Trending Topics Mars Earth cigarettes dust storms sky watching

New Artificial Intelligence Program Helps Scientists Diagnose Tumors -- Is It Reliable?

Mar 07, 2017 10:23 AM EST

The field of artificial intelligence is expanding in ways and speeds that are almost considered inhuman -- and it's all for the benefit of mankind. Scientists have recently created an artificial intelligence (AI) program that can accurately detect the presence of tumor in brain tissues.

This research, courtesy of researchers from the University of Michigan Medical School and Harvard University, will not only classify and identify tumors, but also select the best remedy without hastily identifying one because of surgical risks.

According Nvidia, Daniel Orringer, the study's first author, said the goal of the algorithm is to help neuropathologists from diagnosis up to the surgical stage of various medical tests. He added that this will help patients get diagnosed quicker, eliminating surgeries which are riskier.

Read Also: Scientists Now One Step Closer to Creating Bones From Stem Cells  

The deep learning algorithm has created four "classifications" of tumors during its experiments. Orringer wanted to create a total of eight categories, each for one of the most common tumors that neurosurgeons encounter.

Test show that the new AI program has 90 percent accuracy, which is on par with the 90 percent rates of neuropathologists when identifying tumors themselves.

The AI works by analyzing more than 100 brain tissue samples, through which deep learning was used to check the presence of tumors. As explained by their research, published in Nature, laser imaging techniques called Stimulated Raman Histology or SRH aided this endeavor.

SRH dramatically reduces the time needed for patients to get diagnosed if they have tumors -- from 30 to 40 minutes to just three minutes. The researchers have currently tested the method on more than 370 patients, but they want to test it to as high as 500 patients.

Scientists also used the CUDA parallel computing platform (cuDNN) on its deep learning network and the Nvidia GeForce GTX 1080 GPU for its machine, which is as of today the prototype SRH model. The next stage in their plan is a larger clinical trial, as their current system is developed for research only.

Read Also: Marijuana vs Cancer: Weed Can Kill Brain Cancer, New Study Explains 

© 2017 All rights reserved. Do not reproduce without permission.

Join the Conversation

Email Newsletter
About Us Contact Us Privacy Policy Terms&Conditions
Real Time Analytics