Author: Michael Cuddy, Samuel J. Fahrenholtz, Khushnood Hamdani, Saman Jirjes, Robert G. Paden, Jeremiah W. Sanders, William F. Sensakovic, Wolfgang Stefan, Jeffrey Xiao, Yuxiang Zhou 👨🔬
Affiliation: Mayo, Mayo Clinic Arizona, Mayo Clinic 🌍
Title: An automated system for MRI coil performance evaluation
Purpose: To develop an automated quality control (QC) system for MRI coils to assess element-level signal-to-noise ratio (SNR), artifacts, and image quality on an ongoing basis. A primary goal was to more closely monitor certain coil models which tend to experience element failure at a higher frequency.
Methods: Open-source software tools were used to build an application which processes images generated by vendor-supplied coil QA functions on GE and Siemens scanners. DICOM images were sent to our system and images were processed automatically. SNR was measured for composite and individual channel images generated by the system, and the results were stored in a database for continued monitoring and comparison. Python-based image object detection and segmentation techniques were used to quantify signal levels in both phantom and background regions. Additionally, a fast Fourier transform (FFT) function was implemented to detect and troubleshoot suspected RF “spike noise” artifacts. Results were accessible through a web-based portal, allowing for image visualization, data analysis, and alert notifications.
Results: The application effectively identified and verified element failure in various coils across the system, while also facilitating the detection of RF spike anomalies. The application provided an independent analysis of coil performance which could be compared against the vendor’s results. Automating the image analysis saved time and ensured consistent ROI placement across images, while enabling real-time generation and delivery of reports and alerts.
Conclusion: This work developed an additional diagnostic tool to monitor and evaluate MR coil and system performance. It has proven useful in monitoring certain coil models which experience high rates of element failure over time. Moreover, it also provides a method of characterizing coil performance and creating baselines and expected values for signal, noise, and artifacts.