Accelerated Spine MRI with Deep Learning Based Image Reconstruction: A Prospective Comparison with Standard MRI

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Standard-of-care (SOC) MRI for spinal imaging can be time-consuming.
• Deep learning (DL) algorithms offer potential for accelerated image acquisition and improved image quality.

Purpose of the Study:

• To compare the performance of DL-reconstructed MRI with SOC MRI.
• To evaluate DL MRI in terms of image acquisition time, image quality, and diagnostic interchangeability for spinal abnormalities.

Main Methods:

• Prospective study of 200 participants with spinal discomfort undergoing both SOC and DL-accelerated MRI.
• Quantitative analysis of signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR).
• Radiologist assessment of subjective image quality and lesion detection, with statistical analysis for interchangeability and agreement.

Main Results:

• DL MRI reduced scan time by approximately 40% with significantly higher SNR and CNR (P < 0.001).
• DL MRI demonstrated interchangeability with SOC MRI for detecting spinal abnormalities (equivalence index < 4%).
• No significant difference in subjective image quality or lesion detection rates between DL and SOC MRI.

Conclusions:

• DL-reconstructed MRI provides high-quality diagnostic images while reducing patient examination time.
• DL MRI is a viable and interchangeable alternative to SOC MRI for evaluating spinal abnormalities.