| Objective(s): Ki67 proliferation index is an important index for molecular typing,treatment monitoring and prognosis prediction of breast cancer,which runs through the whole process of diagnosis and treatment of breast cancer.At the end of 2021,the international Ki67 working group released the fifth Ki67 assessment guide for breast cancer,proposed standardized Ki67 detection process,quality control and quality assurance,established new standardized scoring methods,affirmed the positive role of artificial intelligence in assessing breast cancer Ki67,and further promoted the clinical application of breast cancer Ki67 immunohistochemistry.However,the problem of poor consistency between laboratories has not been completely solved,and AI assisted Ki67 interpretation still needs further research to find a software with high accuracy,good consistency,simple operation and strong popularization for daily pathological work.It is still controversial which method is used to evaluate the Ki67 immunohistochemical staining results of the hot spot distribution pattern and the average distribution pattern.This topic interprets and discusses the consistency of the interpretation results by selecting two regions(hot spots and average regions)and using four clinically applicable free pathological image analysis software(Qupath,Clinica Path.AIM 1.0,Path920Ki67.A1.0 semi-automatic and Path920Ki67.A2.0automatic),so as to provide a reference for clinical selection of more appropriate Ki67 interpretation,and provide a theoretical basis for accurate treatment of breast cancer.Methods: 1.Cases: 40 cases of breast cancer from 2020 to 2022 were selected from the Department of Pathology,920 Hospital of the Joint Support Force.All the patients were female,aged 32-79 years old,with a median age of 62 years old.The histological subtypes of tumors were all non-specific invasive breast cancer(IBC-NST).The histological Nottingham classification included 10 cases of Grade I(25%),26 cases of Grade II(65%),and 4 cases of Grade III(10%).Ki67 staining was performed by using a unified standard immunohistochemical staining method and scanned into digital pathological sections.2.Interpretation: Three pathologists with different seniority use Qupath,Clinica Path.AIM 1.0,Path920Ki67.A1.0semi-automatic and Path920Ki67.A2.0 automatic software to interpret respectively.Each software uses two methods,namely,hot area interpretation method and overall area average method,to interpret all 40 cases.Hotspot area method: select 4 areas with the most dense Ki67 positive tumor cells in the whole film for interpretation;The overall regional average method: according to the requirements of the 2019 version of the guidelines,the tumor area is divided into four types according to the Ki67 expression level: high,medium,low and negative,and the region is selected for interpretation in each type.3.Grouping: According to the distribution of Ki67 positive tumor cells,40 cases were divided into two groups: evenly distributed and unevenly distributed.According to the size of Ki67 proliferation index,40 cases were divided into two groups: Ki67 ≤ 30%(low expression group)and Ki67>30%(high expression group).4.Statistical analysis: SPSS software was used to analyze the intra-group correlation coefficient(ICC)of all cases interpreted by three doctors using the same software,and the consistency of the four software was compared.The ICC values interpreted by the four softwares in the evenly distributed and non-uniformly distributed group,the low expression group and the high expression group were calculated respectively.ICC value below 0.4 indicates poor consistency,0.41-0.6indicates general consistency,0.61-0.8 indicates good consistency,and greater than0.8 indicates good consistency.Results: 1.Qupath software uses the overall regional average method to interpret the ICC value of all 40 cases as 0.956,the ICC value of cases with uniformly distributed Ki67 positive tumor cells as 0.967,the ICC value of cases with uneven distribution as 0.957,the ICC value of cases with Ki67 ≤ 30% as 0.821,and the ICC value of cases with Ki67>30% as 0.919;The ICC value of all 40 cases was 0.963 by using the hot spot area method.The ICC value of cases with uniformly distributed Ki67 positive tumor cells was 0.966,the ICC value of cases with uneven distribution was 0.962,the ICC value of cases with Ki67 ≤ 30% was 0.765,and the ICC value of cases with Ki67>30% was 0.942.2.Clinica Path.AIM 1.0 software uses the overall regional average method to interpret the ICC value of all 40 cases as 0.947,the ICC value of cases with uniformly distributed Ki67 positive tumor cells is 0.976,the ICC value of cases with uneven distribution is 0.925,the ICC value of cases with Ki67 ≤30% is 0.754,and the ICC value of cases with Ki67>30% is 0.946.The ICC value of all 40 cases was 0.935 by using the hot spot area method.The ICC value of cases with uniformly distributed Ki67 positive tumor cells was 0.955,the ICC value of cases with uneven distribution was 0.919,the ICC value of cases with Ki67 ≤ 30% was0.840,and the ICC value of cases with Ki67>30% was 0.886.3.Path920Ki67.A1.0semi-automatic software uses the overall regional average method to interpret the ICC value of all 40 cases as 0.936,the ICC value of cases with uniform distribution of Ki67 positive tumor cells as 0.966,the ICC value of cases with uneven distribution as0.918,the ICC value of cases with Ki67 ≤ 30% as 0.735,and the ICC value of cases with Ki67>30% as 0.910;The ICC value of all 40 cases was 0.939 by using the hot spot area method.The ICC value of the cases with uniformly distributed Ki67 positive tumor cells was 0.957,the ICC value of the cases with uneven distribution was 0.932,the ICC value of the cases with Ki67 ≤ 30% was 0.871,and the ICC value of the cases with Ki67>30% was 0.919.4.Path920Ki67.A2.0 automatic software uses the overall regional average method to interpret the ICC value of all 40 cases as 0.933,the ICC value of cases with uniformly distributed Ki67 positive tumor cells as 0.973,the ICC value of cases with uneven distribution as 0.910,the ICC value of cases with Ki67 ≤30% as 0.815,and the ICC value of cases with Ki67>30% as 0.850;The ICC value of all 40 cases was 0.953 by using the hot spot area method.The ICC value of cases with uniformly distributed Ki67 positive tumor cells was 0.964,the ICC value of cases with uneven distribution was 0.941,the ICC value of cases with Ki67 ≤ 30% was0.833,and the ICC value of cases with Ki67>30% was 0.903.5.The time taken by the four software to interpret one case is about 1 min for the hot area method Qupath software,about 2 min for Clinica Path.AIM 1.0,6 min for the Path920Ki67.A1.0semi-automatic software,and 1 min for the Path920Ki67.A2.0 automatic software;The overall area average method Qupath software takes 10 minutes,Clinica Path.AIM1.0 takes 12 minutes,Path920Ki67.A1.0 semi-automatic software takes 14 minutes,and Path920Ki67.A2.0 automatic software takes 10 minutes.Conclusion(s): 1.The results of immunohistochemical interpretation of breast cancer Ki67 by the four softwares are statistically consistent and have their own characteristics.No matter whether Ki67 positive tumor cells are evenly distributed or not,each software has good consistency(ICC>0.8);The consistency of the four softwares in the interpretation of Ki67 high expression case group is better than that of Ki67 low expression case group.2.Path920Ki67.A2.0 and Qupath interpretation take the least time,followed by Clinica Path.AIM 1.0 and Path920Ki67.A1.0,respectively. |
PDF Full Text Request