Image quality of whole-body diffusion MR images comparing deep-learning accelerated and conventional sequences

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Whole-body MRI (WB-MRI) is valuable for staging bone marrow cancer.
• Conventional diffusion-weighted imaging (DWI) sequences are time-consuming.
• Deep learning reconstruction offers potential for accelerated imaging.

Purpose of the Study:

• To compare image quality between deep learning accelerated (DRB) and conventional WB-DWI sequences.
• To evaluate the impact of DRB on image metrics and diagnostic confidence.
• To assess the feasibility of DRB for bone marrow cancer evaluation.

Main Methods:

• Fifty patients with bone marrow cancer underwent WB-MRI with both DRB-accelerated and conventional DWI sequences.
• Two expert radiologists qualitatively assessed image noise, artifacts, fat suppression, and lesion conspicuity using Likert scales.
• Quantitative analysis included signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and apparent diffusion coefficient (ADC) values.

Main Results:

• Radiologists preferred DRB sequences in nearly 80% of cases, especially for patients with higher BMI.
• DRB sequences demonstrated significantly higher SNR in normal tissues and cancer lesions.
• Acquisition time was reduced by over 50% with DRB (6:42 min vs. 14 min).

Conclusions:

• Deep learning reconstruction (DRB) significantly enhances WB-DWI quality, improving image metrics and reducing scan time.
• DRB facilitates improved lesion conspicuity and diagnostic preference, potentially increasing WB-MRI adoption for bone marrow cancer.
• While ADC values in normal tissues differ, cancer lesion diffusivity remains comparable to conventional sequences.