Abdominal Muscle Groups Analysis Software – AI-powered Tool for Cancer Sarcopenia

Booth No: N516

Characteristic

This technique utilizes quantitative analysis on abdominal CT axial images at the L3 vertebral level to infer body composition metrics. A deep learning model is utilized to automatically segment five abdominal muscle groups (rectus abdominis, later wall muscles, psoas muscle, quadratus lumborum and posterior paraspinous muscles), calculating the areas and densities of muscle, visceral fat and subcutaneous fat, metrics of myosteatosis (normal-attenuation muscle area, low-attenuation muscle area, and inter-/intra-muscular adipose tissue), waist circumference, and radiomics features of bone density. The AI models and analysis modules are integrated into a comprehensive software platform. Users can input either complete abdominal CT image sequences or a single pre-selected L3 slice image. If an image sequence is provided, the platform will automatically select the L3 slice and display in the platform's image viewer for user confirmation and manual correction. The platform then performs automatic muscle and vertebral body segmentation, represented by polygons which can be modified by users to better align with the contours of the target objects by moving, adding, or deleting polygon nodes. Subsequently, quantitative analysis generates key body composition parameters related to muscle, fat, myosteatosis, waist circumference, and bone density. The technique provides valuable insights into sarcopenia, metabolic syndrome, and osteoporosis.
The software platform facilitates AI calculations and human-computer collaboration, allowing users to make corrections for image selection or segmentation, thereby generating more accurate quantitative analysis results. The platform also allows users to visualize and export quantitative analysis results in real-time for comprehensive report editing and further analysis. Results are exportable in PDF, image, or CSV formats.