13:30
Imaging - Cardiac System
Chair: Theo van Walsum
13:30
15 mins
|
AUTOMATED THREE-DIMENSIONAL SEGMENTATION AND QUANTIFICATION OF THE RIGHT VENTRICLE
Maartje Nillesen, Arie van Dijk, Han Thijssen, Chris de Korte
Abstract: Assessment of right ventricular function is important for clinical decision making in a variety of diseases, such as pulmonary hypertension. With the introduction of 3D real-time echocardiography, accurate volumetric RV measurements might be performed since inadequate assumptions on the geometry of the right ventricle (RV) are not needed. However, (semi-)automated quantification of RV volumes is difficult in 3D echocardiography because of the complex geometry of the RV. Also, RV quantification often strongly relies on implicit knowledge of the global shape of the RV as supplied by expert users. This might introduce errors in RV volume measurement by overlooking true anatomical details.
This study describes a fully automated segmentation method for the assessment of RV geometry. This method was developed by the authors originally for left ventricular analysis [1] and as no a priori shape information was incorporated, it could be easily adjusted for RV analysis. Real-time 3D transesophageal echocardiographic (TEE) image sequences of 10 patients undergoing percutaneous cryoablation for atrial fibrillation were acquired in radiofrequency format (RF). A 3D adaptive filtering technique based on the (in)homogeneity of the tissue echo signals that optimizes the discrimination between blood and heart muscle [2] was applied. The filtered data were incorporated in a gradient-based deformable model to segment the endocardial surface. Trabeculations were excluded from the RV volume. End-systolic (ES) and end-diastolic (ED) volumes, as well as ejection fraction (EF) were computed from the segmented endocardial surface and compared against volumes manually delineated by an expert cardiologist. The results show that the method yields good correlation and agreement (ED volume: r = 0.86, mean difference ± standard deviation [-8.7 ± 16.3 ml]; ES volume: r = 0.96, [-4.0 ± 11.2 ml]; EF: r = 0.67, [-0.9 ± 9.8 %]) with respect to the results from the reference contours. The technique prevents incorrect RV quantification caused by strong reliance on geometrical assumptions about average shape of the right ventricle.
REFERENCES
[1] M.M. Nillesen, R.G.P. Lopata, W.P. de Boode, I.H. Gerrits, H. J. Huisman, J.M. Thijssen, L. Kapusta, and C.L. de Korte, “In vivo validation of cardiac output assessment in non-standard 3D echocardiographic images”, Phys. Med. Biol., 54(7), pp.1951-1962, (2009).
[2] M.M. Nillesen, R.G.P. Lopata, I.H. Gerrits, H. J. Huisman, J.M. Thijssen, L. Kapusta, and C.L. de Korte, “Segmentation of the heart muscle in 3-D pediatric echocardiographic images”, Ultrasound. Med. Biol., 33(9), pp. 1453-1462 (2007).
|
13:45
15 mins
|
GEOMETRY ASSESMENT OF CAROTID ENDARTERECTOMY SPECIMENS USING 3D ECHO-CT IMAGING
Renate Boekhoven, Richard Lopata, Marc van Sambeek, Frans van de Vosse, Marcel Rutten
Abstract: Endarterectomy is often performed in carotid arteries of patients with advanced atherosclerosis, regardless of knowledge about the stability of the plaque. To date, plaque stability cannot be assessed accurately in vivo, leading to a large number of false positives. To improve the understanding of plaque rupture, techniques to determine the geometrical, morphological, and mechanical properties of atherosclerotic plaques in vitro need to be developed. In this study, as a first step, a new experimental tool to assess 3D plaque geometry using 2D ultrasonography has been evaluated. Micro-CT (µCT) was used for validation.
One polyvinyl-alcohol vessel phantom (asymmetric lumen), one porcine carotid artery and four human atherosclerotic carotid specimens, obtained from endarterectomy (approved by the local ethics committee), were studied. The latter two segment types were pressure fixed. To avoid shadowing effects due to plaque calcifications, longitudinal cross-sections were imaged over 360º in steps of 10º. For this purpose, an ultrasound device (MyLab70, Esaote, Maastricht, The Netherlands) equipped with a linear array probe and RF interface, was used. Next, as a validation, the specimens were imaged in a µCT-scanner (Scanco Medical, Brütisellen, Switserland).
Polar transverse cross-sections containing 360° of near wall and far wall data were created by combining single RF-lines from each longitudinal position. Segmentation of the inner and outer wall was performed with a so-called sustain-attack-filter (SAF). The polar data were projected and interpolated on a Cartesian grid (taking the rotation axis into account) resulting in smooth transverse cross-sections. Normalized 3D cross-correlation was used to register the µCT and US segmentation results, accounting for both translation and rotation. To quantify the quality of the US-based 3D volume, the 3D similarity index (SI) was calculated, regarding the µCT data as the golden standard.
The SAF was able to segment both outer and inner wall of the vessel phantom and carotid segments. The resulting images show the entire arterial wall in 3D at the highest US resolution possible. Results of the image registration indicate good agreement between the ultrasound and the µCT-based geometries of three different segment types. ISI of the phantom was 0.94; ISI of the healthy carotid artery was 0.79; and the average ISI of the human endarterectomy samples was 0.74 (n=4).
In contrast to conventional ultrasound imaging, the method proposed in this study does not suffer from acoustic shadowing effects present when imaging stenotic segments. Furthermore, the method allows future dynamic measurements with high temporal and spatial resolution to determine mechanical properties of atherosclerotic plaque in an in vitro setting.
|
14:00
15 mins
|
SIMULATING CAROTID MRI: HOW ACCURATE CAN WE QUANTIFY ATHEROSCLEROTIC PLAQUE COMPONENTS IN VIVO?
Harm Nieuwstadt, Tom Geraedts, Eline Kooi, Ton van der Steen, Jolanda Wentzel, Frank Gijsen
Abstract: Carotid atherosclerosis is a disease characterized by plaque formation in the carotid bifurcation. Vulnerable plaques, consisting of a large lipid-rich necrotic core (LRNC) separated by a thin fibrous cap (FC) from the lumen, are most prone to rupture and can be visualized in vivo by carotid MRI [1]. How accurate MRI can quantify plaque components such as thin FC’s and LRNC’s in vivo, remains unknown because of the lack of an accurate ground truth on the sub millimeter scale. To circumvent this problem, we chose a novel approach by simulating carotid MRI using the open-source package JEMRIS [2].
We simulated an in vivo T1W gadolinium contrast enhanced MRI protocol, specifically designed to image FC’s. We simulated identical timings, turbo-spin echo factor, acquired in-plane voxel dimensions and k-space filling. A set of 33 ground truth vulnerable plaque geometries derived from cross-sectional histological data from 12 patients were used as 2D sample models for the MRI simulations. Segmentation on simulated images was performed by 3 expert MR readers and measurements were compared to the ground truth by correlation coefficient (R) and within readers by the intraclass correlation coefficient (ICC).
MR readers segmented the lumen with high correlation and excellent agreement (R = 0.996, ICC = 0.99). Measured vessel wall area correlated well (R = 0.96, ICC = 0.94), but was found to be overestimated by 15%. MR readers were found to systematically under predict LRNC area by -31%, but their measurements correlated well (R = 0.95, ICC = 0.94). Measured FC thickness showed a weak correlation (R = 0.71, ICC = 0.69). FC’s smaller than 0.6 mm were severly overestimated in thickness by 201 ± 217%, where FC’s between 0.6 and 0.9 mm were measured more accurate and slightly underestimated: -6 ± 15%. We conclude that in vivo MRI can accurately quantify plaques with regard to vessel wall area and LRNC, but shows limitations for thin FC measurements. This might influence the reliability of in vivo MRI when assessing vulnerable plaque risk by quantifying FC thickness.
This research was supported by the Center for Translational Molecular Medicine and the Netherlands Heart Foundation (PARISk).
REFERENCES
[1] H.R. Underhill et al., Nature Reviews in Cardiology, 7, pp. 165–173, (2010).
[2] T. Stocker et al., Magnetic Resonance in Medicine, 64, pp. 186–193, (2010).
|
14:15
15 mins
|
MEASURING REPERFUSION OF THE HAND OF PATIENTS UNDERGOING CORONARY ARTERY BYPASS SURGERY USING LASER SPECKLE CONTRAST ANALYSIS: AN OBJECTIVE ALLEN’S TEST.
Stefan Sandker, Erwin Hondebrink, Jan Grandjean, Wiendelt Steenbergen
Abstract: Introduction
The radial artery (RA) has become a routinely used conduit for coronary artery bypass graft surgery (CABG) [1]. Prior to surgery the Allen’s test is performed to test the patency of the ulnar artery (UA). Pressure is applied to the RA and UA and the patient is asked to make a fist. Pressure on the UA is released and the red color of the hand should return within 5 seconds, a negative Allen’s test [2]. If the test is positive, the RA should not be used as bypass graft. The predictive value of a positive test is 53% [3]. In this pilot study we investigated if laser speckle contrast analysis (LASCA) provides a more objective determination of the reperfusion time compared to the conventional Allen’s test.
Methods
When the hand is illuminated with coherent laser light, here a 660nm 75mW continuous wave laser, the backscattered light will result in constructive and destructive interference consisting of bright and dark areas, speckles [4]. This speckle pattern will change due to movement of red blood cells. LASCA uses these changes to visualize the perfusion during the Allen’s test at 30 frames/second. The average perfusion value of each frame of a 4x4cm area on the palmar side of the hand is calculated and set against time. The reperfusion time is the time to reach maximal perfusion after releasing pressure of the UA and is calculated using the first derivative of a 6th order polynomial curve fit. The calculated reperfusion time using LASCA is compared to the conventional Allen’s test of patients undergoing CABG (n=17) as performed by the nurse practitioner.
Results
LASCA measurements showed a negative Allen’s test of both hands of eleven patients. Six had a borderline reperfusion time of 5 – 5.5 seconds and/or a positive Allen’s test of one or both hands. These results were consistent with the conventional Allen’s test. Furthermore, differences in reperfusion of different parts in the measured area of the hand were visible.
Conclusion
LASCA is able to visualize reperfusion of the hand and measure a quick, moderate, slow reperfusion response or no reperfusion. It is technically feasible to determine the reperfusion time of the hand. In the future, LASCA could be a useful and objective tool to assess ulnar collateral blood supply to the hand prior to harvesting of the radial artery as a bypass graft.
REFERENCES
[1] Kobayashi J. Radial artery as a graft for coronary artery bypass grafting. Circ J 2009; 73: 1178-1183.
[2] Cable DG, Mullany CJ, Schaff HV. The Allen test. Ann Thorac Surg 1999; 67: 876-877.
[3] Asif M, Sarkar PK. Three-Digit Allen’s Test. Ann Thorac Surg 2007; 84: 686-687.
[4] Draijer M, Hondebrink E, van Leeuwen T, Steenbergen W. Review of laser speckle contrast techniques for visualizing tissue perfusion. Lasers Med Sci 2009; 24: 639-651.
|
14:30
15 mins
|
LUMEN SEGMENTATION OF ATHEROSCLEROTIC CAROTID ARTERIES IN CTA
Hui Tang, Theo van Walsum, Reinhard Hameeteman, Michiel Schaap, Aad van der Lugt, Lucas van Vliet, Wiro Niessen
Abstract: Atherosclerosis is a major vascular disease which is asymptomatic in the early stages of development[1]. Several studies have been performed to predict atherosclerosis progression using information on plaque composition [2, 3] and vessel geometry [4]. In these studies, accurate lumen segmentation of atherosclerotic vessels is required to determine the vessel geometry or to detect a region-of-interest for plaque composition analysis. As a result, lumen segmentation in CTA has received considerable interest. Compared to lumen segmentation in healthy vessels, lumen segmentation in atherosclerotic vessels is far more challenging due to the presence of stenoses and plaques. In this paper, we propose a semi-automatic lumen segmentation method for the segmentation of the lumen in CTA datasets of atherosclerotic carotid arteries. The proposed segmentation method includes two steps: centreline extraction followed by a levelset evolution initialized by the extracted centrelines. We extract the centerline using an iterative minimum cost path approach where the costs are defined by intensity and gradient information [5]. In the second step we utilize a levelset approach, which uses both boundary information and regional intensity information to accurately segment the lumen of atherosclerotic vessels. The method was trained and validated on a publicly available database of 56 carotid arteries. The average Dice similarity coefficient was 90.2%, the mean absolute surface distance was 0.23 mm. Compared to using only boundary information, this method yields better segmentation in severely atherosclerotic vessels.
REFERENCES
1. R. Ross, “Atherosclerosis: an inflammatory disease,” New England Journal of Medicine, vol. 340, pp. 115–126, 1999.
2. J.M.A. Hofman, et al, “Quantification of atherosclerotic plaque components using in vivo mri and supervised classifiers,” Magn. Reson. Med., vol. 55, no. 4, pp. 790–799, 2006.
3. F Liu, et al, “Automated in vivo segmentation of carotid plaque MRI with morphology-enhanced probability maps,” Magn Reson Med, vol. 55, no. 3, pp. 659– 668, Mar. 2006.
4. J.B. Thomas, et al, “Variation in the carotid bifurcation geometry of young versus older adults: Implications for geometric risk of atherosclerosis,” Stroke, vol. 36, no. 11, pp. 2450–2456, Nov. 2005.
5. Hui Tang et al., Multispectral MRI centerline tracking in carotid arteries, Proc. SPIE 7962, 79621N (2011) , 2011.
|
14:45
15 mins
|
CARDIAC SEGMENTATION IN 3D ULTRASOUND IMAGES: BOUNDARY OPTIMIZATION USING TEMPORAL INFORMATION
Anne Saris, Maartje Nillesen, Richard Lopata, Chris de Korte
Abstract: Automated segmentation of three-dimensional (3D) echocardiographic images in patients with congenital heart disease is challenging, because of poor contrast between blood and cardiac tissue locally. Incorporation of a priori knowledge of anatomy is undesirable for patients with congenital malformations. Cardiologists mentally incorporate movement of the heart, using temporal coherence of structures to resolve ambiguities. Therefore, we investigated the merit of temporal cross-correlation for automated segmentation over the entire cardiac cycle.
A 3D iterative cross-correlation algorithm [1] was extensively examined with respect to signal choice (envelope or radio-frequency (RF) data) and window size, in order to obtain optimal contrast of maximum cross-correlation (MCC) values between blood and cardiac tissue in all phases of the cardiac cycle. Both contrast (CNR and Overlap) and boundary-gradient (Acutance) were quantified. Resulting optimal MCC-values were used as additional external force in a deformable model approach [2] to segment the left ventricular cavity in 3D echocardiographic images, in entire systolic phase. MCC-values were tested against and combined with, adaptive filtered, demodulated RF-data. Segmentation results were compared with manually segmented volumes using a 3D Dice similarity index (3D SI).
Results in 3D pediatric echocardiographic images sequences (n = 5) demonstrate that the use of envelope data outperformed RF-data in terms of optimal blood-myocardium Acutance, Overlap and CNR. The use of a relatively small axial window (0.7 – 1.25 mm) resulted in optimal contrast and boundary gradient between the two tissues. Incorporation of MCC-values, either alone or in combination with adaptive filtered, demodulated RF-data, improved automated segmentation of the left ventricular cavity (n = 4). When MCC-values were used as external force in the deformable model, the 3D SI increased in 75% of the cases (average SI increase: 0.71 to 0.82). Results might be further improved by optimizing MCC-contrast locally, in regions with low blood-tissue contrast. Reducing underestimation of the endocardial volume due to MCC processing scheme (choice of window size) and consequential border-misalignment, could also lead to more accurate segmentations.
|
|
