Clinical Impact of MRI Artifacts on Neuroimaging Cases in Nigerian Hospitals: A Retrospective Study with Protocol Optimization Recommendations

Abstract

Magnetic resonance imaging (MRI) artifacts impact diagnostic accuracy in low-resource settings, where scanner access and maintenance differ from high-income regions. Understanding artifacts distribution is crucial to identifying common patterns, sequence-, orientation- and gender-specific issues, allowing targeted improvements. This study analyzes 100 brain MRI scans (50 low-field/0.36T, 50 high-field/1.5T) from Ibadan, Nigeria, to characterize artifact prevalence by sequence, orientation, and gender. Using Python-based heatmaps and interpretable machine learning, we quantified artifact patterns and identified key predictors. Motion artifacts were the most frequent (31.4% in 1.5T; 27.1% in 0.36T), particularly in axial T1/T2 sequences. Notably, hardware-related artifacts (e.g., RF inhomogeneity) were rare, underscoring operational efficiency. Machine learning models, despite limited dataset size, highlight sequence type as the top artifact predictor (feature importance: 0.72). We propose protocol adjustments (e.g., prioritizing FLAIR over T1/T2) may reduce artifacts by 30% for high fields and not less than 26% for low fields (prioritizing T2* over T1/T2). Our findings provide actionable insights for radiologists in resource-constrained environments, bridging a critical gap in global MRI quality assessment.