Medical IP has published a study that proved DeepCatch, its AI-backed whole-body CT analysis software, can predict the prevalence and incidence of type 2 diabetes while simultaneously identifying various cardiometabolic complications using CT scans.

DeepCatch, which has received FDA 510(k) clearance, uses deep learning algorithms to segment and quantify body composition data from CT images automatically.

The study, "Automated comprehensive CT assessment of the risk of diabetes and associated cardiometabolic conditions," was conducted in collaboration with Kangbuk Samsung Hospital and Seoul National University Hospital.

It analyzed CT scans from 32,166 adults aged 27 to 83 who underwent PET-CT examinations between January 2012 and December 2015.

The study found that the visceral fat index was the strongest predictor of diabetes prevalence and incidence. When combined with muscle area index, liver fat ratio, and aortic calcification measurements, the predictive performance improved further.

The AI tool also showed high accuracy in identifying associated conditions, such as fatty liver, coronary artery calcification, sarcopenia, and osteoporosis. For example, the area under the receiver operating characteristic curve (AUROC) for identifying metabolic syndrome using the visceral fat index was 0.81 for men and 0.90 for women, indicating excellent performance.

"This study shows that CT imaging can be used for preventive screening beyond traditional disease diagnosis,” said Professor Chang Yoo-soo at Kangbuk Samsung Hospital. “We expect that CT images will be used to screen for diabetes and other cardiometabolic diseases simultaneously, helping to predict diabetes onset in currently healthy individuals and aid in early detection and prevention of complications."

Medical IP CEO Park Sang-joon said, “Now, individuals can obtain biomarker information for various metabolic and body composition-related diseases from a single CT scan during their health checkup.”

Park added that it will increase the acceptability of CT analysis for medical professionals by minimizing manual work and enabling faster, more accurate analysis.

The results of the research were published in Radiology.