Dual-tracer PET Radiomics for Prostate Cancer Stratification

Dual-tracer PET/CT Interpretable Radiomics for Pathological Classification and Prognostic Stratification of Prostate Cancer

This single-center retrospective study aims to develop an interpretable radiomics model based on dual-tracer PET/CT to preoperatively predict the postoperative pathological Gleason grade group in treatment-naïve prostate cancer patients. A machine learning-based three-class prediction model will be constructed and interpreted using SHAP. Its performance will be compared with systematic biopsy results, assessing grading accuracy and prognostic value for biochemical recurrence-free survival.

Inclusion Criteria:

• Treatment-naïve prostate cancer with subsequent radical prostatectomy (RP)
• Availability of presurgical dual-tracer PET/CT images, biopsy, and RP pathology data

Exclusion Criteria:

• Unavailability of imaging or pathological data
• PET/CT performed after any prostate cancer-related treatment
• Interval between PET and biopsy > 3 months
• Interval between biopsy and RP > 1 month