Head and Neck Cancer Detection in Digitized Whole-Slide Histology Using Convolutional Neural Networks

Abstract

Primary management for head and neck cancers, including squamous cell carcinoma (SCC), involves surgical resection with negative cancer margins. Pathologists guide surgeons during these operations by detecting cancer in histology slides made from the excised tissue. In this study, 381 digitized, histological whole-slide images (WSI) from 156 patients with head and neck cancer were used to train, validate, and test an inception-v4 convolutional neural network. The proposed method is able to detect and localize primary head and neck SCC on WSI with an AUC of 0.916 for patients in the SCC testing group and 0.954 for patients in the thyroid carcinoma testing group. Moreover, the proposed method is able to diagnose WSI with cancer versus normal slides with an AUC of 0.944 and 0.995 for the SCC and thyroid carcinoma testing groups, respectively. For comparison, we tested the proposed, diagnostic method on an open-source dataset of WSI from sentinel lymph nodes with breast cancer metastases, CAMELYON 2016, to obtain patch-based cancer localization and slide-level cancer diagnoses. The experimental design yields a robust method with potential to help create a tool to increase efficiency and accuracy of pathologists detecting head and neck cancers in histological images.

Impact of image analysis and artificial intelligence in thyroid pathology, with particular reference to cytological aspects

“…A subset of studies compared the automated assessment of cytological and histological slides to manual assessment by a pathologist, reporting high correlation values between these two modalities with polyserial correlation coefficient, linear correlation coefficient, and area under the curve . In some studies, there was only a descriptive report of high correlation without measures . The reported correlation coefficient with visual score ranged from 0.76 to 0.83 and the areas under the curve (AUCs) were 0.909‐0.932.…”

“…Most of studies used proprietary software for image analysis. Homegrown development of machine learning algorithms was reported in only a few studies and in these cases the focus was training algorithms to perform specific tasks such as nuclear segmentation, classify follicular groups, predict the Bethesda diagnostic category in FNA samples or recognize cancer in surgical pathology specimens. As clearly explained by Dov et al, the typical size of a single WSI image is tens of gigabytes which is computationally prohibitive when developing and running AI algorithms.…”

Using a 22-layer U-Net to perform segmentation of squamous cell carcinoma on digitized head and neck histological images

“…To the best of our knowledge, this is the first work to investigate the application of fully convolutional architectures, such as U-Net, to the task of SCC segmentation in digitized histological images from head and neck SCC cancers. Previously our group worked to use a standard, patch-based CNN using this dataset [5]. While the patch-based CNN achieved a testing AUC of 0.92, our fully convolutional U-Net CNN achieved a testing AUC of 0.89.…”

“…The dataset used to implement the U-net architecture was a digitized H&N SCC dataset we previously reported [5]. The histology slide images were prepared using a standard procedure, where the tissue samples were fixed, embedded in paraffin, sectioned, and stained with haemotoxylin and eosin, and finally digitized using whole-slide scanning.…”

“…Our previous work focused on using a patch-based Inception V4 CNN to detect SCC using a dataset of 192 digitized histological images from 84 head and neck SCC patients [5]. The method we proposed in the previous work achieved an AUC of 0.91 on the testing group and an AUC of 0.92 on the testing group.…”