QC-Automator: Deep Learning-Based Automated Quality Control for Diffusion MR Images.

Area of Science:

• Medical Imaging
• Neuroimaging
• Artificial Intelligence

Background:

• Diffusion MRI (dMRI) data quality assessment is critical for accurate analysis and reliable results.
• Manual quality control is subjective, time-consuming, and impractical for large-scale studies.
• Automated methods are needed to ensure robust and efficient dMRI data processing.

Purpose of the Study:

• To develop and validate a deep-learning-based automated quality control (QC) tool for dMRI data.
• To address the need for efficient and objective artifact detection in large dMRI datasets.
• To improve the reliability of dMRI image analysis by mitigating artifact impact.

Main Methods:

• Development of QC-Automator, a deep-learning tool utilizing convolutional neural networks and transfer learning.
• Training the model on a large dataset of approximately 332,000 dMRI slices from 155 subjects and 5 scanners.
• The tool is designed to detect a variety of artifacts, including motion, ghosting, and susceptibility-induced distortions.

Main Results:

• QC-Automator achieved 98% accuracy in detecting artifacts in dMRI data.
• The developed method demonstrated high efficiency and speed for processing large datasets.
• The tool proved replicable across datasets with varying acquisition parameters.

Conclusions:

• QC-Automator provides an accurate, fast, and automated solution for dMRI data quality control.
• This tool enhances the efficiency and reliability of dMRI analysis in large-scale research.
• The automated artifact detection facilitates improved data quality and more robust neuroimaging research outcomes.