Junghwa Kang1, Na Young Shin2, and Yoonho Nam1,2
1Division of Biomedical Engineering, Hankuk University of Foreign Studies, Yongin, South Korea, yongin, Korea, Republic of, 2Seoul St.Mary’s Hospital, Department of Radiology, The Catholic University of Korea, Seoul, South Korea, Seoul, Korea, Republic of
1Division of Biomedical Engineering, Hankuk University of Foreign Studies, Yongin, South Korea, yongin, Korea, Republic of, 2Seoul St.Mary’s Hospital, Department of Radiology, The Catholic University of Korea, Seoul, South Korea, Seoul, Korea, Republic of
We proposed to evaluate the degree of motion artifact on high-resolution magnetic susceptibility contrast images for N1 visualization that is sensitive to patient’s motions. We introduced deep CNN-LSTM network. The proposed method could be helpful in clinical use.
Figure 3. The architecture of CNN-LSTM for motion assessment
Figure 4. Confusion Matrix and ROC analysis (validation set) (a), (c) CNN , (b), (d) CNN-LSTM.