New AI Platform Accurately Identifies Lung Cancer Type, Genetic Mutations

A new artificial intelligence (AI) tool has demonstrated 97% accuracy in identifying 2 difficult-to-distinguish lung cancer types, according to new research published in Nature Medicine.

The study showed that the AI program can analyze images of lung tumors, specify cancer types, and identify altered genes driving abnormal cell growth. The program was able to accurately distinguish between adenocarcinoma and squamous cell carcinoma most of the time, which even experienced pathologists have difficulty doing without confirmatory tests, according to the study.

Additionally, the tool showed aptitude in determining whether abnormal versions of 6 genes linked to lung cancer, including EGFR, KRAS, and TP53, were present in cells. Accuracy ranged from 73% to 86% depending on the gene.

Currently, genetic tests are used to confirm the presence of mutations, but these tests typically take weeks to return results, the researchers noted. The study’s findings suggest that use of the AI program could help patients get started on targeted therapies sooner, according to the authors.

For the study, the researchers designed statistical techniques that gave their program the ability to “learn” how to get better at a task, enabling the tool to get “smarter” as the amount of training data grow. The authors trained Google’s Inception vs, a deep convolutional neural network, to analyze slide images obtained from The Cancer Genome Atlas, a database with images of cancer diagnoses that have already been determined.

According to the study results, approximately half of the small percentage of tumor images misclassified by the AI program were also misclassified by pathologists. However, 45 out of 54 of the images misclassified by at least 1 of the pathologists in the study were diagnosed correctly by the machine learning program.

“In our study, we were excited to improve on pathologist-level accuracies, and to show that AI can discover previously unknown patterns in the visible features of cancer cells and the tissue around them,” study author Narges Razavian, PhD, assistant professor in the departments of Radiology and Population Health. “The synergy between data and computational power is creating unprecedented opportunities to improve both the practice and the science of medicine.”

These findings suggest that deep-learning models can assist pathologists in the detection of cancer subtype or gene mutations, the researchers concluded. They intend to continue training the program with data until it can determine which genes are mutated in a given cancer with more than 90% accuracy.

References

Coudray N, Ocampo PS, Sakellaropoulos T, et al. Classification and mutation prediction from non-small cell lung cancer histopathology images using deep learning. Nature Medicine. 2018. https://www.nature.com/articles/s41591-018-0177-5. Accessed September 17, 2018.

Artificial Intelligence Tools Accurately Identifies Cancer Type & Genetic Changes in Each Patient’s Lung Tumor. NYU Langone’s website. https://nyulangone.org/press-releases/artificial-intelligence-tool-accurately-identifies-cancer-type-genetic-changes-in-each-patients-lung-tumor. Accessed September 17, 2018.