Biopsy Image Stitching And Error Correction Technology

In the process of remote consultation, pathological images analysis is one of the main references for doctor to diagnosis. The accuracy of pathological images will have effect on the accuracy of diagnosis, and which is affected by the stitching quality of the pathological images. However, in current stage, remote pathology consultation is only treated as a reference source. From the technical view, it is due to the limitations of image mosaic technology. During the stitching process, slightly mismatch may be occurred. Because of the possible existing errors, some internal factors may be ignored, and then the patients may be misdiagnosed. Therefore, in order to improve pathological diagnosis based on the pathological images, it is necessary to study the height accurate stitching approach of pathology images.This paper studies and summarizes the image mosaic theories and methods of image stitching, and the importance for the remote consultation. The errors classification and the nature of stitching errors in the image mosaic are analyzed and described. The expected value of the accumulation of errors is analyzed. According to the expected values of images stitching quality before and after the accumulation of errors, the effectiveness is theoretically proved. Combined with the characteristics of the pathological images, it is presented of the differences of image mosaic between pathological images and common images. The process of collecting pathological images is explained. An image stitching algorithm based on the SIFT is proposed, and then the LOF anomaly detection algorithm is used to remove the mismatching points. The method is tested with pathological images, and the results display that the method is correct and efficient. Focusing on the accumulation errors, an image mosaic correct method is proposed base on the cell mosaic theory. According to the position relationship of the stitched images, modified formulas of stitching parameters are proposed, then the simplify rule is proposed. The method is tested with sixteen images, four by four. The results of correction present that the mismatch errors are restrained little than one pixel. The technology is suitable for high-accurate situation. It is a high-precision image stitching method, with high practical value. The pathological images with high-accuracy and high-quality have practical value. It not only can enhance the authority of doctors for remote diagnosis and treatment information, but also can provide convenience for the local hospitals to research the previous pathological images.