Texture Analysis of F-18 Fluciclovine PET/CT to Predict Biochemically Recurrent Prostate Cancer: Initial Results

Area of Science:

• Radiology and Oncology
• Medical Imaging Analysis
• Computational Pathology

Background:

• Accurate prediction of prostate cancer recurrence is crucial for timely treatment and improved patient outcomes.
• Inter- and intrareader variability in interpreting F-18 fluciclovine PET/CT scans complicates distinguishing tumor tissue from post-radiation necrosis.
• Standardized interpretation methods are needed to reliably identify biochemical recurrence of prostate cancer.

Purpose of the Study:

• To develop a computational methodology using Haralick texture analysis to standardize and improve F-18 fluciclovine PET/CT interpretation.
• To enhance the prediction of biochemical recurrence in prostate cancer by integrating imaging features with clinical data.

Main Methods:

• Haralick texture analysis was applied to F-18 fluciclovine PET/CT imaging data.
• Variable selection using least absolute shrinkage and selection operator (LASSO) logistic regression and bootstrapping identified key textural features.
• Logistic ridge regression model incorporated selected textural features, age at prostatectomy, PSA level, and time from PSA measurement to PET/CT imaging.

Main Results:

• The proposed model achieved an overfitting-corrected area under the curve (AUC) of 0.94 and a Brier score of 0.12.
• Compared to models using only textural features or clinical information, the combined model showed a 2% and 32% increase in AUC, respectively.
• The combined model demonstrated an 8% and 48% reduction in Brier score compared to the textural imaging and clinical information models, respectively.

Conclusions:

• Combining Haralick textural features from PET/CT imaging with clinical information significantly enhances the prediction of biochemical recurrence in prostate cancer.
• This computational approach offers a promising adjunct tool for improving the standardization and accuracy of F-18 fluciclovine PET/CT interpretation.
• The findings support the potential of advanced imaging analysis for more precise patient management in prostate cancer.