The amount of data in pathology images has previously been too vast for researchers to process easily, but that's changing thanks to advanced machine learning.
A group of researchers from Stanford University were able to more accurately predict lung cancer prognoses by grabbing images from the Cancer Genome Atlas from patients with the disease, and then through those train a computer software program to pinpoint characteristics in the images previously unable to be seen by the human eye, according to an announcement. Their research was published in Nature Communications.
Once the researchers could home in on those specific characteristics, they were able to figure out the cancer subtype, as well as how long a patient would live with that diagnosis.
“Ultimately this technique will give us insight into the molecular mechanisms of cancer by connecting important pathological features with outcome data,” Michael Snyder, Ph.D., a professor and chair of genetics at Stanford, said in the announcement.
He added that machine learning, like the kind used for his story, may help further advancements in cancer genomics and other areas.
In a similar study, researchers from the Regenstrief Institute and Indiana University-Purdue University in Indianapolis used about 7,000 pathology reports from the Indiana health information exchange to attempt to detect cancer cases using already available algorithms and open source machine learning tools.
They concluded that machine learning-based methods are "feasible and practical" ways to extract value from medical data.