CLEAR: Colon Segmentation with Liquid Error Avoidance and Refinement for VR Colonography

Pedro Silva Freire, Anderson Maciel, João Madeiras Pereira, Joaquim Jorge, Catarina Moreira

Abstract: Colon cancer remains a leading cause of cancer-related deaths, emphasizing the importance of early detection and accurate diagnostics. While colonoscopy is the gold standard, its invasiveness deters frequent, widespread use. Desktop colonography offers a less invasive alternative, and Immersive Colonography improves the efficiency of the analysis. However, colonography requires anatomical reconstruction from CTs, and liquid pockets in CT colonography often obscure the colon, necessitating dual-position scans and doubling radiation exposure. We introduce an automatic segmentation method that integrates liquid regions to eliminate the need for a second CT scan. By requiring only one CT image, this approach enhances patient safety and comfort while maintaining accurate segmentation. The proposed method uses advanced filtering and geometry-based techniques to accurately segment air and liquid pockets, ensuring continuity in colon reconstructions. This innovation promises safer, more efficient VR colonography and improved diagnostic capabilities.