Continuous MRI Coil Quality Control Using Clinical Imaging Data

Abstract

PurposeCurrent testing requirement for individual coil elements are only once per year. Over time the coils degrade due to near continuous use and mishandling, causing elements to become faulty or die. Faulty coil elements may not immediately present themselves in clinical images, but do degrade the quality of the image. When enough elements have failed or degraded this can reduce diagnostic ability. This project serves to create an automated quality control process that tracks the performance of individual coil elements across a fleet of MRI scanners on a daily basis using raw clinical MRI data. Methods Utilizing third-party software we are able to collect the raw individual coil data from all localizer scans on the scanners and transfer this data to a network drive. Another computer then accesses this drive to process each of the localizer scans. The images are reconstructed from the raw k-space data, segmented, and the signal to noise ratio (SNR) is calculated. This SNR calculation is corrected for differences in acquisition parameters such that it can be trended across patients and over time. Results MR data was acquired using a variety of acceleration techniques for example parallel imaging and partial Fourier acquisition. Our code was designed to be robust enough to be able to correctly reconstruct, segment, and process 85% of scans collected across four different MRI scanners. Segmentation code was able to automatically and accurately segment scans across a wide range of anatomy without any physicist or technician input. Corrections applied to SNR placed over 90% of functioning coil scans within the 100-300 range for SNR. Setting a threshold at a corrected SNR value of 50 allowed for the detection of three faulty coils in the span of three weeks before clinical indications began to show.Conclusion This work demonstrates that we can detect dead elements from clinical imaging data before otherwise apparent to physicians. Catching these failures early allows for replacement parts to be ordered and prevents patients from needing to be rescheduled. Many flagged scans were identified that were low SNR due to patient positioning rather than faulty coil elements. These cases were easily checked by physics to ensure that the coils were working properly.