Parallel Computation Approach for Efficient Medical Image Analysis

摘要

A new approach of parallel processing for clinical image analysis for endoscopic and ultrasound applications are presented in this paper. Parallel computing in combination with frame selection and image processing enables real-time analysis for assisting physician's decision-making. Instantaneous diagnosis is beneficial for endoscopic and ultrasound examination as off-line detection could be inefficient. Different image partitioning and processor configuration techniques are discussed in the paper. Processors perform calculations concurrently and independently based on the images or sub-images assigned by the Inter-Process Communication Switching Mechanism. Selected color components extraction methods based on previous work have been applied to detect the abnormalities from the endoscopic images. The algorithms are implemented in shared memory architecture and tested. The results of these tests indicate that the implementation of parallel computation utilizing the n-parallelization processor configuration with adaptive-image partitioning algorithm has distinctive speed-up and improvement in run-time for usage in medical image processing and analysis.