Intelligent CT Image Analysis Algorithm for 3D Mo del Construction

Abstract

This paper presents an algorithm and associated software implementing a comprehensive computational pipeline based on the YOLOv8n-seg model (Ultralytics) for automated analysis of human lumbar spine CT images, including localization of anatomical regions, segmentation and quantitative measurements of vertebrae, and 3D visualization. Geometric parameters (total lumbar spine height and segmental angles) are calculated directly from segmentation masks and converted into physical units using the image resolution. This generates an STL file that can be exported to third-party software (e.g., 3D Slicer) for 3D surface reconstruction. A critical aspect of the developed approach is its emphasis on quantitative accuracy and reproducibility of results, rather than solely on visualization: measurements obtained from masks translate the model output into clinically interpretable metrics. This opens up opportunities for subsequent applications such as screening, patient stratification, and longitudinal studies.