Tonghe Wang1, Yang Lei1, Haipeng Tang2, Zhuo He2, Richard Castillo1, Cheng Wang3, Dianfu Li3, Kristin Higgins1, Tian Liu1, Walter J Curran1, Weihua Zhou4, Xiaofeng Yang5. 1. Department of Radiation Oncology and Winship Cancer Institute, Emory University, Atlanta, GA, 30322, USA. 2. School of Computing, University of Southern Mississippi, Long Beach, MS, 39560, USA. 3. Department of Cardiology, The First Affiliated Hospital of Nanjing Medical University, Nanjing, Jiangsu, China. 4. School of Computing, University of Southern Mississippi, Long Beach, MS, 39560, USA. weihua.zhou@usm.edu. 5. Department of Radiation Oncology and Winship Cancer Institute, Emory University, Atlanta, GA, 30322, USA. xiaofeng.yang@emory.edu.
Abstract
BACKGROUND: The performance of left ventricular (LV) functional assessment using gated myocardial perfusion SPECT (MPS) relies on the accuracy of segmentation. Current methods require manual adjustments that are tedious and subjective. We propose a novel machine-learning-based method to automatically segment LV myocardium and measure its volume in gated MPS imaging without human intervention. METHODS: We used an end-to-end fully convolutional neural network to segment LV myocardium by delineating its endocardial and epicardial surface. A novel compound loss function, which encourages similarity and penalizes discrepancy between prediction and training dataset, is utilized in training stage to achieve excellent performance. We retrospectively investigated 32 normal patients and 24 abnormal patients, whose LV myocardial contours automatically segmented by our method were compared with those delineated by physicians as the ground truth. RESULTS: The results of our method demonstrated very good agreement with the ground truth. The average DSC metrics and Hausdorff distance of the contours delineated by our method are larger than 0.900 and less than 1 cm, respectively, among all 32 + 24 patients of all phases. The correlation coefficient of the LV myocardium volume between ground truth and our results is 0.910 ± 0.061 (P < 0.001), and the mean relative error of LV myocardium volume is - 1.09 ± 3.66%. CONCLUSION: These results strongly indicate the feasibility of our method in accurately quantifying LV myocardium volume change over the cardiac cycle. The learning-based segmentation method in gated MPS imaging has great promise for clinical use.
BACKGROUND: The performance of left ventricular (LV) functional assessment using gated myocardial perfusion SPECT (MPS) relies on the accuracy of segmentation. Current methods require manual adjustments that are tedious and subjective. We propose a novel machine-learning-based method to automatically segment LV myocardium and measure its volume in gated MPS imaging without human intervention. METHODS: We used an end-to-end fully convolutional neural network to segment LV myocardium by delineating its endocardial and epicardial surface. A novel compound loss function, which encourages similarity and penalizes discrepancy between prediction and training dataset, is utilized in training stage to achieve excellent performance. We retrospectively investigated 32 normal patients and 24 abnormal patients, whose LV myocardial contours automatically segmented by our method were compared with those delineated by physicians as the ground truth. RESULTS: The results of our method demonstrated very good agreement with the ground truth. The average DSC metrics and Hausdorff distance of the contours delineated by our method are larger than 0.900 and less than 1 cm, respectively, among all 32 + 24 patients of all phases. The correlation coefficient of the LV myocardium volume between ground truth and our results is 0.910 ± 0.061 (P < 0.001), and the mean relative error of LV myocardium volume is - 1.09 ± 3.66%. CONCLUSION: These results strongly indicate the feasibility of our method in accurately quantifying LV myocardium volume change over the cardiac cycle. The learning-based segmentation method in gated MPS imaging has great promise for clinical use.
Authors: Tonghe Wang; Yang Lei; Yabo Fu; Walter J Curran; Tian Liu; Jonathon A Nye; Xiaofeng Yang Journal: Phys Med Date: 2020-07-29 Impact factor: 2.685
Authors: Tonghe Wang; Yang Lei; Sibo Tian; Xiaojun Jiang; Jun Zhou; Tian Liu; Sean Dresser; Walter J Curran; Hui-Kuo Shu; Xiaofeng Yang Journal: Med Phys Date: 2019-05-21 Impact factor: 4.071
Authors: Yupei Zhang; Yang Lei; Richard L J Qiu; Tonghe Wang; Hesheng Wang; Ashesh B Jani; Walter J Curran; Pretesh Patel; Tian Liu; Xiaofeng Yang Journal: Med Phys Date: 2020-04-03 Impact factor: 4.071
Authors: Xianjin Dai; Yang Lei; Yupei Zhang; Richard L J Qiu; Tonghe Wang; Sean A Dresser; Walter J Curran; Pretesh Patel; Tian Liu; Xiaofeng Yang Journal: Med Phys Date: 2020-06-15 Impact factor: 4.071