Comparative evaluation of four reconstruction techniques for prostate T2-weighted MRI: Sensitivity encoding,...

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Prostate cancer diagnosis relies heavily on MRI, particularly T2-weighted imaging (T2WI).
• Image quality and lesion conspicuity are critical for accurate detection and staging.
• Advancements in reconstruction techniques aim to improve diagnostic performance.

Purpose of the Study:

• To compare the image quality and lesion conspicuity of prostate T2WI using four reconstruction methods.
• Evaluate conventional Sensitivity Encoding (SENSE), compressed sensing (CS), deep learning reconstruction (DL), and deep learning super-resolution reconstruction (SR).

Main Methods:

• Retrospective analysis of 49 patients undergoing prostate MRI (mpMRI or bpMRI).
• T2WI acquired with SENSE and CS protocols; CS data reconstructed using CS, DL, and SR.
• Qualitative (Likert scale) and quantitative (SNR, CNR, sharpness) assessments by radiologists.
• Preliminary evaluation of PI-RADS T2WI scores and lesion conspicuity in 18 patients.

Main Results:

• SR consistently outperformed SENSE, CS, and DL in qualitative and quantitative image quality assessments (p < 0.0001).
• SR demonstrated a trend towards improved lesion conspicuity.
• PI-RADS T2WI scores were comparable across all reconstruction methods.

Conclusions:

• Deep learning super-resolution reconstruction (SR) offers superior image quality for prostate T2WI.
• SR shows potential for enhancing lesion visualization in prostate MRI.
• Further investigation in larger cohorts is warranted to confirm SR's benefits.