A GA-based image alignment approach for tissue image matching

摘要

Tissue image matching is important in tissue microarray (TMA) processing, during which massive patient samples are embedded in a single paraffin-based block for simultaneous analysis of pathological features. Prior to TMA processing, the images of the donor block and the corresponding slide must be aligned to determine the desired punching locations. This study developed a genetic algorithm (GA)-based image alignment approach to image superimposition. The similarity between the two images is first evaluated by calculating the dissimilarity area of their binary images using logical operators. The GA is then performed to obtain the optimal translation and rotation parameters for superimposing one image onto another. Experimental results revealed that with both crossover and mutation rates of 0.9, the proposed approach can yield a parameter combination that achieves 100% success of tissue image matching with minimum alignment error.