Multi-Site Concordance of Diffusion-Weighted Imaging Quantification for Assessing Prostate Cancer Aggressiveness

Area of Science:

• Radiology
• Medical Imaging
• Oncology

Background:

• Diffusion-weighted imaging (DWI) is crucial for prostate cancer detection.
• A key challenge is distinguishing aggressive from indolent prostate cancer.
• Standardized DWI analysis is needed for reliable cancer characterization.

Purpose of the Study:

• To compare 14 different parametric fitting methods for DWI data.
• To evaluate how various fitting algorithms affect prostate cancer differentiation.
• To test if cancer detection varies with DWI analysis techniques.

Main Methods:

• Prospective study of 33 patients undergoing prostatectomy.
• Utilized a 3T DWI sequence with optimized field-of-view.
• Compared 14 site-specific DWI parametric fitting implementations using a digital reference object and patient data.
• Analyzed mono-exponential (MEADC), kurtosis (K, DK), and bi-exponential (BID, BID*, F) models.
• Assessed differentiation using receiver operating characteristic area under the curve (ROC AUC).

Main Results:

• Minimal site-to-site variability observed with a digital reference object.
• Kurtosis (K, DK) and MEADC models demonstrated superior prostate cancer detection (AUC 0.72-0.81).
• Bi-exponential models (BID, BID*, F) showed lower detection capability and higher variability (AUC 0.51-0.80).
• Post-processing choices, like cluster size, influenced results (e.g., MEADC AUC from 0.75 to 0.87).

Conclusions:

• Conventional diffusion models (K, DK, MEADC) consistently differentiate prostate cancer from benign tissue.
• DWI post-processing significantly affects diagnostic sensitivity and specificity.
• Standardized DWI analysis protocols are essential for accurate prostate cancer assessment.