A Genetic-Algorithm-Based Temporal Subtraction for Chest Radiographs

Abstract

To assess pathological chest change, radiologists compare the same patient's chest radiographs taken at different times. Supporting radiologists' diagnostics, temporal-subtraction images constructed from the previous and current radiographs have enhanced the visualization of pathological change. This paper presents a genetic-algorithm-based temporal subtraction for chest radiographs. First, we extract ribs from previous and current images and use them for global matching of the two images. Then, we divide the lung area in the current image into many subareas. For individual subarea, we use the genetic algorithm for local matching to find its corresponding area in the previous image efficiently. Results demonstrated that pathological change were accurately enhanced in temporal-subtraction images without major misregistration artifacts, accurately visualizing of pathological change and proving useful in improving radiologists, diagnostic performance.