A statistically optimized regional thresholding method (SORT) for bone lesion detection in ¹⁸F-NaF PET/CT imaging

Area of Science:

• Nuclear Medicine
• Oncology
• Medical Imaging Analysis

Background:

• Identifying bone lesions on 18F-NaF PET scans is challenging due to subjectivity and limitations of current automated methods.
• Accurate characterization of disease burden is crucial for effective treatment planning in metastatic prostate cancer.

Purpose of the Study:

• To develop and validate a novel Statistically Optimized Regional Thresholding (SORT) method for improved bone lesion detection on 18F-NaF PET/CT scans.
• To compare the diagnostic performance of SORT thresholds against commonly used global thresholds (SUV10 and SUV15).

Main Methods:

• Developed SORT, a statistically optimized regional thresholding technique, analyzing 18F-NaF PET/CT scans from 37 prostate cancer patients.
• Divided PET images into 19 skeletal regions, with experienced nuclear medicine physicians identifying and marking lesions.
• Determined region-specific SUV thresholds using receiver operating characteristic (ROC) analysis and compared SORT performance to global SUV thresholds.

Main Results:

• SORT identified region-specific thresholds (SUV range: 3-13 g mL-1), achieving higher sensitivity (95.8%) than global thresholds (SUV10: 82.8%, SUV15: 58.4%).
• SORT maintained high specificity (97.1%), comparable to global thresholds (SUV10: 97.3%, SUV15: 100.0%).
• Image reconstruction settings minimally impacted SORT threshold optimization, with average threshold changes of 10%.

Conclusions:

• Region-specific SUV thresholding using the SORT method significantly enhances bone lesion detection sensitivity in 18F-NaF PET imaging for metastatic prostate cancer.
• The SORT method demonstrates superior performance over current global thresholding techniques, offering a more accurate assessment of disease burden.