Deep Learning-Based Assessment of Histological Grading in Breast Cancer Whole Slide Images
Authors: Esther Albertín Iván Lalaguna Javier García María Blanca Cimadevila Ana Calvo Valery Naranjo
Date: 30.11.-0001
Abstract
Histological grading of breast cancer samples is critical for determining a patient’s prognosis. Automatic grading of pathological cancer images promotes early diagnosis of the disease, as it is a long and tedious task for health professionals. In this paper, we propose an algorithm capable of predicting each component of the histological grade in Hematoxylin and eosin (H&E)-stained Whole-Slide Images (WSIs) of breast cancer. First, the WSI is split into tiles, and a classifier predicts the grade of both tubular formation and nuclear pleomorphism. Experiments are carried out with a proprietary database of 1,374 breast biopsy DICOM
WSIs and evaluated on an independent test set of 120 images. Our model allows us to accurately classify the constitutive components of the histological grade both for all tumour samples and only-invasive samples.
BIB_text
title = {Deep Learning-Based Assessment of Histological Grading in Breast Cancer Whole Slide Images},
pages = {355-364},
keywds = {
Digital pathology · breast cancer · histologic grade · deep learning
}
abstract = {
Histological grading of breast cancer samples is critical for determining a patient’s prognosis. Automatic grading of pathological cancer images promotes early diagnosis of the disease, as it is a long and tedious task for health professionals. In this paper, we propose an algorithm capable of predicting each component of the histological grade in Hematoxylin and eosin (H&E)-stained Whole-Slide Images (WSIs) of breast cancer. First, the WSI is split into tiles, and a classifier predicts the grade of both tubular formation and nuclear pleomorphism. Experiments are carried out with a proprietary database of 1,374 breast biopsy DICOM
WSIs and evaluated on an independent test set of 120 images. Our model allows us to accurately classify the constitutive components of the histological grade both for all tumour samples and only-invasive samples.
}
isbn = {978-981-99-3311-2},
date = {0000-00-00},
}
