Comparing prostate diffusion weighted images reconstructed with a commercial deep-learning product to a deep learning...

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Multiparametric MRI (mpMRI) is crucial for prostate cancer assessment.
• Diffusion-weighted imaging (DWI) quality is vital for accurate diagnosis.
• Current deep learning (DL) models offer improvements but have limitations.

Purpose of the Study:

• To evaluate a novel deep learning based phase corrected (DLPC) reconstruction model.
• To compare DLPC's performance against a commercial DL product for prostate DWI at 1.5T.
• To assess enhancement in image quality, noise reduction, and signal-to-noise ratio (SNR).

Main Methods:

• Retrospective analysis of 30 patients' prostate mpMRI at 1.5T.
• Comparison of DWI datasets reconstructed with DLPC and a commercial DL model.
• Qualitative assessment by three radiologists using a 5-point Likert scale.
• Quantitative analysis of SNR and image noise using edge function and intermediate b-value images.

Main Results:

• Radiologists consistently rated DLPC images as having less noise and better quality (p < 0.05).
• DLPC reconstruction yielded significantly higher SNR (median 49.4 vs 27.5; p < 0.001).
• Edge analysis confirmed significantly reduced noise in DLPC images (p < 0.001), with no significant difference in ADC values.

Conclusions:

• DLPC image reconstruction significantly improves image quality for prostate DWI at 1.5T.
• The DLPC model demonstrates superior noise reduction and SNR enhancement over commercial DL products.
• This advancement holds promise for more accurate prostate cancer diagnosis using MRI.