The application of artificial intelligence in pathology reveals new insights into endometrial cancer diagnostics

The application of artificial intelligence in pathology reveals new insights into endometrial cancer diagnostics

im4MEC deep learning pipeline. (A) Images from the whole slide were segmented and cut into non-overlapping 360 μm square tiles at 40× magnification and scaled to 224×224 pixels. (B) The optimal number of tiles was sampled from each image of the entire slide to create a training dataset for the MoCo-v2 self-supervised learning model. (C) Features were extracted from all image tiles of the entire slide by use of the self-supervised learning encoder, ResNet-50, in the last layer resulting in features of size 2048. (D) The model was trained to molecularly classify the entire slide image, assigning attention scores to each tile and molecular class. Attention heatmaps display from low attention (blue) to high attention (red). (E) The first 20 attended tiles were drawn from the intended attention branch only. Predictions from HoVer-Net, a deep learning model of nuclear segmentation and classification trained on a dataset of endometrial cancer tile images, were used to calculate the counts of the three cell types and the size and shape of the tumor nuclei (Appendix 1 p 29). Subsequent analyzes described these morphological features in association with molecular classes and measured their relative feature importance with a support vector machine. MMRd = mismatch repair protein deficient. NSMP=no specific molecular profile. p53abn=abnormal expression of p53 cell tumor antigen. Credit: The Lancet’s digital health (2022). DOI: 10.1016/S2589-7500(22)00210-2

Research at the pathology department of Leiden University Medical Center (LUMC) shows the power of artificial intelligence (AI) applied to microscopic images of endometrial cancer. Dr. Tjalling Bosse’s group offers new insights that could improve the diagnosis and treatment of uterine cancer. Their results have been published in The Lancet’s digital health.

Endometrial cancer is the most common cancer of the gynecological tract. Both clinical trials and translational research are conducted at the LUMC to improve care for these patients. In recent years, the LUMC has taken a leading role in the development of a new cancer classification system based on molecular alterations, which has resulted in four subtypes of endometrial cancer. Would it be possible to predict these molecular classes, based only on microscope images?

Thousands of images

Bosse and colleagues applied artificial intelligence on microscopic images of thousands of endometrial cancer images from patients who participated. His team developed a model that reliably predicts the four molecular classes of endometrial carcinomas based on an image of a stained microscope slide (hematoxylin and eosin), which is the standard histological stain used in diagnostics for assessing tumor classification and histological subtype.

This model was not “a black box,” but through reverse engineering the researchers were able to show which features of the image were relevant to its predictions. The model has provided the team with important new insights that can be used in future studies to further improve the diagnostics, prognosis and management of endometrial cancer patients.

Emerging application of AI

“The application of AI in pathology is emerging. In this project, we studied the morphology of tumors that shared the same molecular alteration to better understand the effect these changes have on tumor appearance. With this work, the computer model it directed us to the areas inside and outside the tumor that are important,” notes Bosse.

“In cancer diagnostics, the number of variables (molecular, tumor morphology, patient data) has increased exponentially and made predicting patient prognosis more complex. Through training unbiased AI models, AI predictions they can also teach pathologists in return, for example, by identifying new morphological parameters details on microscope slide images with prognostic value,” says Sarah Fremond.

More information:
Sarah Fremond et al, Interpretable deep learning model to predict endometrial cancer molecular classification from hematoxylin and eosin-stained whole-slide images: A pooled analysis of PORTEC randomized trials and clinical cohorts, The Lancet’s digital health (2022). DOI: 10.1016/S2589-7500(22)00210-2

Provided by the University of Leiden

Citation: AI Application in Pathology Reveals New Insights in Endometrial Cancer Diagnostics (2022, Dec 16) Retrieved Dec 16, 2022 from -insights.html

This document is subject to copyright. Except in all propriety for the purpose of private study or research, no part may be reproduced without written permission. The content is provided for informational purposes only.

Add a Comment

Your email address will not be published. Required fields are marked *