The Kalman filter[57] approach to echocardiographic segmentation has been used in several works by for instance Orderud[86], Bersvendsen[11] and Snare[102].
Since its invention in 1960, the Kalman filter has been used for a variety of purposes, like navigation, control of manufacturing processes and prediction of the paths of celestial objects[43]. The filter does have some limitations and assumptions that are often not true.[56]. Firstly, the Kalman filter has an assumption that the next time step is solely determined by the previous, this is called the Markov property. Secondly, the noise in the system is assumed to be additive Gaussian, although modifications can be made if noise is known to be something else. Finally, the process is assumed to be either linear or locally linear.
The Markov property does not hold for a cardiac image, which is cyclical, meaning that if at time step k to k+1 there is an expansion of the ventricle, that expansion will most likely continue into k+2. Knowing how much, if any, expansion there was between previous time steps would give extra information about where in the cardiac cycle the system is in, and could be used to give more accuracy, but the Kalman filter does typically not allow this.
An assumption on the edge detection procedure developed by Orderud[86] is that the individual edge detections are independent of each other. This is not true, as edge detectors close to each other will likely detect the same objects, and so the results of one heavily correlate with the results of the others. Factors
like general contrast and quality of the image influence all edge detectors, giving a further reason why independence is not true.
Finally, the Kalman algorithm requires a good starting guess, and if the guess is too wrong, the entire algorithm can fail. In situations where there is too much variation, this can be fixed by manual input or a special initialization stage.
Despite these theoretical limitations, the model has in practical applications shown good results over a variety of uses, like the segmentation of the LV, RV and LA, the placement of the mitral valveand the estimation of standard views.
It is computationally efficient and allows for an intuitive and adaptable union of known information about shape and size of what is to be estimated, and the measurements from the image.
4.5 Future Improvements and Applications
For the sharp models, it would be an advantage if the Kalman filter could be used to determine the proper weight for any edge or vertex, instead of it being manually determined ahead of time, as was done in paper I. This could possibly be done by determining how sharp the local curvature is based on the edge detectors, and this could be used as input to the Kalman filter.
The sharp Doo-Sabin methods have the potential to be used in other applications of modeling sharp features, especially objects that have a mix of smooth and sharp edges. This could be for medical segmentation, potentially the valves could be modeled with a sharp centre, but also for non-medical applications. In addition, the sharp Doo-Sabin models could be used for the segmentation of the RV in other modalities, like MRI.
One of the advantages of the standard view algorithm constructed in the second paper is that it can run in real time-after initialization. This means it would be possible to update the standard view at every frame to make sure all features are in view. The algorithm could also be modified to find other landmarks in TEE images.
The four-chamber model Doo-Sabin could also be used on other modalities like MRI or CT. Seeing as other four chamber models have been made for those modalities, it would be an interesting point of comparison. The four-chamber model does have the potential for use in cases where the chamber walls are partially obstructed. While the accuracy would be less than if the chamber was in full view, using the placement and size of the fully visible chambers to guide the partially visible ones. Figure 4.3 shows an example of this, but a rigorous test is necessary.
It is possible to expand on the four-chamber model in order to use it in more applications. By setting further landmarks on the models, other features in the image could be identified. It would also be possible to add the aortic outflow tract or other such features to the model. The four-chamber model could also be used for interventional guidance, helping with determining what part of the heart the tools are positioned at. For a robot guidance system for use in cardiac
Future Improvements and Applications
Figure 4.3: A comparison of a single chamber and four-chamber approach in a case with partially obscured LV. The four-chamber approach is shown on the left, and the single on the right. The four-chamber version of the LV is better at determining the apex and the base, likely helped by the RV and LA placement.
intervention, a four-chamber model fitted to an image of the heart could be used to determine landmark positions that the robot could guide itself from.
Other fitting algorithms could be considered. Chen et al.[23] note several algorithms that combine deformable models and deep learning fitting algorithms, which might prove an interesting alternative.
Chapter 5
Conclusions
This thesis has been part of the INIUS project in cooperation with GE Healthcare.
The aim has been to create new solutions to practical problems, and the premise of the thesis was that this can be done by expanding on GE Healthcare’s framework for echocardiographic segmentation.
Following this idea, several improvements have been made to the framework originally described by Orderud et al. [86]. Two sharp versions of the Doo-Sabin surfaces have been created and an RV model has been made based on those surfaces. A technique for solving the initialization problem of the Kalman filter in some situations has been invented, and a four-chamber model with a new implementation of wall thickness between the chambers has been made. While these were made to solve practical problems, they are also theoretical contributions to the fields of mathematical splines, Kalman filters, and echocardiographic segmentations.
Given the aim of the thesis, these expansions of the framework were applied to several practical problems in the field of ultrasound segmentation. The sharp Doo-Sabin models were used to create more accurate segmentation of the RV, accuracy mainly being measured by the model’s ability to estimate the volume and ejection fraction. Accurate segmentation of the RV is useful to determine RV dysfunction, and RV ejection fraction is a predictor for moderate heart failure.
We have created an algorithm for the automatic determination of standard views in TEE images. The solution to the initialization problem of Kalman filters was vital in order to detect the aorta and mitral valve, the location of which was used to determine standard views in TEE images. Automatic standard views would allow for shorter time being spent on manually determining it during the examination of for instance valvular dysfunction and could be useful for guidance during surgery.
Finally, the four-chamber model was used to accurately and simultaneously segment all four chambers of the heart in ultrasound images. Again volumetric measures were used to test how well the four-chamber model worked, but this model could easily be used to extract landmarks. It could also be used to study how the chambers interact.
Through these applications, we believe we have shown that the expansions of Orderud’s framework allow it to be used in new applications, and have real value when it comes to solving practical segmentation problems, leading to an improved workflow and diagnostic accuracy.
Bibliography
[1] Almeida, N. et al. “Left-Atrial Segmentation From 3-D Ultrasound Using B-Spline Explicit Active Surfaces With Scale Uncoupling”. In: IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Controlvol. 63,
no. 2 (2016), pp. 212–221.
[2] An, S. et al. “A category attention instance segmentation network for four cardiac chambers segmentation in fetal echocardiography”. In:
Computerized Medical Imaging and Graphics vol. 93 (2021), p. 101983.
[3] Andreassen, B. S. et al. “Mitral Annulus Segmentation Using Deep Learning in 3-D Transesophageal Echocardiography”. In:IEEE Journal of Biomedical and Health Informatics vol. 24, no. 4 (2020), pp. 994–1003.
[4] Angelini, E. D. et al. “Segmentation of real-time three-dimensional ultrasound for quantification of ventricular function: A clinical study on right and left ventricles”. In:Ultrasound in Medicine & Biologyvol. 31, no. 9 (2005), pp. 1143–1158.
[5] Apostolakis, S, A., and S, K. “The Right Ventricle in Health and Disease:
Insights into Physiology, Pathophysiology and Diagnostic Management”.
In:Cardiologyvol. 121 (2012), pp. 263–273.
[6] Aune, E. et al. “Normal reference ranges for left and right atrial volume indexes and ejection fractions obtained real-time three-dimensional echocardiography”. In:European journal of echocardiography : the journal of the Working Group on Echocardiography of the European Society of Cardiology vol. 10 (June 2009), pp. 738–44.
[7] Avendi, M., Kheradvar, A., and Jafarkhani, H. “A combined deep-learning and deformable-model approach to fully automatic segmentation of the left ventricle in cardiac MRI”. In:Medical Image Analysisvol. 30 (2016), pp. 108–119.
[8] Bach Cuadra, M., Duay, V., and Thiran, J.-P. “Atlas-based Segmentation”.
In:Handbook of Biomedical Imaging: Methodologies and Clinical Research.
Ed. by Paragios, N., Duncan, J., and Ayache, N. Boston, MA: Springer US, 2015, pp. 221–244.
[9] Ballard, D. “Generalizing the Hough transform to detect arbitrary shapes”.
In:Pattern Recognition vol. 13, no. 2 (1981), pp. 111–122.
[10] Bersvendsen, J. et al. “Automated Segmentation of the Right Ventricle in 3D Echocardiography: A Kalman Filter StateEstimation Approach”.
In:IEEE TRANSACTIONS ON MEDICAL IMAGING vol. 35 (2015), pp. 42–51.
[11] Bersvendsen, J. “Segmentation of cardiac structures in 3-dimensional echocardiography”. PhD thesis. University of Oslo, 2016.
[12] Bersvendsen, J. et al. “Semiautomated biventricular segmentation in three-dimensional echocardiography by coupled deformable surfaces”. In:
Journal of medical imaging (Bellingham, Wash.)vol. 4, no. 2 (Apr. 2017), p. 024005.
[13] Besbeas, P. and Morgan, B. J. T. “Kalman filter initialization for integrated population modelling”. In: Journal of the Royal Statistical Society: Series C (Applied Statistics)vol. 61, no. 1 (2012), pp. 151–162.
eprint:https://rss.onlinelibrary.wiley.com/doi/pdf/10.1111.
[14] Bleeker, G. et al. “Assessing right ventricular function: The role of echocardiography and complementary technologies”. In:Heart (British Cardiac Society)vol. 92 Suppl 1 (May 2006), pp. i19–26.
[15] Budoff, M. J., Achenbach, S., and Duerinckx, A. “Clinical utility of computed tomography and magnetic resonance techniques for noninvasive coronary angiography”. In:Journal of the American College of Cardiology vol. 42, no. 11 (2003), pp. 1867–1878.
[16] Budoff, M. J. et al. “Assessment of Coronary Artery Disease by Cardiac Computed Tomography”. In:Circulation vol. 114, no. 16 (2006), pp. 1761–
1791. eprint: https : / / www . ahajournals . org / doi / pdf / 10 . 1161 / CIRCULATIONAHA.106.178458.
[17] Bølviken, H. S. et al. “Modified Doo-Sabin Modeling of the Right Heart Ventricle”. In:Medical Imaging 2020: Ultrasonic Imaging and Tomography.
Feb. 2020.
[18] Bølviken, H. S. and al., et. “Two methods for modifed Doo–Sabin modeling of nonsmooth surfaces—applied to right ventricle modeling”. In: Journal of Medical Imagingvol. 7, no. 6 (2020), pp. 1–17.
[19] Bølviken, H. S. et al. “Automatic alignment of standard views for transesophageal echocardiographic images”. In: Journal of Medical Imaging vol. 9, no. 5 (2022), p. 057001.
[20] Catmull, E. and Clark, J. “Recursively generated b-spline surfaces on arbitrary topological meshes.” In:Computer-Aided Design vol. 10 (1978), pp. 350–355.
[21] Charnoz, A. et al. “A levelset based method for segmenting the heart in 3D+T gated SPECT images”. In: vol. 2674. May 2003.
[22] Chechani, S., Suresh, R., and Patwardhan, K. A. “Aortic Root Seg-mentation in 4D Transesophageal Echocardiography”. In: MEDICAL IMAGING 2018: COMPUTER-AIDED DIAGNOSIS. Ed. by Petrick, N.
and Mori, K. Vol. 10575. Proceedings of SPIE. Conference on Medical Imaging - Computer-Aided Diagnosis, Houston, TX, FEB 12-15, 2018.
SPIE; DECTRIS Ltd. 2018.
Bibliography [23] Chen, C. et al. “Deep Learning for Cardiac Image Segmentation: A
Review”. In:Frontiers in Cardiovascular Medicinevol. 7 (2020).
[24] Cootes, T. et al. “Active Shape Models-Their Training and Application”.
In: Computer Vision and Image Understanding vol. 61, no. 1 (1995), pp. 38–59.
[25] Curiale, A. H. and al., et. “Fully Automatic Detection of Salient Features in 3-D Transesophageal Images”. In:Ultrasound in Medicine & Biology vol. 40, no. 12 (2014), pp. 2868–2884.
[26] Degel, M. A., Navab, N., and Albarqouni, S. “Domain and Geometry Agnostic CNNs for Left Atrium Segmentation in 3D Ultrasound”. In:
Medical Image Computing and Computer Assisted Intervention – MICCAI 2018. Ed. by Frangi, A. F. et al. Cham: Springer International Publishing, 2018, pp. 630–637.
[27] Derose, T., Kass, M., and Truong, T. “Subdivision Surfaces in Character Animation”. In: Proceedings of the ACM SIGGRAPH Conference on Computer Graphics vol. 32 (1998), pp. 85–94.
[28] Dikici, E. “Ultrasound cardiac modeling, segmentation and tracking”.
PhD thesis. Norwegian University of Science and Technology, 2013.
[29] Dikici, E. and Orderud, F. “Graph-cut based edge detection for kalman filter based left ventricle tracking in 3D+T echocardiography”. In:2010 Computing in Cardiology. 2010, pp. 205–208.
[30] Dindoyal, I. et al. “Level Set Segmentation of the Fetal Heart.” In: Jan.
2005, pp. 123–132.
[31] Doherty, J. U. et al. “ACC/AATS/AHA/ASE/ASNC/HRS/SCAI/SCC-T/SCMR/STS 2019 Appropriate Use Criteria for Multimodality Imaging in the Assessment of Cardiac Structure and Function in Nonvalvular Heart Disease”. In:Journal of the American College of Cardiologyvol. 73, no. 4 (2019), pp. 488–516. eprint:https://www.jacc.org/doi/pdf/10.1016/j.
jacc.2018.10.038.
[32] Doo, D. W. H. “A subdivision algorithm for smoothing down irregular shaped polyhedrons”. In: Proceedings on Interactive Techniques in Computer Aided Design (1978), pp. 157–165.
[33] Dyverfeldt, P. et al. “4D flow cardiovascular magnetic resonance consensus statement”. In: Journal of cardiovascular magnetic resonance : official journal of the Society for Cardiovascular Magnetic Resonance vol. 17 (Aug. 2015), p. 72.
[34] Faletra, F. F., Narula, J., and Ho, S. Y. “Atlas of Non-Invasive Imaging in Cardiac Anatomy”. In: Springer, 2020, pp. 91–94.
[35] Fang, W. et al. “Incorporating temporal information into active contour method for detecting heart wall boundary from echocardiographic image sequence”. In:Computerized Medical Imaging and Graphics vol. 32, no. 7 (2008), pp. 590–600.
[36] Flachskampf and al., et. “Recommendations for transoesophageal echocar-diography: update 2010”. In: European Journal of Echocardiography vol. 11, no. 7 (Aug. 2010), pp. 557–576.
[37] Frangi, A., Niessen, W., and Viergever, M. “Three-dimensional modeling for functional analysis of cardiac images, a review”. In:IEEE Transactions on Medical Imaging vol. 20, no. 1 (2001), pp. 2–5.
[38] Gao, Y. et al. “SU-E-U-03: Skin Segmentation Algorithm for Breast Ultrasound Images Using Active Contour Method: Toward Development of Ultrasound-Based Automated-Tissue-Toxicity-Assessment (ATTA) Tool in Breast-Cancer Radiotherapy”. In:Medical Physics vol. 38, no. 6Part25 (2011), pp. 3698–3698.
[39] Genovese, D. et al. “Machine Learning–Based Three-Dimensional Echocar-diographic Quantification of Right Ventricular Size and Function: Valida-tion Against Cardiac Magnetic Resonance”. In: Journal of the American Society of Echocardiography vol. 32, no. 8 (2019), pp. 969–977.
[40] Ghelich Oghli, M. et al. “Left Ventricle Segmentation Using a Combination of Region Growing and Graph Based Method”. In:Iranian Journal of Radiology vol. 14, no. 2 (2017), e42272.
[41] Ghesu, F. C. et al. “Marginal Space Deep Learning: Efficient Architecture for Volumetric Image Parsing”. In:IEEE Transactions on Medical Imaging vol. 35 (2016), pp. 1217–1228.
[42] Goodfellow, I. J. et al.Generative Adversarial Networks. 2014.
[43] Grewal, M. S. and Andrews, A. P. “General Information”. In: Kalman Filtering: Theory and Practice Using MATLAB. 2008, pp. 1–29.
[44] Groote, P. de et al. “Right ventricular ejection fraction is an independent predictor of survival in patients with moderate heart failure”. In:Journal of the American College of Cardiologyvol. 32 (4 1998), pp. 948–954.
[45] Habijan, M. et al. “Overview of the Whole Heart and Heart Chamber Segmentation Methods”. In:Cardiovascular Engineering and Technology vol. 11 (Nov. 2020).
[46] Hadda, F. et al. “Right Ventricular Function in Cardiovascular Disease, Part II Pathophysiology, Clinical Importance, and Management of Right Ventricular Failure”. In: Division of Cardiovascular Medicine (F.H., S.A.H.), Medicine (R.D.), and Pediatrics (D.J.M.) vol. 117 (2008), pp. 1717–1731.
[47] Hahn, R. T. and al., et. “Guidelines for Performing a Comprehensive Transesophageal Echocardiographic Examination: Recommendations from the American Society of Echocardiography and the Society of Cardiovascular Anesthesiologists”. In:Journal of the American Society of Echocardiography vol. 26 (2013), pp. 921–964.
[48] Hautvast, G. et al. “Automatic Contour Propagation in Cine Cardiac Magnetic Resonance Images”. In:IEEE Transactions on Medical Imaging vol. 25, no. 11 (2006), pp. 1472–1482.
Bibliography [49] Hillier, D. et al. “Online 3-D Reconstruction of the Right Atrium From Echocardiography Data via a Topographic Cellular Contour Extraction Algorithm”. In:IEEE Transactions on Biomedical Engineering vol. 57, no. 2 (2010), pp. 384–396.
[50] Hochreiter, S. and Schmidhuber, J. “Long Short-term Memory”. In:Neural computation vol. 9 (Dec. 1997), pp. 1735–80.
[51] Hou, Z.-h. et al. “Prognostic Value of Coronary CT Angiography and Calcium Score for Major Adverse Cardiac Events in Outpatients”. In:
JACC: Cardiovascular Imaging vol. 5, no. 10 (2012), pp. 990–999.
[52] Humpherys, J., Redd, P., and West, J. “A Fresh Look at the Kalman Filter”. In: SIAM Rev.vol. 54, no. 4 (Jan. 2012), pp. 801–823.
[53] Iglesias, J. and Sabuncu, M. “Multi-Atlas Segmentation of Biomedical Images: A Survey”. In:Medical image analysis vol. 24 (Dec. 2014).
[54] Jafari, A., Pszczolkowski, E., and Krishnamurthy, A. “A framework for biomechanics simulations using four-chamber cardiac models”. In:Journal of Biomechanicsvol. 91 (2019), pp. 92–101.
[55] Jain, R. et al. “Diagnostic accuracy of bicuspid aortic valve by echocardiography”. In: ECHOCARDIOGRAPHY-A JOURNAL OF CARDIOVASCULAR ULTRASOUND AND ALLIED TECHNIQUES
vol. 35, no. 12 (Dec. 2018), pp. 1932–1938.
[56] Julier, S. and Uhlmann, J. “Unscented filtering and nonlinear estimation”.
In:Proceedings of the IEEE vol. 92, no. 3 (2004), pp. 401–422.
[57] Kalman, R. “A New Approach to Linear Filtering and Prediction Problems”. In: ASME: Journal of Basic Engineeringvol. 82 (Mar. 1960), pp. 35–45.
[58] Katti, G., Ara, S., and Shireen, D. “Magnetic resonance imaging (MRI) -A review”. In:Intl J Dental Clin vol. 3 (Mar. 2011).
[59] Kim, S. J. W. et al. “Multi-atlas cardiac PET segmentation”. eng. In:
Physica medica vol. 58 (2019), pp. 32–39.
[60] Kingma, D. P. and Welling, M. “An Introduction to Variational Autoencoders”. In:Foundations and Trends® in Machine Learningvol. 12, no. 4 (2019), pp. 307–392.
[61] Knobelsdorff, F. von and Schulz-Menger, J. “Role of cardiovascular magnetic resonance in the guidelines of the European Society of Cardiology”. In:Journal of Cardiovascular Magnetic Resonance vol. 18 (Dec. 2015).
[62] Lancellotti and al., et. “Recommendations for the echocardiographic assessment of native valvular regurgitation: an executive summary from the European Association of Cardiovascular Imaging”. In:European Heart Journal - Cardiovascular Imaging vol. 14, no. 7 (July 2013), pp. 611–644.
[63] Lang, R. M. et al. “Recommendations for Cardiac Chamber Quantification by Echocardiography in Adults: An Update from the American Society of Echocardiography and the European Association of Cardiovascular Imaging”. In:Journal of the American Society of Echocardiographyvol. 28, no. 1 (2015), 1–39.e14.
[64] Leclerc, S. et al. “LU-Net: A Multistage Attention Network to Improve the Robustness of Segmentation of Left Ventricular Structures in 2-D Echocardiography”. In:IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control vol. 67, no. 12 (2020), pp. 2519–2530.
[65] Lee, D. et al. “The growth and evolution of cardiovascular magnetic resonance: A 20-year history of the Society for Cardiovascular Magnetic Resonance (SCMR) annual scientific sessions”. In:Journal of Cardiovas-cular Magnetic Resonance vol. 20 (Dec. 2018).
[66] Leung, K. Y. E. et al. “Sparse Registration for Three-Dimensional Stress Echocardiography”. In:IEEE Transactions on Medical Imagingvol. 27, no. 11 (2008), pp. 1568–1579.
[67] Linderoth, M. et al. “Initialization of the Kalman filter without assumptions on the initial state”. In:2011 IEEE International Conference on Robotics and Automation. 2011, pp. 4992–4997.
[68] Litjens, G. et al. “A survey on deep learning in medical image analysis”.
In:Medical Image Analysisvol. 42 (2017), pp. 60–88.
[69] Loop, C. T. “Smooth Subdivision Surfaces Based on Triangles”. In: 1987.
[70] Mahadi, L. F. et al. “Performance evaluation on mitral valve motion feature tracking using Kanade-Lucas-Tomasi (KLT) algorithm based eigenvalue measurement”. In:AIP Conference Proceedingsvol. 2173, no. 1 (2019), p. 020002. eprint: https://aip.scitation.org/doi/pdf/10.1063/1.
5133917.
[71] Maus, T. M., Nhieu, S., and Herway, S. T. “Transesophageal Echocardio-graphy for Non-cardiac Anesthesiologists”. In: 1st ed. Springer, Cham, 2016. Chap. 2, pp. 15–44.
[72] Mazaheri, S. et al. “Echocardiography Image Segmentation: A Survey”.
In: 2013 International Conference on Advanced Computer Science Applications and Technologies. 2013, pp. 327–332.
[73] Medvedofsky, D. et al. “Three-dimensional echocardiographic quantifica-tion of the left-heart chambers using an automated adaptive analytics algorithm: multicentre validation study”. In: European Heart Journal - Cardiovascular Imaging vol. 19, no. 1 (Feb. 2017), pp. 47–58. eprint:
https://academic.oup.com/ehjcimaging/article-pdf/19/1/47/22951743/
jew328.pdf.
Bibliography [74] Medvedofsky, D. et al. “Three-dimensional echocardiographic
quantifica-tion of the left-heart chambers using an automated adaptive analytics algorithm: multicentre validation study”. In: European Heart Journal - Cardiovascular Imaging vol. 19, no. 1 (Feb. 2017), pp. 47–58. eprint:
https://academic.oup.com/ehjcimaging/article-pdf/19/1/47/22951743/
jew328.pdf.
[75] Mertens, L. and Friedberg, M. “Imaging the right ventricle—Current state of the art”. In:Nature reviews. Cardiology vol. 7 (2010), pp. 551–63.
[76] Montagnat, J., Delingette, H., and Ayache, N. “A review of deformable surfaces: topology, geometry and deformation”. In: Image and Vision Computing vol. 19, no. 14 (2001), pp. 1023–1040.
[77] Mukhopadhyay, P. and Chaudhuri, B. B. “A survey of Hough Transform”.
In:PATTERN RECOGNITION vol. 48, no. 3 (Mar. 2015), pp. 993–1010.
[78] Naqvi, T. Z. “Echocardiography in Percutaneous Valve Therapy”. In:
JACC-CARDIOVASCULAR IMAGING vol. 2, no. 10 (Oct. 2009), pp. 1226–1237.
[79] Narang, A. et al. “Machine learning based automated dynamic quan-tification of left heart chamber volumes”. In: European Heart Journal -Cardiovascular Imaging vol. 20, no. 5 (Oct. 2018), pp. 541–549. eprint:
https://academic.oup.com/ehjcimaging/article-pdf/20/5/541/31623850/
jey137.pdf.
[80] Nillesen, M. M. et al. “Automated Assessment of Right Ventricular Vol-umes and Function Using Three-Dimensional Transesophageal Echocar-diography”. In:Ultrasound in Medicine & Biology vol. 42, no. 2 (2016), pp. 596–606.
[81] Nkomo, V. T. et al. “Burden of valvular heart diseases: a population-based study”. In:The Lancet vol. 368, no. 9540 (2006), pp. 1005–1011.
[82] Noble, J. A. and Boukerroui, D. “Ultrasound Image Segmentation: A Survey”. In:IEEE TRANSACTIONS ON MEDICAL IMAGING vol. 25 (8 2006), pp. 987–1010.
[83] Noble, J. and Boukerroui, D. “Ultrasound image segmentation: A survey”.
In:IEEE transactions on medical imaging vol. 25 (Sept. 2006), pp. 987–
1010.
[84] Nurmaini, S. et al. “Deep Learning-Based Computer-Aided Fetal Echocardiography: Application to Heart Standard View Segmentation for Congenital Heart Defects Detection”. In:Sensors vol. 21, no. 23 (2021).
[85] Oktay, O. et al. “Anatomically Constrained Neural Networks (ACNNs):
Application to Cardiac Image Enhancement and Segmentation”. In:IEEE Transactions on Medical Imaging vol. 37, no. 2 (2018), pp. 384–395.
[86] Orderud, F. “Real-time segmentation of 3D echocardiograms using a state estimation approach with deformable models”. PhD thesis. Norwegian University of Science and Technology, 2010.
[87] Orderud, F. and Inge Rabben, S. “Real-time 3D segmentation of the left ventricle using deformable subdivision surfaces”. In:Proc IEEE Computer Vision Pattern Recognition(2008), pp. 1–8.
[88] Orderud, F., Torp, H., and Rabben, S. “Automatic Alignment of Standard Views in 3D Echocardiograms Using Real-time Tracking”. In:Progress in Biomedical Optics and Imaging - Proceedings of SPIE vol. 7265 (Feb.
2009).
[89] Ostenfeld, E. et al. “Manual correction of semi-automatic threedimen-sional echocardiography is needed for right ventricular assessment in adults; validation with cardiac magnetic resonance”. In:Cardiovascular Ultrasound vol. 10 (2012).
[90] Pace, D. F. et al. “Interactive Whole-Heart Segmentation in Congenital Heart Disease”. In: Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015. Ed. by Navab, N. et al. Cham: Springer International Publishing, 2015, pp. 80–88.
[91] Peterson, P. et al. “Cardiac Imaging Modalities and Appropriate Use”.
In:Primary Care: Clinics in Office Practicevol. 45 (Dec. 2017).
[92] Punithakumar, K. et al. “Regional heart motion abnormality detection:
An information theoretic approach”. In:Medical Image Analysisvol. 17, no. 3 (2013), pp. 311–324.
[93] Qin, X. et al. “Automatic Segmentation of Right Ventricle on Ultrasound Images Using Sparse Matrix Transform and Level Set”. In: Proceedings of SPIE vol. 8669 (Mar. 2013).
[94] Queirós, S. et al. “Aortic Valve Tract Segmentation From 3D-TEE Using Shape-Based B-Spline Explicit Active Surfaces”. In:IEEE Transactions on Medical Imaging vol. 35, no. 9 (2016), pp. 2015–2025.
[95] Rizzo, G. et al. “Satisfactory Rate of Postprocessing Visualization of Stan-dard Fetal Cardiac Views From 4-Dimensional Cardiac Volumes Acquired During Routine Ultrasound Practice by Experienced Sonographers in Peripheral Centers”. In:Journal of Ultrasound in Medicine vol. 30, no. 1 (2011), pp. 93–99. eprint:https://onlinelibrary.wiley.com/doi/pdf/10.7863/
jum.2011.30.1.93.
[96] Ronneberger, O., Fischer, P., and Brox, T. “U-Net: Convolutional Networks for Biomedical Image Segmentation”. In: Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015. Ed. by Navab, N. et al. Cham: Springer International Publishing, 2015, pp. 234–
241.
[97] Sachpekidis, C. et al. “Equilibrium radionuclide ventriculography: Still a clinically useful method for the assessment of cardiac function?” In:
Hellenic Journal of Nuclear Medicine vol. 21 (Sept. 2018), pp. 213–220.
[98] Schneider, R. J. et al. “Mitral Annulus Segmentation From 3D Ultrasound Using Graph Cuts”. In:IEEE Transactions on Medical Imaging vol. 29, no. 9 (2010), pp. 1676–1687.
Bibliography [99] Schroder, J. N. and al., et. “Impact of mitral valve regurgitation evaluated by intraoperative transesophageal echocardiography on long-term outcomes after coronary artery bypass grafting”. In: Circulation vol. 112 (2005).
[100] Sheehan, F. and Redington, A. “The right ventricle: anatomy, physiology and clinical imaging”. In:Heart vol. 94, no. 11 (2008), pp. 1510–1515.
[101] Shibayama, K. et al. “Evaluation of automated measurement of left ventricular volume by novel real-time 3-dimensional echocardiographic system: Validation with cardiac magnetic resonance imaging and 2-dimensional echocardiography”. In:Journal of Cardiology vol. 61, no. 4 (2013), pp. 281–288.
[102] Snare, S. R. and al., et. “Automatic real-time view detection”. In:2009 IEEE International Ultrasonics Symposium. 2009, pp. 2304–2307.
[103] Snare, S. R.Quantitative Cardiac Analysis Algorithms for Pocket-sized Ultrasound Devices. 2011.
[104] Szabo, T. L. “Diagnostic Ultrasound Imaging: Inside Out Second Edition”.
In: Academic Press, 2013, pp. 23–35.
[105] Tavakoli, V. and Amini, A. A. “A survey of shaped-based registration and segmentation techniques for cardiac images”. In: Computer Vision and Image Understanding vol. 117, no. 9 (2013), pp. 966–989.
[106] Tegnander, E. and Eik-Nes, S. H. “The examiner’s ultrasound experience has a significant impact on the detection rate of congenital heart defects at the second-trimester fetal examination”. In:Ultrasound in Obstetrics
& Gynecology vol. 28, no. 1 (2006), pp. 8–14. eprint: https : / / obgyn . onlinelibrary.wiley.com/doi/pdf/10.1002/uog.2804.
[107] Thorne, S. and Bowater, S. “Epidemiology of ACHD”. In:Adult Congenital Heart Disease. Oxford University Press, Sept. 2017, pp. 3–4. eprint:
https : / / academic . oup. com / book / 0 / chapter / 250021994 / chapter ag -pdf/44570778/book\_29665\_section\_250021994.ag.pdf.
[108] Townsend, N. et al. “Cardiovascular disease in Europe: epidemiological update 2016”. In:European Heart Journal vol. 37, no. 42 (Aug. 2016), pp. 3232–3245.
[109] “Transthoracic 3D Echocardiographic Left Heart Chamber Quantifica-tion Using an Automated Adaptive Analytics Algorithm”. In: JACC:
Cardiovascular Imaging vol. 9, no. 7 (2016), pp. 769–782.
[110] Vlachkova, K. and Terziev, P. “A Comparison of Surface Subdivision Algorithms for Polygonal Meshes”. In: AIP Conference Proceedings vol. 1487 (Oct. 2012), pp. 343–350.
[111] Voelkel, N. F. et al. “Right Ventricular Function and Failure”. In:
Circulation vol. 114 (17 2006), pp. 1883–1891.
[112] Wan, E. and Van Der Merwe, R. “The unscented Kalman filter for nonlinear estimation”. In:Proceedings of the IEEE 2000 Adaptive Systems for Signal Processing, Communications, and Control Symposium (Cat.
No.00EX373). 2000, pp. 153–158.
[113] Wang, Y. et al. “Atrial Volume in a Normal Adult Population by Two-dimensional Echocardiography”. In:Chest vol. 86, no. 4 (1984), pp. 595–
601.
[114] Xiaoguang Lu et al. “AutoMPR: Automatic detection of standard planes in 3D echocardiography”. In:2008 5th IEEE International Symposium on Biomedical Imaging: From Nano to Macro. 2008, pp. 1279–1282.
[115] Yuan, B. et al. “Automatic valve segmentation in cardiac ultrasound time series data”. In: Mar. 2018, p. 69.
[116] Zhen, X. et al. “Direct and simultaneous estimation of cardiac four chamber volumes by multioutput sparse regression”. In:Medical Image Analysis vol. 36 (2017), pp. 184–196.
[117] Zheng, Y. et al. “Four-Chamber Heart Modeling and Automatic Segmentation for 3-D Cardiac CT Volumes Using Marginal Space Learning and Steerable Features”. In: IEEE Transactions on Medical Imaging vol. 27, no. 11 (2008), pp. 1668–1681.
[118] Zhuang, X. et al. “Registration-based propagation for whole heart segmentation from compounded 3D echocardiography”. In:2010 IEEE International Symposium on Biomedical Imaging: From Nano to Macro.
2010, pp. 1093–1096.
[119] Østvik, A. et al. “Real-Time Standard View Classification in Transthoracic Echocardiography Using Convolutional Neural Networks”. In:Ultrasound in Medicine and Biology vol. 45, no. 2 (2019), pp. 374–384.
Papers
Paper I
Two Methods for Modified Doo-Sabin Modeling of
Non-Smooth Surfaces - Applied to Right Ventricle Modelling
Håkon Strand Bølviken, Jørn Bersvendsen, Fredrik Orderud, Sten Roar Stange, Pål Brekke, Eigil Samset
Published inJournal of Medical Imaging, December 2020, volume 7, issue 6, pp. 1–17. DOI: 10.1117/1.JMI.7.6.067001.
I
Abstract
Purpose: In recent years there has been increased clinical interest in the right ventricle(RV) of the heart. RV dysfunction is an important prognostic marker for several cardiac diseases. Accurate modeling of the RV shape is important for estimating performance. We have created computationally effective models that allow for accurate estimation of the RV shape.
Approach: Previous approaches to cardiac shape modeling, including modeling the RV geometry, has used Doo-Sabin surfaces. Doo-Sabin surfaces allow for effective computation and adapt to smooth, organic surfaces. However, they struggle with modeling sharp corners or ridges without many control nodes. This paper modifies the Doo-Sabin surface to allow for sharpness using weighting of vertices and edges instead. This was done in two different ways. For validation, we compared the standard Doo-Sabin vs. the sharp Doo-Sabin models in modeling the RV shape of 16 cardiac ultrasound images, against a ground truth manually drawn by a cardiologist. A Kalman filter fitted the models to the ultrasound images, and the difference between the volume of the model and the ground truth was measured.
Results: The two modified Doo-Sabin models both outperformed the standard Doo-Sabin model in modeling the RV. On average the regular Doo-Sabin had 8 ml error in volume, while the sharp models had 7 and 6 ml error respectively.
Conclusions: Compared to the standard Sabin, the modified Doo-Sabin models can adapt to a larger variety of surfaces while still being compact models. They were more accurate on modeling the RV shape, and could have uses elsewhere.