5th Dutch Bio-Medical Engineering Conference 2015
22-23 January 2015, Egmond aan Zee, The Netherlands
13:30   Cardiovascular Mechanics II
15 mins
Ralph Wijshoff, Wouter Peeters, Gerrit Jan Noordergraaf, Massimo Mischi, Ronald Aarts
Abstract: Detecting return of spontaneous circulation (ROSC) during cardiopulmonary resuscitation (CPR) typically involves pulse checks by manual palpation, which is technically challenging and interrupts the chest compressions [1]. In an experimental automated-CPR study on pigs, retrospective spectral analysis shows photoplethysmography (PPG) as a feasible option to measure a spontaneous pulse during compressions [2]. Here, based on automated-CPR porcine data from Ref. [2], we present an algorithm which removes the compression component from the PPG signal. The time trace of the resulting compression-free PPG signal allows the clinician to directly observe absence or presence of a spontaneous pulse during compressions. The compression-free PPG signal was obtained by subtraction of the compression component which was modelled by a harmonic series. The algorithm determined the instantaneous compression rate from the trans-thoracic impedance signal. The instantaneous compression rate was used as the fundamental frequency of the harmonic series. The amplitudes of the harmonic components were estimated via a least mean-squares algorithm. The results obtained from seven animals with ROSC showed that removal of the compression component was feasible. The compression-free PPG signal showed absence of a spontaneous pulse during cardiac arrest, and showed presence of a spontaneous pulse when the heart started beating at a pulse rate different than the compression rate. Detecting absence of a spontaneous pulse may prevent interrupting compressions for futile pulse checks. Detecting presence of a spontaneous pulse may guide stopping compressions to reduce the risk of refibrillation, which is associated with persisting compressions on a beating heart. We conclude that a compression-free PPG signal can be obtained and can facilitate the clinician in detecting absence or presence of a spontaneous pulse during compressions. This may directly improve survival and neurological outcome by limiting no-compression time. Further research should determine the potential of PPG during CPR in humans and the performance of the compression removal algorithm during manual CPR.
15 mins
Jelle Schrauwen, Dion Koeze, Jolanda Wentzel, Frans van de Vosse, Anton van der Steen, Frank Gijsen
Abstract: Introduction: The significance of a human coronary stenosis is determined with the Fractional Flow Reserve, which is derived from the pressure drop 1. A calculated pressure drop has clinical potential when the calculation time is short. We propose a method which quickly calculates the pressure drop based on geometry. Methods: Luminal data of ten mildly diseased human coronary arteries were obtained with intravascular ultrasound. As a first step to evaluate our method, straightened geometries were constructed. The full Navier-Stokes (NS) equations were solved with computational fluid dynamics (CFD) and the resulting pressure drop (Δp_cfd) served as the gold standard. Δp_cfd was computed for Reynolds numbers (Re_in) ranging from 25 to 250. The new method to determine the pressure drop instantaneously (Δp_geometry) relies on the combination of a reduced form of the NS and assumed axial velocity profile in the form of a power law with one free parameter β. The radial profile follows from the continuity equation. Following this approach, the simplified NS can be solved quickly and Δp_geometry only depends on β. We related the local value of β to geometrical features of the coronary arteries with linear regression by comparing Δp_geometry to Δp_cfd. Results: The β-defined velocity profiles performed well compared to CFD. As a result the estimated pressure drop distribution along the centerline followed the gold standard very well. Minor differences in the velocity profile and pressure drop occurred mainly distal to the stenoses. For high physiological flow conditions (Re_in=150) the mean total pressure drop was <Δp_cfd>=479.7 Pa and our method resulted in <Δp_cfd-Δp_geometry>=26.2±32.2 Pa. Overall, there was an excellent agreement with a correlation of r=0.99 at all Re_in 2. Conclusions: The proposed method reduced the calculation time from hours to seconds. It performs very well in capturing the effects of the pressure drop distribution along the centerline. The mean difference in pressure drop between CFD and the proposed method was low, and an excellent correlation was found for the entire flow range. Future work includes applying this method to curved coronary arteries, so it can serve as a geometry-based pressure drop estimation.
15 mins
Antoine Courcelles, André Sclérusat, Philippe Moulin, Guillaume Zahnd
Abstract: Early detection of cardiovascular diseases (CVDs) is a major health issue [1]. However, traditional risk markers such as pulse wave velocity or intima-media thickness have been shown to demonstrate an overall lack of screening performance [2]. During the last decade, longitudinal motion of the arterial wall (i.e. parallel to the blood flow) has been discovered and demonstrated to be associated with cardiovascular (CV) health [3,4]. Accordingly, measures derived from this index are likely to improve the diagnosis of CVDs [5,6]. The aim of this work is to introduce a motion tracking framework, dedicated to investigate the longitudinal wall shear strain (WSS), by assessing the longitudinal motion of the tissues at different depths in the arterial wall, within the layers of the common carotid artery (CCA) in ultrasound (US) B-mode imaging. The method introduced is devised to track several points spread in a mesh, in order to overcome the challenge of tracking a single point within an homogeneous region such as the adventitia layer. A dense displacement field is thus estimated, using a grid overlying the three layers of the arterial wall. Speckle tracking is performed via a specific block matching approach. An a priori information, using a previously validated single-point tracking method [7], is then exploited to guide each tracked points by associating penalties to outliers. With an heuristic algorithm, the final motion is computed depth by depth in the arterial wall. The grid undergoes elastic deformation to fit the movement of tissues. However, the initial shape of the grid is preserved via a term that minimizes the variation between neighboring points as well as via the segmentation of the luminal interface [8]. Finally, a novel parameter α is also introduced to quantify the WSS and corresponds to the angle between two regressions curves, computed from points in the intima-media complex and in the adventitia layer, respectively. The framework was applied on 56 healthy subjects and 20 at-risk patients. Evaluation of the method was carried out by visual inspection of the tracking quality by one analyst. Fourty trackings out of 56 for healthy subjects, and 13 out of 20 for diseased subjects, have been labelled as correct, as the mesh did follow well the actual tissue movement over consecutive card­iac cycles. Tracking quality was better in the healthy population: images are more challenging in at-risk patients due to the lower amplitude of the longitudinal motion and a lower signal to noise ratio. As for the shear, a near-significantly difference could be observed between the two populations using the introduced parameter α (Kruskal-Wallis test p-value: 0.06). This indicates that arteries of at-risk patients are subject to a lower shearing motion, which could be explained by an increased arterial stiffness. In conclusion, we introduced a new mesh-based framework to estimate an index of WSS in B-mode US images of the CCA. This method represents a new potential screening technique to assess CV risk, and the α parameter may be used as a surrogate marker for atherosclerosis. REFERENCES [1] World Health Organization. Cardiovascular diseases (CVDs), Fact sheet number 317. http://www.who.int/mediacentre/factsheets/fs317/en/index.html, March 2013. [2] A. Simon, G. Chironi and J. Levenson. Performance of subclinical arterial disease detection as a screening test for coronary heart disease. Hypertension, 48(3):392–396, 2006. [3] S. Golemati, A. Sassano, M.J Lever, A.A. Bharath, S. Dhanjil, and A.N. Nicolaides. Carotid artery wall motion estimated from b-mode ultrasound using region tracking and block matching.Ultrasound in Medicine and Biology, 29(3):387–399, 2003. [4] M. Cinthio, Å.R. Ahlgren, J. Bergkvist, T. Jansson, H.W. Persson, and K. Lindström. Longitudinal movements and resulting shear strain of the arterial wall.American Journal of Physiology , 291(1):H394–H402, 2006. [5] S. Svedlund, C. Eklund, P. Robertsson, M. Lomsky, and L.M. Gan. Carotid artery longitudinal displacement predicts 1-year cardiovascular outcome in patients with suspected coronary artery disease. Arteriosclerosis, Thrombosis, and Vascular Biology, 31(7):1668–1674, 2011. [6] G. Zahnd, L. Boussel, A. Marion, M. Durand, P. Moulin, A. Sérusclat, and D. Vray. Measurement of two-dimensional movement parameters of the carotid artery wall for early detection of arteriosclerosis: a preliminary clinical study. Ultrasound in Medecine & Biology, 37(9):1421–1429, 2011. [7] G. Zahnd, M. Orkisz, A. Sérusclat, P. Moulin and D. Vray. Evaluation of a Kalman-based block matching method to assess the bi-dimensional motion of the carotid artery wall in B-mode ultrasound sequences, Medical Image Analysis, 17(5):573–585, 2013. [8] G. Zahnd, M. Orkisz, A. Sérusclat, P. Moulin and D. Vray. Simultaneous extraction of carotid artery intima-media interfaces in ultrasound images - Assessment of wall thickness temporal variation during the cardiac cycle, International Journal of Computer Assisted Radiology and Surgery, 9(4):645–658, 2014.
15 mins
Renate Boekhoven, Louise Marais, Peter Brands, Frans van de Vosse, Richard Lopata, Pierre Boutouyrie
Abstract: Recent studies have shown that arterial stiffness is related to age and the degree of atherosclerosis. It has been hypothesized that there is no difference in arterial stiffness between normotensive (NT) and hypertensive (HT) patients of same sex and age. This was demonstrated in a previous study by analysing pressure-diameter curve data and estimating the incremental Young’s modulus [1]. In this study, 2D ultrasound strain imaging and elastography were applied to verify these findings. Radio-frequency data of the common carotid artery from 19 NT and 19 HT patients were acquired, using the Mylab 70 US system (ESAOTE, NL). Patient couples were analysed, which were matched to age and sex (mean age: 52 ± 11 yrs; male n = 22; female n = 16). Moreover, applanation tonometry was used to acquire the patient’s carotid pressure curve. 2-D RF-based motion and strain imaging was performed on the RF data. From the displacements of the vessel wall the radial strain in the wall were estimated as well as the mechanical properties, being 1) the distensibility and 2) the incremental shear modulus. For the latter, the artery was modelled as a neo-Hookean solid, and the incremental shear modulus was estimated for each patient couple within a pressure range available for both patients, to allow comparison. Bland-Altman analyses was performed to investigate any possible differences in this population. Strains were analysed at the patient specific systolic pressure, these strains were normalized according to their pulse pressure, again to allow comparison. At systole, lower strains were found in the HT group, although one-way ANOVA indicates no significant difference, (NT = 8.4% ± 3.6; HT = 7.4% ± 3.4. Distensiblity was found to be significantly lower in the HT population, which was in agreement with literature [2]. Regarding the arterial stiffness, no fixed bias was found from the analyses, with (Gdiff = GNT – GHT) Gdiff ¬= 18.7 kPa ± 96.0 kPa, and GNT = 218 kPa ± 63.7 kPa; GHT = 198 kPa ± 71.4 kPa. Moreover, one-way Anova indicates that there is no significant difference in shear moduli between the NT and HT groups. These results confirm the hypothesis that the stiffness in HT patients is similar to NT patients when matching age and sex.
15 mins
Shaoxiong Sun, Rick Bezemer, Jens Muehlsteff, Ronald Aarts
Abstract: Continuous non-invasive blood pressure (BP) monitoring has been under intensive investigation by the clinical and technical community. The photoplethysmography (PPG) signal, measured unobtrusively at distal sites, serves as a popular surrogate. This is due to the fact that it indicates pulsatile blood volume changes and is morphologically similar to the continuous BP signal. From PPG and ECG, pulse arrival time (PAT) based methods are widely studied. However, PAT is determined not only by BP but also vasoregulation and vascular properties, making it an unstable estimator per se. Multiple features derived from a PPG signal have been used to estimate BP, but they either require calibration, fail to validate its ability to track BP, or ignore the abundant information contained within the PPG waveform [1,2]. In this study, we introduce BP changes in 18 healthy subjects, explore a number of features from various perspectives, and determine the performance of tracking Systolic Blood Pressure (SBP) in the absence of calibration. Besides PAT, we include timing features (time from systolic slope to dicrotic notch), and amplitude features (normalized systolic slope mean, variance and diastolic slope mean, variance), spectral features (Teager energy mean, variance, skewness). We fuse these features using Random Forest as regression model, and obtain an estimate of SBP every 30 seconds. By using leave-one-out cross-validation (LOOCV), recursively training on 17 subjects and testing on 1 subject, the resulting correlation coefficient is 0.85 and R-squared is 0.31. With a very limited number of subjects, we build a universal model to estimate SBP. This model is applied to all subjects without calibration, to estimate and track variations in their SBP. The result is promising and further work should investigate the potential of sensor fusion and PPG waveform analysis for blood pressure estimation. Sources of improvement might be found in involving more subjects and creating an appropriate model for fitting and analyzing PPG waveforms.
15 mins
Harm Nieuwstadt, Lambert Speelman, Stein Fekkes, Chris de Korte, Jolanda Wentzel, Anton van der Steen, Frank Gijsen
Abstract: Introduction: Biomechanical analysis is an effective approach to assess which carotid atherosclerotic plaques are at risk of rupture which can lead to stroke. In vivo magnetic resonance imaging (MRI) allows plaque segmentation of various components such as lipid-rich necrotic core (LRNC) and fibrous tissues. Segmentation data are used as input for finite element analysis (FEA) to compute the peak cap stress. Apart from geometrical information, knowledge of the patient-specific tissue constitutive relations is vital for FEA. We propose a new, noninvasive, patient-specific elasticity estimation methodology which consists of obtaining the plaque geometry with MRI, measuring the intraplaque spatial strain distribution with ultrasound elastography, and finally solving the inverse FEA problem. Methods: To assess the feasibility of our methodology, we constructed a pipeline for a fully simulated analysis providing us a ground truth comparison (see Figure). Histological cross-sections from carotid plaques (n=12) were obtained. First, 2D biomechanical models (i.e., ground truths) were generated by segmenting the high-resolution histological images and assigning neo-Hookean material properties (Youngs moduli: lipid 30 kPa, intima 720 kPa, wall 1500 kPa). 2D strain fields under constant intraluminal pressure (80 to 120 mmHg) were computed with Abaqus 6.11, and re-sampled to an ultrasound elastography resolution (0.15x0.15 mm2) to mimic an elastography measurement. Realistic elastographic noise was added (SNRe = 4 at the strain-filter plateau). The geometries of the 2D biomechanical models were subjected to MRI simulations (T1-weighted, 0.62 mm in-plane voxel size) in the Jülich extensible MRI simulator (JEMRIS). Manual segmentation was used in the minimization procedure. For the inverse problem, a grid scatter-shot technique was used to minimize the strain least-squares objective function. Results: The results are shown in the Figure. The plaque studied (thick wall, large LRNC) exhibited typical deformations (peak axial engineering strain: 2.5%, average strain: 0.4%). In the MRI, the insufficient contrast between intima (fibrous) and wall (media/adventitia) tissue correctly prevented the inclusion of the wall in the MRI-based model. The inverse FEA problem yielded a unique minimum (cumulative strain residual of 9%): the stiffness of the lipid was estimated as 60 kPa and the stiffness of the intima as 840 kPa. Conclusion: We simulated (1) the cross-sectional deformations of a carotid plaque with realistic biomechanical properties, (2) an in vivo ultrasound elastography measurement, and (3) MR imaging with a clinically-used MRI protocol. Our fully numerical simulation methodology was successfully implemented. While both lipid and intima stiffness values were overestimated, they were in the correct order of magnitude. The remaining 11 patient cross-sections need to be analyzed to draw definitive conclusions. We demonstrated the feasibility of estimating patient-specific, heterogeneous carotid plaque component Youngs Moduli noninvasively using ultrasound elastography, MRI and inverse FEA. This research was supported by the Center for Translational Molecular Medicine and the Netherlands Heart Foundation (PARISk).