Evaluating Machine Learning Based Automatic Quality Assurance of Magnetic Resonance Images in Healthcare System

Magnetic Resonance Imaging (MRI) is essential for diagnosing and monitoring neurological disorders due to its superior soft tissue contrast and high spatial resolution; however, the slow acquisition process poses challenges to clinical efficiency. Advancements in temporal encoding, compressed sensing, and parallel imaging have improved acquisition speed, but consistent image quality is crucial for accurate diagnosis and treatment. This study focuses on enhancing automated quality assurance (QA) in MRI brain imaging using machine learning (ML) models to improve tumour classification accuracy and reduce manual intervention. A Kaggle MRI brain scan dataset is preprocessed through resizing, cropping, and data augmentation, followed by the evaluation of multiple models, including MobileNetV2, ResNet50, and Random Forest, using metrics such as accuracy, precision, recall, and F1-score. MobileNetV2 outperforms the other models with a classification accuracy of 96%, demonstrating the potential of lightweight architectures for reliable QA automation. The research contributes to streamlining clinical workflows and improving diagnostic precision by showcasing the effectiveness of AI-driven MRI quality enhancement frameworks.