Thromboembolic disease is an important vascular disease with a high incidence in the United States. Estimates of incidence in the United States are 398,000 cases of deep venous thrombosis per year and 347,000 pulmonary embolus cases per year. Deaths due to PE are estimated to be about 235,000 per year. The mortality due to untreated clinically apparent PE is approximately 30%. However, if correctly diagnosed and anticoagulant therapy is initiated, mortality drops to below 3%. The advent of multi-slice helical CT imaging has provided a robust, rapid and minimally invasive means to visualize the both the pulmonary and peripheral vasculature with exquisite detail. However, the amount of image data generated for review is large and can be tedious to review, thus potentially limiting diagnostic accuracy and efficiency. Computer aided detection (CAD) techniques could be of potential use in this environment, increasing both the efficiency and accuracy of expert review of the volumetric CT images. In this proposal we will test the feasibility of using a CAD tool to automatically diagnosis thromboembolic disease as manifested in volumetric pulmonary and peripheral CT angiography (CTA) and CT venography (CTV) images. We will construct a database of 450 CTA/V that were acquired to rule out thromboembolic disease, along with corresponding clinical data which will be used to establish the true disease state. We will develop a set of vascular analysis algorithms that will be applied to the CTA/V images to automatically extract a centerline model of the relevant vascular trees. Detailed morphological measurements of the vasculature will be made automatically using the centerline model. These measurements will form the input vector to a computer-aided detection (CAD) algorithms based on artificial neural networks. The CAD algorithms will be used to determine the presence of thrombus or emboli within the vascular trees. The efficacy of the CAD algorithms will be assessed using FROC methods. Performance level of the CAD algorithms will be compared to the clinical radiologists'interpretations of the performance experiment.
|Shukla, Gaurav; Klatzky, Roberta L; Wu, Bing et al. (2017) Integrating images from a moveable tracked display of three-dimensional data. Cogn Res Princ Implic 2:34|
|Chapman, Brian E; Berty, Holly P; Schulthies, Stuart L (2015) Automated generation of directed graphs from vascular segmentations. J Biomed Inform 56:395-405|
|Berty, Holly L; Simon, Marc; Chapman, Brian E (2012) A semi-automated quantification of pulmonary artery dimensions in computed tomography angiography images. AMIA Annu Symp Proc 2012:36-42|
|Park, Sang Cheol; Chapman, Brian E; Zheng, Bin (2011) A multistage approach to improve performance of computer-aided detection of pulmonary embolisms depicted on CT images: preliminary investigation. IEEE Trans Biomed Eng 58:1519-27|
|Chapman, Brian E; Lee, Sean; Kang, Hyunseok Peter et al. (2011) Document-level classification of CT pulmonary angiography reports based on an extension of the ConText algorithm. J Biomed Inform 44:728-37|