Improving the accuracy of reporting Ki-67 IHC by using an AI tool

Area of Science:

• Oncology
• Computational Pathology
• Medical Imaging

Background:

• Ki-67 proliferative index (PI) scoring is crucial for tumor proliferation assessment using immunohistochemical (IHC) slides.
• Manual Ki-67 PI scoring is labor-intensive, time-consuming, and prone to interobserver variability among pathologists.

Purpose of the Study:

• To develop an AI-based method for automating Ki-67 PI scoring.
• To enhance diagnostic concordance among pathologists through AI-assisted scoring.

Main Methods:

• Utilized watershed algorithm for nuclear segmentation on 440 regions of interest (ROIs) from 88 sarcoma cases.
• Three pathologists scored Ki-67 PI on ROIs with and without AI assistance.

Main Results:

• AI-assisted scoring demonstrated significant concordance among pathologists.
• Inter-pathologist discordance was reduced by 82.1% after AI assistance.

Conclusions:

• AI-based Ki-67 PI scoring effectively improves inter-pathologist agreement.
• The developed AI method offers a promising solution for objective and efficient cancer diagnosis.