A computer that analyzes images of breast cancer tumors can predict patient survival better than human pathologists do, according to a study published in Science Translational Medicine.

The computer "provides information above and beyond what the physician provides, using the same data," said Daphne Koller, senior author of the study and professor of computer science at Stanford University.

Currently, pathologists analyze breast cancer by examining the tumor under a microscope and giving it a score based on an established scale. Patient prognosis and treatment are based on this information. This process has not significantly changed in 80 years, said Koller.

The computer, called Computational Pathologist (C-Path), also analyzes microscopic images of the tumor, but was able to pinpoint additional factors to consider when determining patient survival. In the current study, C-Path analyzed over 6,000 cellular factors and discovered that certain characteristics of cells surrounding the cancer are important in predicting survival.

"We found 11 [factors] ultimately showed the most robust association with survival," said Dr. Andrew H. Beck, lead author of the study when he was at Stanford (now he's assistant professor of pathology at Harvard Medical School in Boston).

Researchers have not begun commercial development of the method, Koller said.