MRI pyschophysics: an experimental framework relating image quality to diagnostic performance metrics

Area of Science:

• Radiology
• Medical Imaging
• Orthopedic Surgery

Background:

• Knee MRI is crucial for diagnosing internal derangements.
• Arthroscopy remains the gold standard for evaluating knee injuries.
• Determining the minimum image quality for accurate MRI diagnosis is essential.

Purpose of the Study:

• To establish the minimal image quality, specifically signal-to-noise ratio (SNR), required for accurate knee MRI diagnosis.
• To compare MRI diagnostic performance with arthroscopic findings.

Main Methods:

• Synthetic noise was added to knee MRI scans from five patients.
• Images were analyzed across five standardized signal-to-noise ratios (SNRs).
• Musculoskeletal radiologists assessed cartilage, ligaments, and menisci using a 5-point scale, with results compared to arthroscopy.

Main Results:

• Optimal diagnostic performance for meniscal tears was achieved at SNR = 5 (CNR = 3.2).
• Optimal diagnostic performance for cartilage abnormalities was achieved at SNR = 10 (CNR = 4.2).
• Reader confidence peaked at SNR = 5 for menisci and SNR = 10 for ligaments and cartilage.

Conclusions:

• For 3D knee MRI, images with SNR less than 10 or CNR less than 10 are considered nondiagnostic.
• Maintaining adequate image quality is critical for reliable knee MRI interpretation.