Leveraging redundancy in simultaneous multislice acquisitions to improve spike detection. Issue 6 (9th January 2022)

• Record Type: Journal Article
• Authors: Sidhu, Jagjit Singh; Sakaie, Ken; Shin, Wanyong; Lowe, Mark
• Is Part Of: Magnetic resonance in medicine. Volume 87:Issue 6(2022)
• Page Start: 2972
• Page End: 2978
• Publication Date: 2022-01-09
• Subjects: functional -- outlier detection -- quality assurance -- simultaneous multislice -- spike noise -- MRI; Nuclear magnetic resonance -- Periodicals; Electron paramagnetic resonance -- Periodicals; 616.07548
• DOI: 10.1002/mrm.29150
• Languages: English
• ISSNs: 0740-3194

Abstract

Abstract : Purpose: To improve the performance of low‐level spike noise artifact detection for daily quality assurance protocols by taking advantage of redundancy in simultaneous multislice (SMS) acquisitions. Methods: Magnitude images were transformed into pseudo k‐space images. Time series at each pseudo k‐space point were detrended. A slice was determined to contain spiking artifact if it exceeded an intensity threshold and if all simultaneously acquired slices contained outliers. Results: A total of 401 112 slices were inspected. Of these, 42 showed a spike artifact, based on visual inspection of image data and k‐space data. With an intensity threshold of 4.6 SDs over time for each pseudo k‐space point, all slices containing artifact were correctly flagged, and only 30 slices were incorrectly flagged when using the SMS criterion. Without the SMS criterion, 12 908 slices were incorrectly flagged as containing artifact. Without the SMS criterion, sensitivity to artifact would have to be sacrificed to substantially reduce the number of incorrectly flagged slices. Conclusion: This study demonstrates that the SMS criterion reduced the number of outliers reported to a manageable level while accurately identifying low‐level spike artifacts. Successfully identifying low‐level spikes allows early detection of hardware problems that can be fixed before the problem becomes debilitating and corrupts data. As part of a daily quality assurance protocol, the method prevents the need to retrospectively carry out time‐intensive despiking and reanalysis of data.