Related Experiment Video
Updated: Oct 4, 2025
High-resolution Functional Magnetic Resonance Imaging Methods for Human Midbrain
Published on: May 10, 2012
Evaluation of motion artifacts in brain magnetic resonance images using convolutional neural network-based prediction
Hajime Sagawa1,2, Koji Itagaki1, Tatsuhiko Matsushita1,3
1Kyoto University Hospital, Division of Clinical Radiology Service, Kyoto, Japan.
Abstract:
Purpose: Motion artifacts in magnetic resonance (MR) images mostly undergo subjective evaluation, which is poorly reproducible, time consuming, and costly. Recently, full-reference image quality assessment (FR-IQA) metrics, such as structural similarity (SSIM), have been used, but they require a reference image and hence cannot be used to evaluate clinical images. We developed a convolutional neural network (CNN) model to quantify motion artifacts without using reference images. Approach: The brain MR images were obtained from an open dataset. The motion-corrupted images were generated retrospectively, and the peak signal-to-noise ratio, cross-correlation coefficient, and SSIM were calculated. The CNN was trained using these images and their FR-IQA metrics to predict the FR-IQA metrics without reference images. Receiver operating characteristic (ROC) curves were created for binary classification, with artifact scores indicating the need for rescanning. ROC curve analysis was performed on the binary classification of the real motion images. Results: The predicted FR-IQA metric having the highest correlation with the subjective evaluation was SSIM, which was able to classify images requiring rescanning with a sensitivity of 89.5%, specificity of 78.2%, and area under the ROC curve (AUC) of 0.930. The real motion artifacts were classified with the AUC of 0.928. Conclusions: Our CNN model predicts FR-IQA metrics with high accuracy, which enables quantitative assessment of motion artifacts in MR images without reference images. It enables classification of images requiring rescanning with a high AUC, which can improve the workflow of MR imaging examinations.
Insights
A new convolutional neural network (CNN) model quantifies motion artifacts in brain MR images without a reference. This AI tool accurately predicts image quality and identifies scans needing rescans, improving MRI workflow.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Image Quality Assessment
Background:
- Motion artifacts in MR imaging are subjectively evaluated, leading to poor reproducibility and high costs.
- Existing full-reference image quality assessment (FR-IQA) metrics require a reference image, limiting their use in clinical settings.
- Developing an automated, objective method for motion artifact quantification is crucial for efficient MR imaging.
Purpose of the Study:
- To develop and validate a convolutional neural network (CNN) model for quantifying motion artifacts in brain MR images without requiring a reference image.
- To assess the model's ability to predict FR-IQA metrics and classify images that necessitate rescanning.
- To enhance the workflow efficiency of MR imaging examinations through objective artifact assessment.
Main Methods:
- Brain MR images were sourced from an open dataset, with motion artifacts introduced retrospectively.
- A CNN model was trained to predict FR-IQA metrics (e.g., SSIM) using corrupted images and their calculated metrics.
- Receiver operating characteristic (ROC) curves were employed for binary classification of images requiring rescanning.
Main Results:
- The CNN model accurately predicted FR-IQA metrics, with SSIM showing the highest correlation to subjective evaluations.
- The model achieved high performance in classifying images needing rescans, with an AUC of 0.930 for predicted metrics and 0.928 for real motion artifacts.
- Sensitivity and specificity for rescanning classification reached 89.5% and 78.2%, respectively.
Conclusions:
- The developed CNN model provides accurate, quantitative assessment of motion artifacts in MR images without a reference.
- This AI-driven approach enables reliable classification of images requiring rescanning, potentially optimizing MR imaging workflows.
- The model offers a reproducible and efficient alternative to subjective artifact evaluation.
