Detection of Samples Out of Training Distribution: Rejection of Potential Erroneous Local SAR Predictions
Ettore Flavio Flavio Meliado1,2,3, Alexander A.J. Raaijmakers1,2,4, P.R. A.J. Luijten1, and C.A.T. A.J. van den Berg2,5
1Department of Radiology, University Medical Center Utrecht, Utrecht, Netherlands, 2Computational Imaging Group for MR diagnostics & therapy, Center for Image Sciences, University Medical Center Utrecht, Utrecht, Netherlands, 3Tesla Dynamic Coils BV, Zaltbommel, Netherlands, 4Biomedical Image Analysis, Dept. Biomedical Engineering, Eindhoven University of Technology, Eindhoven, Netherlands, 5Department of Radiotherapy, Division of Imaging & Oncology, University Medical Center Utrecht, Utrecht, Netherlands
Recognize potential hazardous situations far from the modeled MR examination scenarios is crucial for RF safety applications and especially for deep-learning-based approaches. Last year we presented a Bayesian deep-learning approach for local SAR assessment. This approach allowed accurate local SAR estimations and returned reliable feedbacks on the error/uncertainty of the estimates for the MR examination scenario observed during training. However, it also showed the dangers of using the predicted uncertainty to identify out-of-training MR examination scenarios. In this study, we propose a simple approach to detect/reject potential erroneous local SAR predictions due to out-of-training transmit array and/or anatomical variations.
This abstract and the presentation materials are available to members only;
a login is required.