Accelerated diffusion-weighted imaging of the prostate using deep learning image reconstruction: A retrospective...

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence

Background:

• Multiparametric MRI of the prostate is crucial for diagnosing prostate cancer, the most common male solid cancer.
• Current MRI systems face capacity limitations, impacting acquisition times for essential sequences like diffusion-weighted imaging (DWI).
• Deep learning image reconstruction (DLIR) offers a potential solution to accelerate MRI acquisition.

Purpose of the Study:

• To evaluate the efficacy of DLIR in accelerating prostate DWI acquisition.
• To determine if DLIR can maintain diagnostic image quality despite reduced acquisition times.
• To assess the impact of DLIR on image noise and signal-to-noise ratios.

Main Methods:

• Retrospective analysis of prostate DWI data from 35 patients.
• Reconstruction of DWI sequences using standard and DLIR methods.
• Simulated 39% reduction in acquisition time by decreasing the number of signal averages.
• Assessment of image quality by three radiologists and objective metrics.

Main Results:

• DLIR resulted in perceived lower image noise in b=0 s/mm² images and ADC maps, with good inter-reader agreement.
• Overall signal-to-noise ratios remained comparable between standard and DLIR methods.
• A discrete reduction in signal-to-noise ratio was observed in the transitional zone with DLIR.

Conclusions:

• Deep learning image reconstruction facilitates a 39% reduction in prostate DWI acquisition time.
• Diagnostic image quality is preserved when using DLIR for prostate DWI.
• DLIR is a feasible technique for accelerating prostate MRI while maintaining diagnostic performance.