Comparison between PUN and Tofts models in the quantification of dynamic contrast-enhanced MR imaging

Area of Science:

• Medical Imaging
• Oncology
• Biophysics

Background:

• Dynamic contrast-enhanced MRI (DCE-MRI) visualizes tissue vascularization and tumor aggressiveness.
• Pharmacokinetic models quantify contrast uptake but rely on physiological assumptions.
• Empirical functions offer an alternative for fitting DCE-MRI data, especially in automated diagnostic systems.

Purpose of the Study:

• To compare the Tofts pharmacokinetic model with the Phenomenological Universalities (PUN) class EU1 for analyzing prostate DCE-MRI data.
• To evaluate model implementation complexity, fitting accuracy, diagnostic performance, and computational cost.
• To assess the utility of PUN-based models as a diagnostic tool in computer-aided detection systems.

Main Methods:

• DCE-MRI data from prostate studies were analyzed using both the Tofts pharmacokinetic model and the EU1 model derived from PUN.
• Key performance metrics including goodness-of-fit (R²), classification accuracy (AUC), implementation complexity, and processing time were compared.
• The study focused on discriminating malignant from benign tissues.

Main Results:

• The EU1 model achieved a mean R² of 0.96, slightly higher than the Tofts model's 0.90.
• Classification performance for discriminating malignant from benign tissues was excellent for both models, with AUCs of 0.92 for EU1 and 0.91 for Tofts.
• The EU1 model demonstrated simpler implementation and reduced computational time (6 minutes vs. 8 minutes for Tofts).

Conclusions:

• The EU1 model, based on Phenomenological Universalities, provides a robust and efficient alternative to traditional pharmacokinetic models for DCE-MRI analysis in prostate cancer.
• PUN-based models offer a simplified approach with comparable diagnostic accuracy and reduced computational burden.
• This suggests a valuable role for PUN in enhancing computer-aided detection systems for oncological imaging.