Novel computer-aided systems for interpreting immunohistochemistry (IHC) results in breast cancer based on deep learning algorithms: A systematic review

Abstract

Breast cancer is a prevalent disease worldwide and the accurate diagnosis and prognosis of breast cancer are essential for the development of effective treatment plans. Pathology remains the gold standard for diagnosis and prognosis but with limitations such as time-consuming manual scoring and some error-prone results. Recently, deep learning techniques, especially convolutional neural networks (CNN), have been proposed for the interpretation of immunohistochemistry (IHC) results in breast cancer. The objective of this systematic review is to critically assess the existing literature on computer-aided systems for the interpretation of IHC results in breast cancer based on deep learning algorithms. We included studies with models that use novel approaches such as deep learning for quantitative measurements of immunohistochemically stained Ki-67, ER, PR, and HER2 images. We systematically searched PubMed, Scopus, and web of science up to September 2022. 15 studies (seven HER2, seven Ki67, and one ER/PR scoring studies) met our inclusion criteria. Various AI-based assays have been developed for different applications in breast pathology, including diagnostic and prognostic applications, as well as predictive values and responses to treatment. These algorithms have shown promise in improving the accuracy of breast cancer diagnosis and prognosis. It is essential to consider the differences in training and inter-observer variability while designing tools, and there is an urgent need to integrate the detection and analysis of various biomarkers at the same place and time to facilitate the formation of patients' reports and treatment.

Keywords

• breast cancer
• deep learning
• computer-aided systems
• IHC