Repeatability of Quantitative Imaging Features in Prostate Magnetic Resonance Imaging

Area of Science:

• Radiology
• Oncology
• Medical Imaging

Background:

• Multiparametric MRI (mpMRI) is a non-invasive tool for prostate cancer diagnosis and monitoring.
• Quantitative imaging metrics can help differentiate clinically significant prostate cancers.
• Reproducibility of these metrics is crucial for clinical application.

Purpose of the Study:

• To evaluate the reproducibility of quantitative imaging features from repeated mpMRI scans in the same patients.
• To assess the impact of different analysis approaches (radiologist-drawn regions vs. habitat regions) on feature reproducibility.
• To identify reproducible imaging features for improved prostate cancer risk stratification.

Main Methods:

• Retrospective analysis of deidentified mpMRI scans from patients undergoing two scans within two weeks.
• Extraction of 307 quantitative features from T2-weighted (T2w) and apparent diffusion coefficient (ADC) sequences.
• Calculation of concordance correlation coefficients (CCCs) to assess feature reproducibility.
• Removal of redundant features using coefficient of determination (R²).

Main Results:

• Habitat region characterization improved repeatability in ADC maps.
• 18 T2w and 15 ADC features in habitat regions were reproducible (CCC ≥0.65) and non-redundant (R² ≥ 0.99).
• Z-transformation of images prior to feature extraction did not negatively impact reproducibility.

Conclusions:

• Quantitative imaging features in prostate mpMRI are reproducible across sequential acquisitions.
• A habitat approach enhances feature repeatability, particularly in ADC maps.
• Validated reproducible features can aid in disease risk stratification and potentially serve as a surrogate for biopsies.