Effect of artificial intelligence-assisted teaching on diagnostic accuracy of prostate cancer among resident pathologists:a prospective self-controlled study
ZHI Xingqi
WU Jingpeng
NIU Yun
HAN Yutong
GUO Jia
ZHONG Dingrong
Abstract：Objective To quantitatively evaluate the effect of artificial intelligence(AI)-assisted teaching on the diagnostic accuracy of prostate cancer among resident pathologists.Methods This was a prospective self-controlled study.A total of 249 hematoxylin-eosin(H&E)slides were obtained from 51 patients who underwent prostate biopsies at China-Japanese Friendship Hospital between May and July 2025.Four resident pathologists in standardized training participated in a three-stage evaluation:(1)initial diagnosis;(2)review and AI-assisted teaching;and(3)post-teaching diagnosis.Slides from May were used for stages(1)and(2),and slides from June were used for stage(3)to minimize memory bias.The gold standard for pathological diagnosis of prostate biopsy was determined by two senior pathologists using histology and confirmatory immunohistochemistry.Diagnostic accuracy,sensitivity,and specificity were compared between stages(1)and(3).Results After AI-assisted teaching,the third-year resident's accuracy increased from 83.3%to 94.6%,and sensitivity increased from 57.9%to 100%;the second-year resident's accuracy increased from 73.9%to 92.8%,and specificity increased from 72.0%to 96.2%(all P<0.05).After AI-assisted teaching,diagnostic performance of first-year resident pathologists was significantly improved.Resident C showed overall enhancement in diagnostic accuracy(from 54.35%to 69.37%),sensitivity(from 81.58%to 90.63%),and specificity(from 44.00%to 60.76%).Resident D shifted from a defensive"all-positive"pattern(accuracy 27.54%,specificity 0%)to a more balanced diagnosis(accuracy 65.77%,specificity 67.09%).Conclusion Preliminary quantitative data demonstrates that short-term AI-assisted teaching can significantly improve the diagnostic accuracy,sensitivity,and specificity of junior pathologists in prostate cancer.It is operable and worthy of promotion,and further validation in larger cohorts is needed.
Keywords：prostate cancerartificial intelligence-assisted teachingpathological diagnosisstandardized training of residentsdiagital pathology
Publication Date：2026-01-28
Online Publishing Date：2026-03-31(First online date of this platform, not the publication date of the document)
Pages：6( 60-65 )