Deep learning trained on H&E tumor ROIs predicts HER2 status and Trastuzumab treatment response in HER2+ breast cancer

Abstract

The current standard of care for many patients with HER2-positive breast cancer is neoadjuvant chemotherapy in combination with anti-HER2 agents, based on HER2 amplification as detected by in situ hybridization (ISH) or protein immunohistochemistry (IHC). However, hematoxylin & eosin (H&E) tumor stains are more commonly available, and accurate prediction of HER2 status and anti-HER2 treatment response from H&E would reduce costs and increase the speed of treatment selection. Computational algorithms for H&E have been effective in predicting a variety of cancer features and clinical outcomes, including moderate success in predicting HER2 status. In this work, we present a novel convolutional neural network (CNN) approach able to predict HER2 status with increased accuracy over prior methods. We trained a CNN classifier on 188 H&E whole slide images (WSIs) manually annotated for tumor regions of interest (ROIs) by our pathology team. Our classifier achieved an area under the curve (AUC) of 0.90 in cross-validation of slide-level HER2 status and 0.81 on an independent TCGA test set. Within slides, we observed strong agreement between pathologist annotated ROIs and blinded computational predictions of tumor regions / HER2 status. Moreover, we trained our classifier on pre-treatment samples from 187 HER2+ patients that subsequently received trastuzumab therapy. Our classifier achieved an AUC of 0.80 in a five-fold cross validation. Our work provides an H&E-based algorithm that can predict HER2 status and trastuzumab response in breast cancer at an accuracy that is better than IHC and may benefit clinical evaluations.

convoHER2: A Deep Neural Network for Multi-Stage Classification of HER2 Breast Cancer

“…However, no benchmark dataset contains ultrasound pictures of the breast tumour location. Hence, further study was focused on detecting HER2 from tissue slices to mimic the typical HER2 status detection method better in [21][22][23][24][25][26][27]. Images of HE-stain were utilized for HER2 status detection in [21 -23].…”

Strategies for Enhancing the Multi-Stage Classification Performances of HER2 Breast Cancer from Hematoxylin and Eosin Images

“…H&E stain images were used in [ 7 , 26 ] to determine the HER2 status. U-Net was utilized in the framework in [ 27 ] to find nuclei locations in the WSI regions of the H&E-stain.…”

The evolving role of morphology in endometrial cancer diagnostics: From histopathology and molecular testing towards integrative data analysis by deep learning

“…Landmark studies have recently provided the proof of principle for the prediction of genetic mutations from H&E whole slide images by DL in several types of cancer (28-52, 54, 55), albeit more frequently in colorectal cancer (29)(30)(31)(32)(33)(34)(35)(36)(37) and breast cancer (35,(38)(39)(40)(41)(42)(43)(44). For example, the feasibility of predicting TP53 mutation status has been explored across breast, colorectal, lung, stomach, pan-gastrointestinal, bladder, and liver cancer (36,48,51,52,55).…”

“…For example, the feasibility of predicting TP53 mutation status has been explored across breast, colorectal, lung, stomach, pan-gastrointestinal, bladder, and liver cancer (36,48,51,52,55). In breast cancer, prediction of hormone receptor status (38,39,42) and homologous recombination deficiency (35,40) has also been investigated. A common task for DL has also been the prediction of microsatellite instability, particularly in colorectal cancer (29,(31)(32)(33)(34)(35)(36) and gastrointestinal cancer (40,45,46).…”

“…A number of proof-ofconcept papers showed the potential of DL models to aid the diagnosis and molecular classification of cancers (28-55) or predicting patient prognosis (56-60), by recognizing phenotypes on H&E-stained tumor slide images. Image-based DL models have been frequently trained onto colorectal cancer (29)(30)(31)(32)(33)(34)(35)(36)(37) and breast cancer (35,(38)(39)(40)(41)(42)(43)(44). However, in EC, only two studies (53, 54) so far have been published using DL for predicting one to various EC molecular alterations or gene mutations from publicly available datasets.…”