Image analysis with deep learning to predict breast cancer grade, ER status, histologic subtype, and intrinsic subtype

By Heather D. Couture, Lindsay A. Williams, J. Geradts, S. Nyante, Eboneé N. Butler, James Stephen Marron, Charles M. Perou, M. Troester, Marc Niethammer. npj Breast Cancer, 2018

303 citations recorded by the source corpus.

RNA-based, multi-gene molecular assays are available and widely used for patients with ER-positive/HER2-negative breast cancers. However, RNA-based genomic tests can be costly and are not available in many countries. Methods for inferring molecular subtype from histologic images may identify patients most likely to benefit from further genomic testing. To identify patients who could benefit from molecular testing based on H&E stained histologic images, we developed an image analysis approach using deep learning. A training set of 571 breast tumors was used to create image-based classifiers for tumor grade, ER status, PAM50 intrinsic subtype, histologic subtype, and risk of recurrence score (ROR-PT). The resulting classifiers were applied to an independent test set (n = 288), and accuracy, sensitivity, and specificity of each was assessed on the test set. Histologic image analysis with deep learning distinguished low-intermediate vs. high tumor grade (82% accuracy), ER status (84% accuracy), Basal-like vs. non-Basal-like (77% accuracy), Ductal vs. Lobular (94% accuracy), and high vs. low-medium ROR-PT score (75% accuracy). Sampling considerations in the training set minimized bias in the test set. Incorrect classification of ER status was significantly more common for Luminal B tumors. These data provide proof of principle that molecular marker status, including a critical clinical biomarker (i.e., ER status), can be predicted with accuracy >75% based on H&E features. Image-based methods could be promising for identifying patients with a greater need for further genomic testing, or in place of classically scored variables typically accomplished using human-based scoring. Artificial intelligence can accurately predict the histological subtype and molecular marker status of breast tumors from pathology slide pictures. Heather Couture from the University of North Carolina at Chapel Hill and coworkers developed a deep learning algorithm to analyze various molecular features found within pathology slide images from breast cancer tissue samples. They trained the algorithm on a set of 571 tumors with defined grade, subtype and hormone receptor status. They then tested the model on another set of 288 breast tumors, finding that the deep-learning method predicted assorted molecular features with an accuracy exceeding 75%. Artificial intelligence applied to breast tumor images in this way could thus help clinicians identify those patients who stand to benefit most from further RNA-based genomic testing, an expensive but more definitive diagnostic tool for disease classification.

Subfields: Artificial Intelligence, Cancer Research, Radiology, Nuclear Medicine and Imaging

View Paper: https://www.semanticscholar.org/paper/2b35c6d7bbb11f6500999f05562593eb43deba9a

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Cancer Research

In Radiology, Nuclear Medicine and Imaging

Citations in the corpus

Artificial intelligence for digital and computational pathology (2023, background): “Differently, microsatellite instability — a genomic biomarker detected in 15% of colorectal cancers and for which immunotherapy treatment exists — can be detected with high”
Deep learning in cancer diagnosis, prognosis and treatment selection (2021, methodology): “eosin (H&E) sections to predict molecular subtypes of lung [49], colorectal [50], breast [51, 52] and bladder”
Deep neural network models for computational histopathology: A survey (2019, methodology): “Following up patients prospectively is a time consuming and expensive process and for this reason several studies use existing clinically validated risk models, or genomics assays as a proxy for long term outcomes; for example, the use of PAM50 scores (Veta et al., 2019; Couture et al., 2018) in bre”
Emerging role of deep learning‐based artificial intelligence in tumor pathology (2020, methodology): “Couture et al. [78] developed a DL-based algorithm to complete multiple tasks in HE-stained tissue microarray images of breast cancer, including identifying histological subtypes (ductal or lobular) with an accuracy of 94%, classifying histological grade (low-intermediate grade or high grade) with…”
Artificial intelligence in histopathology: enhancing cancer research and clinical oncology (2022, result): “Although prediction performance did not reach clinical standards, this work provided seminal proof of principle and gave rise to follow-up studies on a number of solid tumor types, including breast cancer 73 , prostate cancer 74 and gastro-intestinal tumors 61,75,76 .”
Deep learning in histopathology: the path to the clinic (2021, background): “…currently impossible to capture by sole visual inspection of histopathology slides, such as prediction of response to chemotherapy or immunotherapy, or even future events such as recurrence or survival 97–104 as well as the presence of genetic mutations 95,105–110 or molecular subtypes 30,111,112 .”