A National Institutes of Health-funded research team has developed a machine learning model to assess CT scans.
The model, named Merlin, was able to accomplish “tasks as simple as identifying anatomical features to as complex as predicting disease onset years in advance.” The model was trained on a “unique set of patient CT scans linked to radiology reports and medical diagnosis codes collected from the Stanford University School of Medicine. The researchers note that it is the largest collection of abdominal CT data to date.”
Reading results of CT scans and performing any additional tests or clinical assessments necessary is a lengthy task growing more difficult in the wake of the growing shortage of physicians in the U.S. Merlin was meant to counteract this; it was initially trained on “more than 15,000 3D abdominal CT scans paired with…radiology reports and nearly one million diagnostic codes.”
Across 692 diagnostic codes, “Merlin successfully predicted which of two scans was more likely to be associated with a particular code over 81% of the time, outperforming several variants of two other models.” It was also able to “identify patients who were at higher risk of developing a particular [chronic] disease in the next five years 75% of the time, versus 68% for the other model [tested].”
