CT), ultrasonography, and magnetic resonance imaging (MRI). The interpretation of medical images frequently necessitates careful examination, which can be laborious and prone to human mistakes because of tiny variations in tissue density, texture, and contrast. This research introduces a novel hybrid approach that combines the strengths of YOLO-NAS and the Segment Anything Model (SAM) to enhance the accuracy and efficiency of medical image segmentation. By leveraging the robust object detection capabilities of YOLO-NAS and the precision of SAM, the proposed model achieves superior performance across diverse medical imaging datasets, demonstrating it’s adaptability and potential for real-world clinical applications. In this study, we increased segmentation accuracy for the Liver-Disease dataset from 41.35% to 57.25% (Dice) and 35.22% to 55.00% (IoU), and for the Brain-Tumor dataset from 65.55% to 78.25% (Dice) and 72.55% to 85.55% (IoU) using our YOLO-NAS and SAM hybrid model. The results underscore the hybrid model’s effectiveness in improving segmentation accuracy, paving the way for advancements in automated medical imaging technologies and contributing to more reliable diagnostic and therapeutic outcomes.