Mathematical method could automatize, enhance breast cancer diagnosis

By Allen Cone  |  May 15, 2018 at 2:43 PM
share with facebook
share with twitter

May 15 (UPI) -- Computers, through mathematical models, are being taught how to diagnose breast cancer.

Researchers at Kaunas University of Technology in Lithuania are applying a deep learning method that they hope will partially automatize and enhance the accuracy of recognizing malignant lesions in breasts.

Dr. Tomas Lesmantas, a postdoctoral researcher at KUT, plans to introduce the results of the research at the 15th International Conference on Image Analysis and Recognition in Portugal from June 27-29.

"Often in cancer diagnosis oncologists rely on visual information -- the image of the tissue in question is being analyzed in order to determine the nature of the lesions," Lesmantas said in a press release. "This process is time consuming and the probability of mistake is not eliminated, which, in the case of cancer can be fatal."

The American Cancer Society says early detection is key to beating the disease, including by getting regular mammograms because most women with the disease never have symptoms.

While stage 2 breast cancer has a 93 percent five-year survival rate, stage 4 breast cancer has just a 22 percent five-year rate, according to the American Cancer Society.

Lesmantas and postdoctoral research supervisor Professor Robertas Alzbutas have analyzed 100 microscope images of breast tissue from University of Porto, Portugal, by using a capsule neural network method that was introduced by the British researcher Geoffrey Hinton, one of the founding fathers of deep learning.

They want to classify the images into the four types: non-cancerous tissue, non-malignant tumor tissue, non-invasive and invasive carcinomas.

"The early results are very promising -- we have achieved 85 percent accuracy rate," Lesmantas said.

He expects one day that computers will be taught to diagnose lesions in lungs, to recognize metastasis in lymph nodes and to localize brain tumors.

"The research is not only conducted on theoretical level, there are some cases where these methods have already been applied in clinical practice," Lesmantas said. "Even though digitalization will not replace human judgment, I believe that automatized computer diagnosis will become more common with time and will help to more accurately identify and diagnose certain types of cancer."

Related UPI Stories
Trending Stories