Deep learning-based predictive biomarker of pathological complete response to neoadjuvant chemotherapy from histological images in breast cancer

Abstract

Background Pathological complete response (pCR) is considered a surrogate endpoint for favorable survival in breast cancer patients treated with neoadjuvant chemotherapy (NAC). Predictive biomarkers of treatment response are crucial for guiding treatment decisions. With the hypothesis that histological information on tumor biopsy images could predict NAC response in breast cancer, we proposed a novel deep learning (DL)-based biomarker that predicts pCR from images of hematoxylin and eosin (H&E)-stained tissue and evaluated its predictive performance. Methods In total, 540 breast cancer patients receiving standard NAC were enrolled. Based on H&E-stained images, DL methods were employed to automatically identify tumor epithelium and predict pCR by scoring the identified tumor epithelium to produce a histopathological biomarker, the pCR-score. The predictive performance of the pCR-score was assessed and compared with that of conventional biomarkers including stromal tumor-infiltrating lymphocytes (sTILs) and subtype. Results The pCR-score derived from H&E staining achieved an area under the curve (AUC) of 0.847 in predicting pCR directly, and achieved accuracy, F1 score, and AUC of 0.853, 0.503, and 0.822 processed by the logistic regression method, respectively, higher than either sTILs or subtype; a prediction model of pCR constructed by integrating sTILs, subtype and pCR-score yielded a mean AUC of 0.890, outperforming the baseline sTIL-subtype model by 0.051 (0.839, P = 0.001). Conclusion The DL-based pCR-score from histological images is predictive of pCR better than sTILs and subtype, and holds the great potentials for a more accurate stratification of patients for NAC.

A Spatial Attention Guided Deep Learning System for Prediction of Pathological Complete Response Using Breast Cancer Histopathology Images

“…By contrast, our proposed prediction system overcomes these limitations by a human prior knowledge guided deep learning attention framework with integrative use of serial pathology slides in multiple stains, achieving much better performance with an odds ratio higher than 50. Except for our previous study on this topic (34), only one study using deep learning for this prediction problem is found to our best knowledge (14). The Inception-V3 based CNN architecture is reported to predict pCR from NAC from pre-determined tumor epithelium regions in H&E pathology images, achieving 0.84 by AUC.…”

“…WSIs can serve as a decisive image modality that is already routinely used by pathologists for clinical diagnosis and prognosis in clinical settings. While there are few published studies on pCR prediction with pathology images (12–14), the overall prediction performances are limited due to the use of a single information source, e.g., images from a single stain. With the rapid development of deep learning techniques, especially convolutional neural networks (CNNs), recent years have witnessed a significant advance not only in natural image processing research (15), but also in a wide range of biomedical image analysis tasks (16), such as abnormal region detection (17), cell segmentation (18), and survival prediction (19).…”

Application of Transfer Learning and Ensemble Learning in Image-level Classification for Breast Histopathology

“…Recent fast ensemble deep learning techniques [60,61] can be applied in whole slide medical image analysis (f.e. predicting pCR from H&E stained whole slide images [62]) to reduce time and space overheads, at the expense of certain accuracy. However, we believe that the performance improvement is much more important in this task, so we choose some usual ensemble methods.…”

Predicting neoadjuvant chemotherapy benefit using deep learning from stromal histology in breast cancer

“…More recently, deep learning (DL) methods based on medical images were used to develop novel biomarkers that were found to be predictive of the prognosis and chemotherapy response [16][17][18][19][20].…”

“…In a previous study, we proposed an image-derived biomarker for predicting pCR in breast cancer, which revealed hidden predictive information from tumor epithelium [20]. Nevertheless, the tumor-associated stroma, .…”