Development of a Hybrid Feature Extraction Model for Efficient Content-Based Medical Image Retrieval Using Support Vector Machines

Abstract

Medical image retrieval technologies are fundamental in modern healthcare since they allow doctors to rapidly obtain relevant past cases and support accurate diagnosis. This paper offers a new hybrid feature extracting method designed to increase the efficiency and effectiveness of content-based medical image retrieval (CBMIR). By using local and global feature descriptors, the model provides a full and discriminative representation of medical images. Combining texture-based global features including Gray-Level Co-occurrence Matrix (GLCM) and Local Binary Patterns (LBP) with local features such Scale-Invariant Feature Transform (SIFT) and Speeded-Up Robust Features (SURF provides strong picture characterisation. To maximise computing speed, dimensionality is reduced using Principal Component Analysis (PCA); Support Vector Machines (SVMs) are utilised for classification thus making use of their capacity in managing high-dimensional, non-linear data. Experimental evaluation on many medical imaging datasets including MRI, CT, and X-ray modalities shows ret retrieval accuracy, reaction speed, and scalability, thereby underlining the benefits of the approach. With an average query response time of 0.8 seconds and a retrieval accuracy of 93.5%, the hybrid approach outperforms traditional single-feature-based systems. Providing a significant contribution to clinical workflows, research, and instructional uses, this development addresses critical issues in medical picture handling including variability across modalities and scalability to large datasets. The outcomes show the opportunities of hybrid feature extraction and machine learning integration to alter medical image retrieval systems, therefore paving the door for further advances employing domain-specific adaptations and deep learning techniques.