Prosthetic Heart Valve Localization and Quantification in CT

PHV in CT Prosthetic heart valves (PHV’s) should in principle last as long as the patient lives, but in practice complications related to these implanted devices may arise. Therefore, patients with such devices undergo regular imaging to assess the implanted device and its functioning. PHVs are routinely assessed by echocardiography and fluoroscopy. For the past decade Cardiac Computed Tomography has also been used [6] since it has been shown to provide additional diagnostic value in specifying the dysfunction cause as well as providing additional anatomical information assisting planning in case of reoperation. Assessment of the device state is a laborious task, which requires manual reformatting of the images and manual measurements. Automating this task reduces the time spent by radiologists on this task. Additionally, these images offer the possibility to assess/quantify various other characteristics of valves, such as motion and deformation, which in turn may be used in studies investigating malfunctioning of these devices.

Our purpose is to develop and evaluate algorithms that can automatically detect, segment and/or quantify parameters of artificial heart valves (e.g. opening /closing angles, deformation of valve/stent over the cardiac cycle). Such software serves two purposes: it may help reduce time spent by the radiologists in assessing the heart valve status in follow-up CTA images, and it may provide information that can be used in studies investigating causes of complications/failures with artificial heart values.

PHV in CT Androulakis et al., TMI 2018 : Cardiac computed tomography (CT) is a valuable tool for functional mechanical heart valve (MHV) assessment. An important aspect of bileaflet MHV assessment is evaluation and measurement of leaflet opening and closing angles. Performed manually however, it is a laborious and time consuming task. In this work, we propose an automated approach for bileaflet MHV leaflet angle computation. This method consists of four steps. After a one click selection of the MHV region on an axial image, an automatic MHV extraction using thresholding and connected component analysis based on voxel intensities is performed. Then the MHV component (valve ring and two leaflets) positions are identified using random sample consensus and least square fitting. Finally, the angles are automatically computed based on the orientation of the components in each timeframe. Five multiphase CT scans from patients with a bileaflet MHV containing between 14 and 17 timepoints were used for development and another fifteen were used for evaluation. The detected MHV components were scored for their overlap with real components as successful or unsuccessful. For successful results, the angles were compared to those measured by a radiologist. Qualitatively evaluated on a dataset of 222 images, a total of 398 out of 444 angle computations (89.6%) were rated as successful. Compared to the angles measured by the radiologist, the successful angles showed a mean difference of 0.54 0 ± 3.63 0 from the manual calculations. The method provides a high success rate and an accurate computation of leaflet opening angles compared to manual measurements.

Associated People