# Clinical Image Quality and Reader Variability in 3D Synthetic Brain MRI Compared with Conventional MRI

## TL;DR

This study compares 3D synthetic MRI with conventional MRI for brain imaging, finding that while synthetic MRI is robust to motion and efficient, it has limitations in detail and reader interpretation.

## Contribution

The study introduces a detailed evaluation of 3D synthetic MRI's clinical usability and variability in reader interpretation compared to conventional MRI.

## Key findings

- 3D synthetic MRI provides motion-robust images but has lower cortical contrast and SNR compared to conventional MRI.
- Experienced readers show higher confidence in conventional MRI for subtle abnormalities.
- Synthetic MRI has broader vascular signal range but reduced sensitivity to vascular abnormalities.

## Abstract

This study evaluates whether modern three-dimensional synthetic MRI (3D SI) provides reliable diagnostic information compared with conventional MRI (cMRI) in routine neuroradiology. Using matched datasets and multiple blinded readers, it assesses image quality, diagnostic confidence, and detection of key brain findings. Results show that 3D SI delivers usable, motion-robust images, though some anatomical features appear differently due to its technical principles, requiring readers to learn their distinct “synthetic” appearance. While cMRI offers higher confidence for subtle abnormalities, this advantage is tied to reader experience, which affects consistency across both techniques. The study outlines strengths, limitations, and practical guidance for integrating synthetic imaging into clinical workflows.

Background/Objectives: This study evaluated the clinical image quality of three-dimensional synthetic MRI (3D SI) compared with conventional MRI (cMRI), focusing on tissue contrast, anatomical detail, and motion sensitivity. Methods: Patients with nonspecific neurological symptoms were included. Both cMRI and 3D SI were acquired on single-vendor 1.5 T and 3 T scanners with slice thicknesses of 1.0–1.7 mm. Two experienced neuroradiologists and one fellow independently evaluated matched scans using a 0–100 scale. Assessed parameters included signal-to-noise ratio (SNR), gray–white matter contrast, artifacts, motion robustness, and confidence in detecting perivascular spaces, white matter lesions, and subtle pathology. Interrater agreement was measured using Krippendorff’s alpha and ICC2. Multiple linear regression analyzed associations between image quality ratings and imaging method. Results: Images of 31 patients were analyzed. Three-dimensional SI demonstrated sufficient-to-good overall image quality and high robustness to motion. Cortical-surface-to-cerebrospinal-fluid contrast on FLAIR was rated lower for 3D SI than for cMRI. False-positive lesion detection occurred more frequently on 3D SI FLAIR, particularly among experienced readers. cMRI achieved significantly higher T1-weighted SNR than 3D SI (8.76 points, p < 0.001). Experienced readers consistently rated SNR and tissue contrast higher than the fellow. Vascular signal range was broader on 3D SI, reducing sensitivity to vascular abnormalities. Conclusions: Three-dimensional synthetic MRI provides clinically usable image quality and fulfills its primary diagnostic purpose, offering advantages in acquisition efficiency and robustness to motion. Nevertheless, limitations in cortical contrast, vascular signal characterization, and reader-dependent interpretive variability constrain its reliability for subtle or detail-critical findings.