Feb. 23 (UPI) -- A machine learning algorithm accurately identifies premalignant colorectal polyps on CT scans, according to a study published Tuesday by the journal Radiology.
Using radiomics, the system correctly differentiated between benign and potentially cancerous polyps more than 80% of the time, the data showed.
Radiomics is an analytical process allowing additional information to be gleaned from medical images, providing views of scans that are impossible with the naked eye.
"These results serve as proof-of-concept that machine learning-based image analysis allows the noninvasive differentiation of benign and premalignant colorectal polyps," study co-author Dr. Sergio Grosu said in a statement.
"This method works well," said Grosu, a radiologist at the University Hospital of Ludwig Maximilian University of Munich in Germany.
Colorectal cancer is among the most common causes of cancer-related death among men and women globally, research has shown.
Most colorectal cancers, including colon cancer, begin with the formation of polyps, or gland-like growths, on the mucous membrane that lines the large intestine.
These polyps typically develop over several years, and early detection and removal of precancerous polyps can help prevent these cancers from forming.
Over the past 20 years, CT colonography has emerged as a noninvasive alternative to colonoscopy in screening for colorectal cancer, according to the researchers.
The technology is comparable to colonoscopy at detecting most polyps and is effective at visualizing portions of the colon that cannot always be evaluated by colonoscopy.
However, CT colonography cannot provide a definite differentiation between benign and premalignant polyps, which is crucial in determining treatment, researchers said.
For this study, Grosu and his colleagues applied their non-invasive, radiomics-based machine learning method on CT colonography images from a group of asymptomatic patients at average risk for colorectal cancer.
The machine learning algorithm was trained on a set of more than 100 colorectal polyps in 63 patients and then tested on a set of 77 polyps in 59 patients.
In the test set, the machine learning approach enabled non-invasive differentiation of benign and pre-malignant colorectal polyps with 82% accuracy.
Additional studies with larger numbers of patients are needed to confirm the findings, the researchers said.
"[Our approach] could further improve the clinical significance of CT colonography-based colorectal cancer screening by allowing for a more precise selection of patients eligible for subsequent polypectomy," Grosu said.
However, "further refinement of the machine learning-based image analysis is necessary to achieve higher precision in polyp differentiation, as well as workflow optimization for better applicability in clinical routine," he said.