(Sunday, 7/29/2018) 3:00 PM - 6:00 PM
Room: Exhibit Hall
Purpose: MRI is a powerful noninvasive imaging modality with >60 million scans performed each year, and 25% of those are brain scans. MRI is susceptible to number of artifacts that degrade image quality and reduce diagnostic confidence. Evaluating image quality at the time of acquisition would enable early detection and re-acquisition of the affected images, thus avoiding the need for patient callbacks. We seek to develop, demonstrate, and evaluate a framework for automated and real-time image quality assurance (QA) for MRI.
Methods: A fast image analysis platform was integrated into the workflow of a 3.0 Tesla MRI scanner to perform near-real-time assessment of image quality. Immediately after acquisition of the 3D T1-weighted brain MRI, the images were extracted and transferred to the processing platform. The quality assessment pipeline extracted image features depicting various aspects of image quality (e.g. signal-to-noise ratio, ghosting, contrast, signal leakage, etc.). The features were used as input to a pre-trained classifier for evaluating image quality. The results from quality assessment were imported back into the scanner and displayed to the MRI operator, flagging images of suboptimal quality. As assessment was performed while the patient was still in the scanner, repeat scanning of the affected images is feasible.
Results: Same-session quality assessment was successfully performed on patient images. The analysis was completed in 13-15 minutes, providing same-session evaluation (the study typically requires 45-60 min), and allowing the re-acquisition of corrupted images in the same scan session.
Conclusion: We have demonstrated the feasibility of same-session quality assessment to assure the quality of structural brain images. To further decrease the turn-around time to few minutes, improved computational resources, efficient algorithms, and streamlined workflow will be investigated. Real-time QA is expected to reduce patient callbacks and delays in instituting treatment, and increase the diagnostic confidence and quality of quantitative MRI.
Keywords
Quality Control, Image Artifacts, Image Analysis
Taxonomy
Contact Email
