Development of a classifier for [18F]fluorodeoxyglucose extravasation severity using semi-quantitative readings from...

Area of Science:

• Nuclear Medicine
• Radiopharmaceutical Imaging
• Medical Instrumentation

Background:

• Radiotracer extravasations occur in up to 23% of 18F-fluorodeoxyglucose (FDG) PET/CT scans, impacting image quality and quantification.
• Conventional assessment of extravasation severity using PET has performance limitations and practical drawbacks.

Purpose of the Study:

• Develop and assess a novel topical detector-based classifier for FDG extravasation severity.
• Calibrate the classifier using semi-quantitative PET measurements for improved accuracy.

Main Methods:

• Retrospective study using topical gamma-ray detectors during FDG PET/CT imaging.
• Quantified extravasation activity from PET images as ground truth for calibration.
• Developed a classifier using calibrated detector readings for four severity classes.

Main Results:

• Strong correlation (R²=0.75) between image-derived and topical detector activity estimates.
• Topical detector classifier achieved 95.8% accuracy for binary extravasation detection.
• Topical detector classifier significantly outperformed qualitative radiological analysis (MCC 0.87 vs. 0.30).

Conclusions:

• Topical detector classifier significantly improves extravasation detection compared to qualitative PET image analysis.
• Calibration with quantitative PET measurements enhances classifier performance.