Medical IP, a company specializing in digital twin-based AI solutions, said on Wednesday that it obtained the FDA 510(k) certification for DeepCatch, an AI software for analyzing CT-based whole-body composition.

DeepCatch is a product that segments human body components from CT images and automatically produces an analysis report containing 3D images and quantitative information.

Accordingly, DeepCatch now stands as the only product that automatically analyzes various body components such as skin, bone, muscle, visceral fat, and subcutaneous fat through full-body CT among medical AI software programs with FDA certification.

The software is a differentiated technology that is not limited to one disease but can provide biomarker information for a variety of diseases related to body composition, such as various adult diseases, cardiovascular diseases, metabolic diseases such as diabetes, and sarcopenia in the elderly, and cancer treatment patients, in a single CT scan, explained a company official.

The software can also provide accurate body composition analysis results for anyone, regardless of race, gender, or age.

The performance testing of DeepCatch was validated on multiple CT equipment manufacturers, medical institutions, and  ethnicities in the U.S. The results demonstrated the accuracy of body composition analysis using AI, such as measuring the volume and area of each body component and abdominal circumference, and validated the product's universal performance and safety.

Unlike conventional bioelectrical impedance analysis (BIA), which are affected by the measurement environment and physical condition, CT can derive clinically valid three-dimensional body composition analysis results, which is expected to establish a new global standard for body composition analysis.

"As DeepCatch can be introduced to all medical institutions around the world where CT scans are performed, we will aim to create global standards for opportunistic screening and body composition analysis,” said Medical IP CEO Park Sang-joon.