Cross-Scanner Harmonization of AI/DL Accelerated Quantitative Bi-Parametric Prostate MRI

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Quantitative MRI

Background:

• Clinical adoption of AI/DL-aided quantitative bi-parametric MRI (q-bpMRI) requires cross-vendor validation and harmonization.
• Prostate cancer (PCa) diagnosis benefits from quantitative MRI metrics like T2 and apparent diffusion coefficient (ADC) mapping.

Purpose of the Study:

• To implement and validate an AI/DL-accelerated q-bpMRI protocol for prostate cancer.
• To assess image quality, quantitative accuracy, and harmonization potential across different MRI vendors.

Main Methods:

• An AI/DL-accelerated q-bpMRI protocol (5-echo T2, 4-b-value ADC) was tested on two vendors' 3T scanners in six PCa patients.
• Image quality was compared to standard-of-care (SOC) MRI using a Likert-like scale.
• Quantitative validation was performed using a multiparametric phantom with known T2 and diffusion kurtosis values.

Main Results:

• The 6-minute AI/DL q-bpMRI achieved diagnostic quality comparable to SOC.
• AI/DL showed improved signal-to-noise ratio (SNR) for ADC mapping compared to SOC.
• Acquisition parameter-dependent biases were identified for T2 and ADC, but were consistent across vendors after AI/DL reconstruction.
• Bias correction successfully harmonized lesion T2 and ADC values across different AI/DL acquisitions.

Conclusions:

• AI/DL-accelerated q-bpMRI is a viable tool for prostate cancer assessment, offering comparable diagnostic quality to SOC.
• The developed bias correction workflow enables harmonization of quantitative T2 and ADC mapping across multi-vendor platforms.
• This facilitates the clinical integration of AI/DL-enhanced quantitative MRI.