Methods of acceleration for feature extractions in medical imaging using GPU processing

Nowadays, computers can scale the workload of domain experts in many medical areas and even outmatch the human interpretation of data by extracting the important information using specific analysis methods. This information can be presented in a perceivable way to the medical experts through visualization methods. The present work contribute to medical image analysis and processing with applications in pulmonary oncology. Our contribution consists in new approaches of accelerating the investigation of radiographic images which performs processing by applying feature extraction algorithms as Canny Edge Detection and Hough Transform for lines and circles. Our research proposes new ways to reduce the computing times of Canny Edge, Hough Transform with circle and lines algorithms, either through parallel implementations with CUDA and based on the particularities of combining CUDA shared memory and texture rebinding memory for the investigated images. Our result demonstrates that the proposed strategy is time efficient by reducing the computation times between 75% and 99%.