CLEVELAND, Sept. 16 (UPI) -- For doctors, it can be difficult to tell the difference between brain cells killed by cancer treatment and recurrent brain cancer using patient's MRI scans. Computers, however, appear to be far more accurate.
A computer program accurately diagnosed dead cells and brain cancer at nearly twice the rate of doctors, suggesting similar software could be used to improve treatment selection, according to a study published in the American Journal of Neuroradiology.
Radiation used as part of brain cancer treatment causes radiation necrosis, or brain cell death, and requires treatment very different from that used in cases of recurrent cancer, making correct diagnosis vital.
For the most part, these diagnoses are conducted using MRI scans, but in some cases must be confirmed using costly, invasive, risky brain biopsies that have high rates of morbidity and mortality. If there is an easier way, researchers say, the benefits for patients are huge.
Related
"One of the biggest challenges with the evaluation of brain tumor treatment is distinguishing between the confounding effects of radiation and cancer recurrence," Pallavi Tiwari, an assistant professor of biomedical engineering at Case Western Reserve and leader of the study, said in a press release.
For the study, the researchers designed software using machine learning algorithms and radiomics, or features extracted from images using computer algorithms, teaching the system to identify radiomic features telling the difference between cancer and dead cells.
To compare the computer's ability to diagnose cancer to that of doctors looking at MRIs, the researchers had two doctors and the computer analyze MRI scans for 15 patients. One doctor got seven diagnoses correct and the other got eight correct -- but the computer was right 12 out of 15 times.
"What the algorithms see that the radiologists don't are the subtle differences in quantitative measurements of tumor heterogeneity and breakdown in microarchitecture on MRI, which are higher for tumor recurrence," Tiwari said.
The next step will be to further validate the accuracy of the algorithm using larger collections of scans, with the hope of developing a system doctors can use to assist in diagnosis and treatment when making treatment decisions with patients.
