Low-field strength MRI (0.55T) for stereotactic and functional neurosurgery using deep learning-based reconstruction algorithm: Preliminary experiences

BACKGROUND: There has been a trend toward exclusive use of high MRI field strength (1.5 T or above) for stereotactic neurosurgery imaging. However, low field strength (0.55 T) MRI may have advantages related to availability, cost, image distortion, and artifact. Low field strength MRI recently has been shown to be effective for diagnosis of brain tumor and stroke, but the benefits and limitations of 0.55 T MRI in stereotactic neurosurgery remain unclear. METHODS: Three consecutive scans using 0.55 T, 1.5 T, and 3.0 T field strength were performed in a healthy adult participant and a deep learning reconstruction algorithm (Deep Resolve Boost) was used to optimize imaging parameters for four MRI sequence protocols (T1-weighted 3D Flash, T1-weighted 3D MPRAGE, T2-weighted 2D TSE, and White matter nulling IR 2D TSE). Subsequently, ten additional healthy adult subjects underwent imaging using these optimized parameters at 0.55 T. Images were independently assessed by four blinded investigators (8-items questionnaire; two-way random-effects model with absolute agreement; ICC) to assess resolution, contrast, and visualization of anatomic structures relevant to stereotactic neurosurgery (commissural lines, basal ganglia, internal capsule, thalamus, and striatum). RESULTS: Higher field strength was associated with higher resolution and shorter scan times due to lower acquisition time, echo time and increased inversion time (TI). Particularly for T1-weighted 3D Flash, T1-weighted 3D MPRAGE, and White matter nulling IR 2D TSE. Image quality was rated similarly across all field strengths with an ICC was 0.947(SD±0.531), indicating excellent agreement among the four raters along with a low inter-individual variability. The 0.55 T MRI protocol using a deep learning-based reconstruction algorithms was found to allow sufficient visualization of all relevant structures in all 3 planes (transversal, coronal, sagittal), including location of AC-PC and functional targets, in all participants. CONCLUSIONS: Our preliminary findings suggest that 0.55 T MRI is feasible in visualization of relevant stereotactic anatomical landmarks in healthy subjects, however warrants further evaluation in real-world clinical settings.