Abdomen disease diagnosis in CT images using flexiscale curvelet transform and improved genetic algorithm.

摘要: This paper presents an abdomen disease diagnostic system based on the flexi-scale curvelet transform, which uses different optimal scales for extracting features from computed tomography (CT) images. To optimize scale of we propose improved genetic algorithm. The conventional algorithm assumes that fit parents will likely produce healthiest offspring leads to least accumulating at bottom population, reducing fitness subsequent populations and delaying solution search. In our algorithm, combining chromosomes a low-fitness high-fitness individual increases probability producing offspring. Thereby, all parent are combined with high next population. this way, leftover weak cannot damage populations. further facilitate search solution, adopts modified elitism. proposed method was applied 120 CT abdominal images; 30 images each normal subjects, cysts, tumors stones. extracted by transform were more discriminative than methods, demonstrating potential as tool diseases.