Machine learning is making a difference in the healthcare sector by identifying patients who have colorectal cancer. A new study reveals that machine learning can also give a clear picture of the severity of the disease and survival.
The non-invasive technique adds to the recent progress in technology that examines circulating tumor DNA (ctDNA) and can be useful in spotting colorectal cancer risk at early stages.
Like other melanomas, colorectal cancers are treatable, if detected at an early stage, especially before they are metastasized to other tissues.
Colonoscopies have been a technique used to diagnose colorectal cancers, but the invasive technique most times is uncomfortable and can lead to other complications that scare patients and leaves them less willing to undergo the necessary screening procedures.
Meant for a study published in the journal Science Translational Medicine, researcher Huiyan Luo from the University Cancer Center in China and colleagues used machine learning techniques to build a less invasive diagnostic technique to detect colorectal patients.
The technology developed by them screens for methylation markers, which are DNA modifications often found in tumors.
More about the innovation
Researchers first developed a diagnostic model for 9 methylation markers related to colorectal cancer, after identifying them by conducting a study of plasma from over samples of 801 patients with colorectal cancer and 1,021 controls.
This model was able to precisely differentiate patients from healthy individuals and outpaced the regular clinical blood test called CEA.
Furthermore, researches stated that only 1 methylation marker, in particular, helped spot cases of colorectal cancer and precancerous lesions found in a prospective study of about 1,493 individuals at risk.
The researchers concluded that studies with longer follow-up periods will be necessary to proceed with further assessment for the model’s reliability for clinicians and patients.
