Head and neck target delineation using a novel PET automatic segmentation algorithm

Area of Science:

• Radiotherapy
• Medical Imaging
• Artificial Intelligence

Background:

• Head and Neck (H&N) radiotherapy treatment (RT) planning relies on accurate delineation of Gross Tumour Volumes (GTVs).
• Manual segmentation of GTVs from CT/MRI can be subjective and time-consuming.
• Advanced imaging techniques like ¹⁸F-FDG-PET/CT offer functional information crucial for precise tumor outlining.

Purpose of the Study:

• To assess the feasibility and impact of a novel automated PET auto-segmentation method (ATLAAS) in H&N RT planning.
• To evaluate the reliability and clinical utility of ATLAAS-generated contours compared to manual delineations.

Main Methods:

• The Automatic decision Tree-based Learning Algorithm for Advanced Segmentation (ATLAAS) was applied to ¹⁸F-FDG-PET/CT scans of 20 H&N patients.
• ATLAAS contours (GTVp_ATLAAS) were compared with manually delineated CT/MRI contours (GTVp_CT/MRI) and final RT planning GTVs (GTVp_final).
• Qualitative and quantitative analyses, including a conformity metric, were used for comparison.

Main Results:

• ATLAAS contours demonstrated reliability and usefulness, with an average conformity index of 0.70 against final GTVs.
• GTVp_ATLAAS was smaller than GTVp_CT/MRI in 70% of cases.
• ATLAAS insights led to adjustments in GTVp_CT/MRI in 17 out of 20 patients, contributing up to 33% to the final GTV volume.

Conclusions:

• ATLAAS provides operator-independent PET segmentation, augmenting clinical outlining based on CT and MRI.
• This automated method shows potential utility for improving H&N radiotherapy planning in future clinical studies.