CGBO-Net: Cruciform structure guided and boundary-optimized lymphoma segmentation network.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Accurate lymphoma segmentation is crucial for diagnosis and treatment.
• Current automatic methods struggle with precise tumor boundary and location identification.
• Semi-automatic methods often require manual input like bounding boxes or points.

Purpose of the Study:

• To develop a precise lymphoma segmentation network (CGBS-Net).
• To enhance tumor boundary and location accuracy.
• To improve segmentation results compared to existing semi-automatic methods.

Main Methods:

• Proposed a cruciform structure guided and boundary-optimized lymphoma segmentation network (CGBS-Net).
• Utilized a cruciform structure extracted from PET images as additional input.
• Employed a boundary gradient loss function for tumor boundary optimization.
• Developed an axial context-based cruciform structure extraction (CCE) method.
• Designed a network (CGBO-Net) using PET/CT and cruciform structures for segmentation.

Main Results:

• Achieved high performance metrics: Dice (90.7%), Precision (89.4%), Recall (92.5%), IOU (83.1%), and RVD (4.5%).
• Demonstrated superior performance compared to state-of-the-art semi-segmentation methods on lymphoma and head and neck datasets.
• Produced promising and accurate segmentation results.

Conclusions:

• CGBS-Net effectively improves lymphoma segmentation accuracy.
• The integration of cruciform structures and boundary optimization enhances tumor boundary and location detection.
• The proposed method offers a promising advancement in semi-automatic medical image segmentation.