Computer-aided Evaluation Of Pathological Section Ki-67 Of Appendiceal Neuroendocrine Neoplasms

Appendiceal neuroendocrine neoplasms neoplasms(ANENs)account for a large proportion of various malignant appendiceal tumors,with obscure biological behavior and low clinical specificity.One of the important diagnostic indexes is the value of Ki-67 proliferation index.Ki-67 index is extremely important for judging the malignant degree and prognosis of tumors and guiding the adjuvant treatment of tumors after surgery.At present,the calculation of Ki-67 index mainly uses manual counting,which depends on the experiences of the pathologist.The counting process is cumbersome and the workload is large.This topic aims to use computer image analysis algorithms to automatically identify and count the number of positive and negative cells from ANENs pathology images,and to calculate Ki-67 indicators.Developed a pathological image processing system to assist pathologists in comprehensive assessment and diagnosis,which has certain practical application value.The main work content includes:(1)In this study,the image acquisition of ANENs pathological slices was completed in the Baoding Hospital,Beijing Children’s Hospital.The structure processing program loop body is used to process the pathological images of different lesion areas in batches.Use different color model conversion,histogram equalization,color image filtering,sharpening and other methods for image preprocessing to improve the contrast between positive and negative cells.(2)The correlation of color components in ANENs pathologic images was high,and the histogram distribution was not obvious.In this paper,the pre-processed image is transformed into YUV color model image,and the threshold is set based on Y,U and V channels to segment the positive cells and negative cells.the corresponding shapes of the positive and negative cells in the image were measured and extracted by the mathematical morphology algorithm according to their morphology,chromaticity,area and related pathological features.(3)Aiming at the over-segmentation problem of the traditional watershed algorithm,the distance transformation is improved by using “the forced minimum technique” to segment and count the positive and negative cells in the pathological images,the recognition and segmentation accuracy of adhesion cells were improved.Comparing the experimental data with the standard number of pathologists,the average accuracy of the improved watershed algorithm for segmenting negative adhesion cells was 93.4%,and the average over-segmentation rate was reduced from 10.3% to 3.3%.The average accuracy rate of Ki-67 index for computer-assisted processing of ANENs pathological images was 93.2%,the average error rate was 6.8%,and the average time for image processing was reduced from57.4s to 29.5s,which improved work efficiency.(4)Using the graphical user interface GUI of Matlab,an on-line ANENs pathology slice Ki-67 auxiliary evaluation system was developed,mainly includinguser password login,ANENs pathological image preprocessing,feature extraction,positive and negative cell segmentation,experimental results storage and so on.The system can save the pathological image,positive cell number,negative cell number and Ki-67 index data in real time.