Skip to content
Stanford University School of Medicine

Predicting lung cancer type and patient survival with computers

My father taught me to play chess when I was in grade school and I can recall many quiet evenings playing with him in the living room after dinner. So we had a lot to talk about when IBM's supercomputer Deep Blue defeated world chess champion Garry Kasparov in 1996. Kasparov went on to win four out of six games in the first match, but that first defeat marked the bittersweet end of an era and hinted that computers could prove to be better than humans in other tasks as well. I recalled this event as I was working on one of our recent press releases describing work conducted in the laboratories of geneticist Michael Snyder, PhD, and radiologist Daniel Rubin, MD. The researchers, led by graduate student Kun-Hsing Yu, MD, devised a way to teach computers to teach themselves how to assess thin sections of lung cancer tissue in an effort to make the process faster and more accurate. They published their findings today in Nature Communications. As Snyder explained in our release:

Pathology as it is practiced now is very subjective. Two highly skilled pathologists assessing the same slide will agree only about 60 percent of the time. This approach replaces this subjectivity with sophisticated, quantitative measurements that we feel are likely to improve patient outcomes.

In short order Yu and his colleagues found the computers were able to pinpoint thousands of minute differences between normal and cancerous tissue. They also could tell between two main hard-to-discern lung cancers types -- adenocarcinoma and squamous cell carcinoma -- as well as predict patient survival time better than the traditional approach of classifying tumors by grade and stage. From our release:

Identifying previously unknown physical characteristics that can predict cancer severity and survival times is also likely to lead to greater understanding of the molecular processes of cancer initiation and progression. In particular, Snyder anticipates that the machine-learning system described in this study will be able to complement the emerging fields of cancer genomics, transcriptomics and proteomics. Cancer researchers in these fields study the DNA mutations and the gene and protein expression patterns that lead to disease. "We launched this study because we wanted to begin marrying imaging to our 'omics' studies to better understand cancer processes at a molecular level," Snyder said. "This brings cancer pathology into the 21st century and has the potential to be an awesome thing for patients and their clinicians."

Previously: Automating genetic analysis could speed diagnosis of rare diseases, Exploring the "ridiculously exciting" opportunities for artificial intelligence and Tracking resistance mutations in lung cancer patients

Photo by Adrian Askew

Popular posts