A new AI program developed by researchers at Peter the Great St. Petersburg Polytechnic University in Russia is able to identify and mark lung nodules on CT scans in under 20 seconds!
The new program is called Doctor AI-zimov and works by spotting the lung nodules on segmented CT scans, and using a recently patented “chord” method to connect a series of points randomly drawn onto each nodule. Doctor AI-zimov places the nodule into a virtual cube where it then composes perpendicular lines from the edges of the cube to the surface of the nodule. By the end of the process, the AI software has resulted in the generation of simple histograms that reflect the shape and structure of the nodules, and is able to then use the histograms to analyze each nodule – from its size to potential malignancy – finally sending its results to clinicians for review.
Since the histograms are so much smaller than an average CT scan, it allows Doctor AI-zimov to complete its tasks within seconds. And according to lead researcher Lev Utkin, PhD, the AI is even capable of detecting lung nodules as small as 2 mm!
Utkin explained: “Initially, we set up an algorithm to search for nodules starting from 6 mm, because radiologists themselves start the treatment of tumors of this size, but the system is so smart that it was able to find nodules of even smaller size.”
Doctor AI-zimov was trained by Utkin and his colleagues using 1,000 CT scans from the Lung Image Database Consortium (LIDC), and was also tested on CT scans of 60 patients at their institution’s oncology center to favorable results.
Looking to the future, the researchers are planning to continue open testing of the AI software at the St. Petersburg Clinical Research Center. One day they hope Doctor AI-zimov will be able to analyze both x-rays and ultrasound scans of various organs. We can certainly hope!