Computers beat humans when predicting breast cancer survival

Sending samples to a pathologist to examine under a microscope is a 80-year-old way of diagnosing cancer. But a computer can make more accurate predictions of breast cancer outcomes, according to a study.

Using algorithms that can look at microscopic images and predict patient survival, Stanford scientists have taught computers to predict breast cancer prognosis.

"The computer strips away that bias and looks at thousands of factors to determine which matter most in predicting survival," Stanford professor Daphne Koller said in a statement. The program is called Computational Pathologist, or C-Path.

Traditionally, pathologists consider -- how much of the tumor has tube-like cells, the type of nuclei in the other cells, and how often they divide -- and come up with a score that determines the patient's treatment and rate of survival.

Human pathologists use a scale to judge the type and diversity of the cancer, a method that USA Today calls "subjective."

A computer pathologist, on the other hand, can look at 6,000 cellular factors and consider the environment of the cells around the cancer.

"Through machine learning, we are coming to think of cancer more holistically, as a complex system rather than as a bunch of bad cells in a tumor," Matt van de Rijn said in a statement.

The study was published in Science Translational Medicine and demonstrates why it may be necessary to treat cancer as an ecosystem.

In the future, computers could be trained to help doctors predict type of treatment or recommend which drugs patients would respond best to.

Beyond that, the aid of computer pathologists in the developing world could bring healthcare to areas that lack professional talent and open up the playing field in the treatment of breast cancer in those areas.

This post was originally published on Smartplanet.com