Sheng Chen1, Min Zhang2, Liping Yao3, Wentao Xu2. 1. School of Optical Electrical and Computer Engineering & Engineering Research, Center of Optical Instrument and System, Ministry of Education, University of Shanghai for Science and Technology, Shanghai, China. chnshn@hotmail.com. 2. School of Optical Electrical and Computer Engineering & Engineering Research, Center of Optical Instrument and System, Ministry of Education, University of Shanghai for Science and Technology, Shanghai, China. 3. Xinhua Hospital, School of Biomedical Engineering, Shanghai Jiaotong University, Shanghai, China. itaa2004@sina.com.
Abstract
PURPOSE: To present an automated method for detecting endotracheal (ET) tubes and marking their tips in portable chest radiography (CXR) for intensive care units (ICUs). METHODS: In this method, the lung region is first estimated and then the spine is detected between the right lung and the left lung. Because medical tubes are inserted into the body through the throat, the region of interest (ROI) is obtained across the spine. A seed point is determined in the cervical region of the ROI, and then the line path is selected from the seed point. In order to detect ET tubes, the ICU CXR image is preprocessed by contrast-limited adaptive histogram equalization. Then, a feature-based threshold method is applied to the line path to determine the tip location. A comparison to the method by use of Hough transform is also presented. The distance (error) between the detected locations and the locations annotated by a radiologist is used to evaluate the detection precision for the tip location. RESULTS: The proposed method is evaluated using 44 images with ET tubes and 43 images without ET tubes. The discriminant performance for detecting the existence of ET tubes in this study was 95 %, and the average of detection error for the tip location was approximately 2.5 mm. CONCLUSIONS: The proposed method could be useful for detecting malpositioned ET tubes in ICU CXRs.
PURPOSE: To present an automated method for detecting endotracheal (ET) tubes and marking their tips in portable chest radiography (CXR) for intensive care units (ICUs). METHODS: In this method, the lung region is first estimated and then the spine is detected between the right lung and the left lung. Because medical tubes are inserted into the body through the throat, the region of interest (ROI) is obtained across the spine. A seed point is determined in the cervical region of the ROI, and then the line path is selected from the seed point. In order to detect ET tubes, the ICU CXR image is preprocessed by contrast-limited adaptive histogram equalization. Then, a feature-based threshold method is applied to the line path to determine the tip location. A comparison to the method by use of Hough transform is also presented. The distance (error) between the detected locations and the locations annotated by a radiologist is used to evaluate the detection precision for the tip location. RESULTS: The proposed method is evaluated using 44 images with ET tubes and 43 images without ET tubes. The discriminant performance for detecting the existence of ET tubes in this study was 95 %, and the average of detection error for the tip location was approximately 2.5 mm. CONCLUSIONS: The proposed method could be useful for detecting malpositioned ET tubes in ICU CXRs.
Keywords:
Computer-aided detection; Endotracheal tube; ICU; Tip detection
Authors: Michal Sofka; Jens Wetzl; Neil Birkbeck; Jingdan Zhang; Timo Kohlberger; Jens Kaftan; Jérôme Declerck; S Kevin Zhou Journal: Med Image Comput Comput Assist Interv Date: 2011