New AI Algorithm Offers Robust Way to Detect Tumor Cells

A new software tool developed by researchers at UT Southwestern Medical Center utilizes artificial intelligence (AI) to detect cancer cells from digital pathology images. This provides clinicians a robust way to predict outcomes in patients.

This illustration of the ConvPath software workflow shows how the AI algorithm automatically recognizes each cell in the pathology image (upper image) as a tumor cell (orange), stromal cell (green), or lymphocyte (blue), then converts the image into a spatial map (middle image). Clusters of tumor cells are further identified as tumor regions (orange areas in the bottom image). Image Credit: UT Southwestern Medical Center.

The growth pattern of cancer, the association of cancer with the surrounding microenvironment, and the immune response of the body can be revealed by the spatial distribution of varying types of cells. However, all the cells in a pathology slide cannot be manually detected because this process is not only prone to errors, but it is also highly labor-intensive.

As there are usually millions of cells in a tissue sample, a pathologist can only analyze so many slides in a day. To make a diagnosis, pathologists usually only examine several ‘representative’ regions in detail, rather than the whole slide. However, some important details could be missed by this approach.

Dr Guanghua “Andy” Xiao, Professor, Department of Population and Data Sciences, UT Southwestern Medical Center

Xiao is also the corresponding author of the study published in the EBioMedicine journal.

According to Dr Xiao, the human brain cannot easily pick up slight morphological patterns. Hence, automatic classification of different types of cells and quantification of their spatial distributions represent a significant technical challenge when it comes to the systematic analysis of the tumor microenvironment.

ConvPath—the AI algorithm developed by Dr Xiao and his team—overcomes these barriers by utilizing AI to categorize different types of cells from the pathology images of lung cancer.

This is how it works—the new ConvPath algorithm “looks” at the cells and finds out their types depending on their appearance in the pathology images through an AI algorithm that, in turn, gets information from human pathologists.

The ConvPath algorithm successfully transforms a pathology image into a “map” displaying the spatial distributions and communications of lymphocytes (that is, the white blood cells), stromal cells (that is, the connective tissue cells), and tumor cells in the tumor tissue.

Whether tumor cells spread into stromal lymph nodes or whether they cluster well together is a factor that reveals the immune response of the body. Therefore, learning that data will allow physicians to tailor the treatment strategies and thus determine the right immunotherapy.

Eventually, the ConvPath algorithm aids pathologists to get the most precise analysis of cancer cells—in a relatively faster manner.

It is time-consuming and difficult for pathologists to locate very small tumor regions in tissue images, so this could greatly reduce the time that pathologists need to spend on each image.

Dr Guanghua “Andy” Xiao, Professor, Department of Population and Data Sciences, UT Southwestern Medical Center

Dr Xiao is a member of both the Harold C. Simmons Comprehensive Cancer Center at the UT Southwestern Medical Center and the Quantitative Biomedical Research Center (QBRC). He also has an appointment in the Lyda Hill Department of Bioinformatics.

The ConvPath software—which integrates deep learning, image segmentation, and feature extraction algorithms—can be accessed freely.

The lead authors of the study include Shidan Wang, QBRC Data Scientist II; Dr Tao Wang, Assistant Professor of Population and Data Sciences and in the Center for the Genetics of Host Defense; and Dr Donghan M. Yang, QBRC Project Manager.

Other co-authors of the study from UT Southwestern Medical Center include Dr Yang Xie, Professor of Population and Data Sciences, a Professor in the Lyda Hill Department of Bioinformatics, and Director of the QBRC; and Dr John Minna, Professor of Pharmacology and Internal Medicine and Director of the Hamon Center for Therapeutic Oncology Research.

Dr Minna holds the Max L. Thomas Distinguished Chair in Molecular Pulmonary Oncology and the Sarah M. and Charles E. Seay Distinguished Chair in Cancer Research.

The National Institutes of Health and the Cancer Prevention and Research Institute of Texas funded the research.


Tell Us What You Think

Do you have a review, update or anything you would like to add to this news story?

Leave your feedback
Your comment type