Assessing inter-observer variability in prostate and GTV segmentation on mpMRI: A comparison between radiation...

Area of Science:

• Radiotherapy
• Medical Imaging
• Oncology

Background:

• Accurate prostate and intraprostatic gross tumor volume (GTV) delineation on multiparametric MRI (mpMRI) is crucial for radiation therapy planning.
• Inter-observer variability in mpMRI interpretation challenges radiation oncologists (ROs), especially those with limited MRI training.
• Lack of radiologist input during treatment planning due to absent diagnostic mpMRI compromises GTV delineation.

Purpose of the Study:

• To assess the impact of radiology reports on ROs' prostate GTV contours.
• To compare an automated segmentation tool (AIRC) with ROs' contours for potential use when radiologist input is unavailable.
• To evaluate methods for enhancing focal prostate GTV delineation on mpMRI for radiotherapy.

Main Methods:

• ROs performed prostate and GTV segmentation twice: initially with biopsy data only, then with radiology reports.
• An automated tool (AIRC) segmented the same mpMRI images.
• Segmentation accuracy was compared to consensus contours using metrics like Dice Similarity Coefficient (DSC) and Hausdorff Distance (HD).

Main Results:

• Prostate gland segmentation showed high agreement and minimal variability for both ROs and AIRC (mean DSC ~0.90).
• GTV segmentation by ROs without radiology reports showed significant variability and errors.
• Access to radiology reports significantly improved ROs' GTV segmentation (p<0.01) and positive predictive value (p=0.03).
• AIRC performance was comparable to ROs' initial segmentation without reports.

Conclusions:

• Significant inter-observer variability in GTV segmentation highlights challenges in mpMRI interpretation.
• Providing radiology reports and standardized guidelines can improve segmentation accuracy and consistency.
• Automated segmentation tools can assist ROs in initial target definition for radiotherapy planning.