Content-Based CT Image Retrieval for Emphysema Using Texture and Shape Features

The growing volume of medical imaging data presents significant challenges in terms of storage, management, and utilization. Content-based medical image retrieval (CBMIR) systems serve as a supportive tool for radiologists and doctors, enabling healthcare professionals to efficiently compare current cases with previous ones, thereby facilitating accurate and timely diagnoses and treatments. This paper explores various clinically correlated features using the k-nearest neighbors algorithm to retrieve Computed Tomography images of emphysema and with no emphysema based on query content. We examine and compare several combinations of feature extraction techniques, including local binary patterns (LBP), local ternary patterns (LTP), and Zernike moments (ZMs), to evaluate retrieval performance. Our results demonstrate that the integration of LBP, LTP, and ZMs features performs well. However, the retrieval performance of ZMs alone is also quite encouraging compared to the other feature sets.