Visualization of the Trochlear Nerve Using Deep Learning-enhanced 3D T2-weighted MR Imaging at 3T

Area of Science:

• Neuroradiology
• Medical Imaging
• Deep Learning Applications

Background:

• Cranial nerve imaging at 3 Tesla (3T) MRI often uses 3D fast imaging employing steady-state acquisition (3D-FIESTA).
• 3D-FIESTA has limitations in spatial resolution and tissue contrast for cranial nerves versus vascular structures.
• The trochlear nerve, the smallest cranial nerve, presents unique imaging challenges due to its long intracranial course.

Purpose of the Study:

• To evaluate 3D T2-weighted imaging (T2-CUBE) with deep learning-based reconstruction (DLR) for cranial nerve visualization.
• To compare the efficacy of T2-CUBE with DLR against T2-CUBE without DLR and 3D-FIESTA.
• To focus on the visualization of the trochlear nerve, including its various anatomical segments.

Main Methods:

• Ten healthy male volunteers underwent 3T MRI scans.
• Scans included T2-CUBE with and without DLR, and 3D-FIESTA.
• Two neuroradiologists assessed trochlear nerve visualization across four segments using a 3-point scale.
• Signal-to-noise ratio (SNR) of the pons (SNR_PONS) and cerebrospinal fluid (SNR_CSF) were calculated.

Main Results:

• T2-CUBE with DLR achieved 100% visualization of all trochlear nerve segments, significantly outperforming other methods (P < 0.025).
• T2-CUBE without DLR showed 67.5%-100% visualization, while 3D-FIESTA showed 32.5%-80% visualization, with no visualization in cavernous segments.
• DLR increased SNR_PONS by 1.8-2.5 times (14.1 vs 5.7) and SNR_CSF by 1.8-2.5 times (31.8 vs 17.5) (P < 0.001).
• T2-CUBE with DLR showed significantly higher SNR_PONS than 3D-FIESTA (14.1 vs 6.4, P < 0.001).

Conclusions:

• 3D T2-CUBE with DLR at 3T significantly enhances trochlear nerve visualization.
• This technique shows potential for improved imaging of other cranial nerves and neurovascular compression.
• T2-CUBE with DLR may become the new standard for cisternal imaging.