A semiautomatic segmentation method framework for pelvic bone tumors based on CT-MR multimodal images

Area of Science:

• Medical Imaging
• Oncology
• Computer-Aided Surgery

Background:

• Pelvic bone tumor resection is challenging due to complex anatomy and poorly defined margins.
• Inaccurate tumor segmentation leads to prolonged surgery and potential failure.
• A precise method for pelvic bone tumor segmentation is essential for surgical planning.

Purpose of the Study:

• To develop and evaluate a semiautomatic segmentation method for pelvic bone tumors.
• To improve the accuracy of tumor margin identification in pelvic bone tumor surgery.
• To provide a tool for assisting pelvic bone tumor preservation surgery.

Main Methods:

• A semiautomatic segmentation approach utilizing CT-MR multimodal images.
• Integration of medical prior knowledge with advanced image segmentation algorithms.
• 3D visualization of segmentation results for surgical planning.

Main Results:

• The method achieved high performance metrics: accuracy (0.9358), recall (0.9278), IOU (0.8697), Dice (0.9280), and AUC (0.9632).
• Segmentation accuracy was consistent across various tumor locations and sizes.
• The average error of the 3D model was within surgical tolerance.

Conclusions:

• The proposed algorithm accurately segments pelvic bone tumors from multimodal images.
• This method offers a reliable solution for identifying tumor resection margins.
• It has the potential to enhance pelvic bone tumor preservation surgery outcomes.