Deep learning reconstruction for temporomandibular joint MRI: diagnostic interchangeability, image quality, and scan...

Area of Science:

• Medical Imaging
• Artificial Intelligence in Radiology
• Temporomandibular Joint Disorders

Background:

• Conventional MRI for temporomandibular joint (TMJ) disorders can be limited by long scan times, potentially increasing patient discomfort and motion-related artifacts.
• Deep learning (DL) reconstruction offers a potential solution to improve MRI efficiency and quality.

Purpose of the Study:

• To compare the diagnostic interchangeability, image quality, and scan time of DL-reconstructed MRI versus conventional MRI for the TMJ.
• To assess the impact of DL reconstruction on the evaluation of TMJ disc shape, position, and joint effusion.

Main Methods:

• Patients with suspected TMJ disorders underwent both conventional and DL-reconstructed sagittal proton density-weighted (PDW) and T2-weighted fat-suppressed (T2W FS) MRI.
• Three oral radiologists independently assessed diagnostic parameters, and image quality was evaluated both qualitatively (scores) and quantitatively (SNR, CNR).
• Diagnostic interchangeability was defined by interobserver agreement within a 95% confidence interval of ±5%.

Main Results:

• DL-reconstructed MRI demonstrated diagnostic equivalence to conventional MRI for disc shape, position, and effusion (equivalence indices <3%).
• DL reconstruction significantly reduced noise and improved signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) (p<0.05), except for disc SNR in PDW.
• Scan time was reduced by 49.2% with DL reconstruction.

Conclusions:

• DL-reconstructed TMJ MRI is diagnostically interchangeable with conventional MRI.
• DL reconstruction offers improved image quality and a significantly shorter scan time, making TMJ MRI more efficient and tolerable.
• This technology facilitates broader clinical adoption by overcoming limitations of long scan times and motion artifacts.