In computed tomography scans, automatic exposure control (AEC) is frequently used to minimize radiation dose exposed to patients while keeping image quality [1]. Traditional AEC modulates tube current of each projection in both angular and longitudinal direction according to the statistics of projection. However, patient risk is not fully considered by traditional AEC, as patient risk is mainly defined by organ-specific dose during the scan. Organ-wise TCM is then proposed to evaluate the patient risk when modulating the tube current and for this end, segmentation of all organs-at-risk are needed [2][3]. In this talk, some research on multi-organ segmentation using deep learning (U-Net [4][5]) is discussed, together with the result of some preliminary experiments and the discussion.

References:
[1] Kalra, Mannudeep K., et al. "Strategies for CT radiation dose optimization." Radiology 230.3 (2004): 619-628.
[2] ICRP, 2007. Managing Patient Dose in Multi-Detector Computed Tomography (MDCT). ICRP Publication 102. Ann. ICRP 37 (1).
[3] ICRP, 2007. The 2007 Recommendations of the International Commission on Radiological Protection. ICRP Publication 103. Ann. ICRP 37 (2-4).
[4] Kerfoot, Eric, et al. "Left-ventricle quantification using residual U-Net." International Workshop on Statistical Atlases and Computational Models of the Heart. Springer, Cham, 2018.
[5] Ronneberger, Olaf, Philipp Fischer, and Thomas Brox. "U-net: Convolutional networks for biomedical image segmentation." International Conference on Medical image computing and computer-assisted intervention. Springer, Cham, 2015.