Artificial intelligence reveals features associated with breast cancer neoadjuvant chemotherapy responses from multi-stain histopathologic images

Abstract

Advances in computational algorithms and tools have made the prediction of cancer patient outcomes using computational pathology feasible. However, predicting clinical outcomes from pre-treatment histopathologic images remains a challenging task, limited by the poor understanding of tumor immune micro-environments. In this study, an automatic, accurate, comprehensive, interpretable, and reproducible whole slide image (WSI) feature extraction pipeline known as, IMage-based Pathological REgistration and Segmentation Statistics (IMPRESS), is described. We used both H&E and multiplex IHC (PD-L1, CD8+, and CD163+) images, investigated whether artificial intelligence (AI)-based algorithms using automatic feature extraction methods can predict neoadjuvant chemotherapy (NAC) outcomes in HER2-positive (HER2+) and triple-negative breast cancer (TNBC) patients. Features are derived from tumor immune micro-environment and clinical data and used to train machine learning models to accurately predict the response to NAC in breast cancer patients (HER2+ AUC = 0.8975; TNBC AUC = 0.7674). The results demonstrate that this method outperforms the results trained from features that were manually generated by pathologists. The developed image features and algorithms were further externally validated by independent cohorts, yielding encouraging results, especially for the HER2+ subtype.

A population-level computational histologic signature for invasive breast cancer prognosis

“…High average nuclear chromatin clumping is an adverse prognostic biomarker, while the TlL-to-TIL variation in chromatin clumping is favorable. The biological signi cance of these ndings is unclear but could represent different lymphoycte subsets and degrees of differentiation 75 . Future work using IHC/ISH-based analysis would address this; we focused on H&E to ensure applicability in routine settings 76 .…”

Computational pathology to improve biomarker testing in breast cancer: how close are we?

“…Hence, artificial intelligence-powered spatial TILs analysis could assess TIL densities in the cancer area and surrounding stroma of TNBC, Moreover, TILs density scores could predict pCR after NAC (Lee et al ., 2022). These new technological advances facilitated the applications of computational pathology for clinical diagnosis and prognosis (Whiteside, 2008); however, predicting clinical outcomes from pre-treatment histopathologic images remains a challenging task, limited by the poor understanding of tumor immune micro-environments (Huang et al ., 2023).…”

“…An attempt to this has been done by Huang et al . where the feasibility of artificial intelligence-based algorithms to predict neoadjuvant chemotherapy (NAC) outcomes in human epidermal growth factor receptor 2 (HER2)+ and triple-negative breast cancer (TNBC) patients using hematoxylin and eosin (H&E) and multiplex immunohistochemistry (PD-L1, CD8+, and CD163+) images was tested by using automatic feature extraction approaches (Huang et al ., 2023).…”

“…Modern medicine necessitates the ability to forecast cancer patient outcomes using computational pathology has been made possible by improvements in computer tools and algorithms; however, predicting clinical outcomes from pre-treatment histopathologic imaging is not feasible. An attempt to this has been done by Huang et al where the feasibility of artificial intelligence-based algorithms to predict neoadjuvant chemotherapy (NAC) outcomes in human epidermal growth factor receptor 2 (HER2)+ and triple-negative breast cancer (TNBC) patients using hematoxylin and eosin (H&E) and multiplex immunohistochemistry (PD-L1, CD8+, and CD163+) images was tested by using automatic feature extraction approaches (Huang et al, 2023).…”

Leveraging medical Twitter to build a visual–language foundation model for pathology AI

“…Recent advances in artificial intelligence (AI) algorithms in computational pathology can help to distinguish cell or tissue types, generate diagnoses, and retrieve relevant images from routinely stained hematoxylin and eosin (H&E) images [1][2][3][4][5]. However, computational progress is heavily bottlenecked by the insufficient number of well-annotated pathology images, as the scale of publicly available annotated pathology datasets falls below the common standards in other AI domains [6].…”