ML provides pathologist-level classification of lung cancer slides

Dartmouth researchers have developed a deep neural network to classify different types of a common lung cancer that they contend is as accurate as pathologists.

In a study published on Monday in Scientific Reports, the research team found that the machine learning model performed at an accuracy level on par with three practicing pathologists.

“All evaluation metrics for our model and the three pathologists were within 95 percent confidence intervals of agreement,” concludes the study’s authors. “If confirmed in clinical practice, our model can assist pathologists in improving classification of lung adenocarcinoma patterns by automatically pre-screening and highlighting cancerous regions prior to review.”

Whole Slide Images Lung Cancer-CROP.jpg

Also See: New tech could change how pathologists view tissue samples

In addition, the authors—whose code is publicly available on GitHub to promote new research and collaborations—contend that their approach “can be generalized to any whole-slide image classification task.”

Researchers at Dartmouth’s Norris Cotton Cancer Center, led by Saeed Hassanpour, assistant professor of biomedical data science and adjunct assistant professor of computer science at Dartmouth’s Geisel School of Medicine, developed the machine learning model to classify different types of lung adenocarcinoma on histopathology slides.

“If validated through clinical trials, our neural network model can potentially be implemented in clinical practice to assist pathologists,” says Hassanpour, who is also a member of the Cancer Population Science Research Program at Norris Cotton Cancer Center. “Our machine learning method is also fast and can process a slide in less than one minute, so it could help triage patients before examination by physicians and potentially greatly assist pathologists in the visual examination of slides.”

Besides testing the model for classifying lung adenocarcinoma in a clinical environment, researchers also intend to apply the method to other challenging histopathology image analysis tasks in breast, colorectal, and esophageal cancer.

For reprint and licensing requests for this article, click here.