A research team from University of North Carolina Lineberger Comprehensive Cancer Center has successfully made use of a computer to accurately analyze and classify a breast cancer tumor from images using machine learning techniques.
More Insights Into the Breast Cancer Identification Study
The team used technologies that are similar to the ones used in smartphones and other smart devices that enable speech and facial recognition features. The tumor was classified on the basis of complex molecular and genomic features. As per the study’s first author, Heather Couture, a smartphone can efficiently analyze and interpret speech and find and identify faces in a photograph. Couture also is a graduate research assistant from the UNC Chapel Hill Department of Computer Science. According to the first author, the team is using a technology similar to one used in smartphone cameras for clicking images, but in a completely different way.
The study mainly consisted of making computers learn how to identify cancer tumors from about 571 images at Carolina Breast Cancer Study. This identification was then used to classify tumors in terms of estrogen receptor status, tumor grade, PAM50 intrinsic subtype, recurrence risk scores, and histologic subtype, among other factors. After this, the researchers also used a set of 288 images to test the computer’s ability differentiate between each tumor based on grade and subtype. The team of researchers also used separate tests for identification of gene expression subtypes.
Furthermore, the researchers also made software programs that can learn to perfectly predict labels from images used by using a training set. In this way, they ensured that new images can be processed in the same manner. After this, the researchers used a different set of 288 images to test the computer’s ability to distinguish features of the tumors on its own. This was done by comparing the computer’s response to the findings of a pathologist.