A Lightweight Deep Learning Framework for Automatic MRI Data Sorting and Artifacts Detection.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Deep Learning

Background:

• Managing large datasets of brain MRI scans is challenging.
• Automated sorting and artifact detection are crucial for efficient data analysis.
• Existing methods may require significant computational resources.

Purpose of the Study:

• To develop a lightweight, deployable deep learning system for automated brain MRI sorting.
• To enable accurate detection of common MRI artifacts.
• To support big medical imaging data research and database management.

Main Methods:

• Utilized a pre-trained EfficientNetB0 as a feature extractor.
• Employed a multilayer perceptron (MLP) with four hidden layers as a classifier.
• Evaluated performance using precision, recall, F1-Score, accuracy, trainable parameters, and FLOPs.

Main Results:

• Achieved high overall performance: precision (0.983), recall (0.926), F1-Score (0.950), and accuracy (0.991).
• Outperformed four other CNN-based models in classification and model size.
• Demonstrated the smallest number of trainable parameters and FLOPs among compared models.

Conclusions:

• The proposed model accurately sorts brain MRI scans and detects artifacts.
• It requires minimal computational resources, making it easily deployable.
• The system can serve as a valuable tool for big medical imaging data research.