The research team led by Associate Professor Tsai Yi-Shan at College of Medicine, National Cheng Kung University (NCKU) has developed the "iLement Abdominal CT Body Composition Quantification System." This system uses deep learning technology to automatically analyze abdominal CT images, providing quick and precise assessments of cancer patients' muscle mass, muscle density, and fat content. This assists physicians in the early diagnosis of sarcopenia and enables timely intervention, with an accuracy rate of over 90%.

AI Rapidly Diagnoses Sarcopenia in Cancer Patients! iLement System Aids Physicians in Early Treatment

Cancer has been the leading cause of death in Taiwan for several consecutive years, with 121,254 new cancer cases reported in 2019. 47% of cancer patients can have their conditions monitored through abdominal CT scans. Utilizing existing abdominal CT images of patients is the most suitable method for diagnosing sarcopenia in cancer patients. Traditional semi-automated commercial software takes an average of 20 minutes to process each image. The team trained an AI model using a method that traces the fascia for quantifying abdominal muscle and fat, delivering results within seconds.

The iLement Abdominal CT Body Composition Quantification System boasts three major technological breakthroughs:

1. AI-Based Selection of Correct L3 Images in Abdominal CT Series: Traditional methods are prone to projection interference from structures other than the spine, and individual differences can result in incomplete reconstruction information. The iLement system employs dynamic MIP (Maximum Intensity Projection) and bi-directional MIP projection techniques to more accurately select L3 images.
2. Muscle Segmentation in CT Images: Traditional manual segmentation of muscle groups is time-consuming and labor-intensive, with inconsistent quality. The iLement system uses a two-stage model training method and fascia-tracing techniques to quickly and precisely segment individual muscle groups.
3. Diverse Quantification Results Presentation: The iLement system links specific muscle groups to potential disease factors, providing information on intermuscular and intramuscular fat infiltration. It quantifies muscle fat degeneration, extrapolates muscle function assessments, detects early changes in sarcopenia, and offers more comprehensive body composition quantification results.

Associate Professor Tsai Yi-Shan stated that iLement uses AI technology to automatically execute muscle segmentation, significantly reducing the time and effort required for manual annotation. This system serves as a foundation for large-scale research and development applications. Clinical trials conducted at NCKU Hospital show that the system has an accuracy rate of over 90%, effectively aiding physicians in diagnosing sarcopenia in cancer patients. The innovative annotation method proposed by the team provides a consistent annotation consensus and is currently undergoing multi-national patent applications. In the future, the team will continue to promote the iLement system to other hospitals and develop more body composition quantification analysis applications, aiming to help more patients receive better treatment.

Resource (mandarin): 成大研發身體組成定量系統 速判癌友肌少症 助提升存活率！