Assembling High-quality Lymph Node Clinical Target Volumes for Cervical Cancer Radiotherapy using a Deep...

Area of Science:

• Medical Imaging
• Radiotherapy Oncology
• Artificial Intelligence in Medicine

Background:

• Accurate delineation of lymph node clinical target volumes (CTV) is crucial for effective cervical cancer radiotherapy.
• Current manual segmentation methods can be time-consuming and prone to inter-observer variability.

Purpose of the Study:

• To develop and evaluate a deep learning-based approach for automated segmentation of lymph node sub-regions.
• To assemble high-quality CTVs for cervical cancer radiotherapy using CT images.

Main Methods:

• A 3D encoder-decoder network was trained on CT images from 152 cervical cancer patients.
• Manual delineation of seven lymph node regions (sub-CTV) was performed.
• The model was optimized for individual structures and validated on an additional 64 cases.
• Quantitative evaluation used Dice Similarity Coefficient (DSC) and Hausdorff Distance (HD).

Main Results:

• The deep learning model achieved high accuracy in segmenting various lymph node regions, with mean DSC values ranging from 0.784 to 0.874.
• Hausdorff distances were generally low, indicating precise contouring (e.g., 4.7mm for common and internal iliac nodes).
• The model demonstrated feasibility across six different clinical conditions, with mean DSC of 0.877 and HD of 4.4mm in one scenario.

Conclusions:

• A deep learning-based approach effectively automates lymph node sub-region segmentation for CTV assembly in cervical cancer radiotherapy.
• This method enhances the efficiency of cervical cancer detection and treatment planning.
• The developed model provides a reliable tool for generating high-quality CTVs, supporting clinical decision-making.