Ekokardiyografi Görüntülerinde Aort Kapak Kalsifikasyon Segmentasyonu için Veri Artırma Yöntemlerinin İrdelenmesi

Yıl 2024, , 1640 - 1653, 15.09.2024

Elif Baykal Kablan

https://doi.org/10.31466/kfbd.1534186

Öz

Aort kapak kalsiyum skoru, aort darlığı tanısında, tedavisinde, takibinde ve koroner arter hastalığı riskinin belirlenmesinde oldukça önemlidir. Güncel kılavuzlar, düşük akım ve düşük gradyanlı aort darlığı tanısında aort kapak kalsiyum skorlarının dikkate alınmasını önermektedir. Aort kapak kalsiyumunun ölçümünde altın standart yöntem bilgisayarlı tomografidir (BT). Agatston skoru, kalsiyum alanı ile BT dansitesinin çarpılmasıyla hesaplanan yarı otomatik bir yöntem olmakla birlikte BT pahalı ve radyasyon riski taşımaktadır. Alternatif olarak, ekokardiyografi, daha ucuz ve radyasyon içermeyen bir yöntem olup bu görüntüleme üzerinde yapılan çalışmalar gözleme dayalı ve yarı kantitatif olup, objektif sonuçlar vermekte zorlanmaktadır. Bu çalışmada, aort kapak kalsifikasyon ölçümü için gerekli olan kalsifikasyon bölgelerini belirlemek üzere derin öğrenme tabanlı U-Net modeli çeşitli veri artırma teknikleri ile değerlendirilmiştir. Bu amaçla yeni bir veri seti oluşturulmuş ve renk, rijid ve rijid olmayan geometrik dönüşümler gibi farklı artırma yöntemlerinin etkinliği analiz edilmiştir. Elde edilen sonuçlar değerlendirildiğinde, rijid olmayan geometrik dönüşümlerin segmentasyon performansını en anlamlı şekilde artırdığı gözlemlenmiştir.

Anahtar Kelimeler

Aort kapak, Kalsifikasyon, Kalsiyum skorlama, Ekokardiyografi, Veri artırma, Derin öğrenme, U-Net

Etik Beyan

KTÜ Farabi Hastanesi'nden (2022/190 nolu) etik kurul onayı alınmıştır.

Destekleyen Kurum

TÜBİTAK

Proje Numarası

222S110

Analysis of Data Augmentation Methods for Aortic Valve Calcification Segmentation in Echocardiographic Images

Öz

Aortic valve calcium scoring is crucial for diagnosing, treating, monitoring aortic stenosis, and assessing coronary artery disease risk. Current guidelines recommend incorporating aortic valve calcium scores in the diagnosis of low-flow and low-gradient aortic stenosis. The gold standard for measuring aortic valve calcium is computed tomography (CT). The Agatston score is a semi-automatic method for calculating calcium scores by multiplying the calcium area by CT density. However, CT is expensive and carries radiation risks. As an alternative, echocardiography, which is cheaper and radiation-free, has been explored. However, studies on echocardiography are observational and semi-quantitative, and they struggle to provide objective results. In this study, the deep learning-based U-Net model was evaluated for identifying calcification regions necessary for aortic valve calcification measurement using various data augmentation techniques. A new dataset was created for this purpose, and the effectiveness of different augmentation methods, including color adjustments, rigid transformations, and non-rigid geometric transformations, was analyzed. The results indicate that non-rigid geometric transformations significantly enhance segmentation performance.

Anahtar Kelimeler

Aortic valve, Calcification, Calcium scoring, Echocardiography, Data augmentation, Deep learning, U-Net

Proje Numarası

222S110

Kaynakça

• Agatston, A. S., Janowitz, W. R., Hildner, F. J., Zusmer, N. R., Viamonte, M., & Detrano, R. 1990. “Quantification of coronary artery calcium using ultrafast computed tomography”, Journal of the American College of Cardiology, 15(4), 827-832.
• Amer, A., Ye, X., & Janan, F. (2021). ResDUnet: A deep learning-based left ventricle segmentation method for echocardiography. IEEE Access, 9, 159755-159763.
• Azad, R., Aghdam, E. K., Rauland, A., Jia, Y., Avval, A. H., Bozorgpour, A., ... & Merhof, D. (2024). Medical image segmentation review: The success of u-net. IEEE Transactions on Pattern Analysis and Machine Intelligence.
• Buslaev, A., Iglovikov, V. I., Khvedchenya, E., Parinov, A., Druzhinin, M., & Kalinin, A. A. (2020). Albumentations: fast and flexible image augmentations. Information, 11(2), 125.
• Cakir, M., Ekinci, M., Kablan, E. B., & Sahin, M. (2024, July). Automated Aortic Valve Calcific Area Segmentation in Echocardiography Images Using Fully Convolutional Neural Networks. In 2024 47th International Conference on Telecommunications and Signal Processing (TSP) (pp. 96-100). IEEE.
• Cakir, M., Ekinci, M., Kablan, E. B., & Şahin, M. (2024). AVD-YOLOv5: a new lightweight network architecture for high-speed aortic valve detection from a new and large echocardiography dataset. Medical & Biological Engineering & Computing, 1-18.
• Chang S, Kim H, Suh YJ, Choi DM, Kim H, Kim DK, Kim JY, Yoo JY, Choi BW. 2021. “Development of a deep learning-based algorithm for the automatic detection and quantification of aortic valve calcium”, Eur J Radiol. 137:109582.
• Clavel, M. A. 2017. “Cardiac imaging for assessing low-gradient severe aortic stenosis”, JACC Cardiovascular Imaging, 10(2), 185-202.
• Elvas, L. B., Almeida, A. G., Rosario, L., Dias, M. S., & Ferreira, J. C. (2021). Calcium identification and scoring based on echocardiography. An exploratory study on aortic valve stenosis. Journal of Personalized Medicine, 11(7), 598.
• ESC, A. K. D., & Kardiyotorasik, A. 2017. “ESC/EACTS Kalp Kapak Hastalıkları Tedavi Kılavuzu”.
• Falk, V., Baumgartner, H., Bax, J. J., De Bonis, M., Hamm, C., Holm, P. J., ... & Zamorano, J. L. 2017. “2017 ESC/EACTS Guidelines for the management of valvular heart disease”, European Journal of Cardio-Thoracic Surgery, 52(4), 616-664.
• Gaibazzi, N., Baldari, C., Faggiano, P., Albertini, L., Faden, G., Pigazzani, F., ... & Reverberi, C. (2014). Cardiac calcium score on 2D echo: correlations with cardiac and coronary calcium at multi-detector computed tomography. Cardiovascular ultrasound, 12, 1-9.
• Gaibazzi, N., Porter, T. R., Agricola, E., Cioffi, G., Mazzone, C., Lorenzoni, V., ... & Faggiano, P. (2015). Prognostic value of echocardiographic calcium score in patients with a clinical indication for stress echocardiography. JACC: Cardiovascular Imaging, 8(4), 389-396.
• Gao, X., Li, W., Loomes, M., & Wang, L. (2017). A fused deep learning architecture for viewpoint classification of echocardiography. Information Fusion, 36, 103-113.
• Hardas, S., Titar, P., Zanwar, I., & Phalgune, D. S. (2021). Correlation between echocardiographic calcium score and coronary artery lesion severity on invasive coronary angiography. Indian Heart Journal, 73(3), 307-312.
• Koos, R., Mahnken, A. H., Sinha, A. M., Wildberger, J. E., Hoffmann, R., & Kühl, H. P. 2004. “Aortic valve calcification as a marker for aortic stenosis severity: assessment on 16-MDCT”, American Journal of Roentgenology, 183(6), 1813-1818.
• Lessmann, N., van Ginneken, B., Zreik, M., de Jong, P. A., de Vos, B. D., Viergever, M. A., & Išgum, I. 2017. “Automatic calcium scoring in low-dose chest CT using deep neural networks with dilated convolutions”, IEEE transactions on medical imaging, 37(2), 615-625.
• Lung, B. 2011. “Epidemiology of valvular heart disease in the adult”, Nature Reviews Cardiology, 8(3), 162-172.
• Mortada, M. J., Tomassini, S., Anbar, H., Morettini, M., Burattini, L., & Sbrollini, A. (2023). Segmentation of anatomical structures of the left heart from echocardiographic images using Deep Learning. Diagnostics, 13(10), 1683.
• Osnabrugge, R. L. 2013. “Aortic stenosis in the elderly: disease prevalence and number of candidates for transcatheter aortic valve replacement: a meta-analysis and modeling study”, Journal of the American College of Cardiology, 62(11), 1002-1012.
• Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., ... & Chintala, S. (2019). Pytorch: An imperative style, high-performance deep learning library. Advances in neural information processing systems, 32.
• Pawade, T., Clavel, M. A., Tribouilloy, C., Dreyfus, J., Mathieu, T., Tastet, L., ... & Dweck, M. R. 2018. “Computed tomography aortic valve calcium scoring in patients with aortic stenosis”, Circulation: Cardiovascular Imaging, 11(3), e007146.
• Saha, S. A., Beatty, A. L., Mishra, R. K., Whooley, M. A., & Schiller, N. B. (2015). Usefulness of an echocardiographic composite cardiac calcium score to predict death in patients with stable coronary artery disease (from the Heart and Soul Study). The American journal of cardiology, 116(1), 50-58.
• Singh, G., Al’Aref, S. J., Lee, B. C., Lee, J. K., Tan, S. Y., Lin, F. Y., ... & Credence And Iconic Investigators. 2021. “End-to-end, pixel-wise vessel-specific coronary and aortic calcium detection and scoring using deep learning”, Diagnostics, 11(2), 215.
• Tang, L., Wang, X., Yang, J., Wang, Y., Qu, M., & Li, H. (2024). DLFFNet: A new dynamical local feature fusion network for automatic aortic valve calcification recognition using echocardiography. Computer Methods and Programs in Biomedicine, 243, 107882.
• van Assen, M., Martin, S. S., Varga-Szemes, A., Rapaka, S., Cimen, S., Sharma, P., ... & Schoepf, U. J. 2021. “Automatic coronary calcium scoring in chest CT using a deep neural network in direct comparison with non-contrast cardiac CT: A validation study”, European Journal of Radiology, 134, 109428.
• Wegner, F. K., Benesch Vidal, M. L., Niehues, P., Willy, K., Radke, R. M., Garthe, P. D., ... & Orwat, S. (2022). Accuracy of deep learning echocardiographic view classification in patients with congenital or structural heart disease: importance of specific datasets. Journal of Clinical Medicine, 11(3), 690.
• Wolterink, J. M., Dinkla, A. M., Savenije, M. H., Seevinck, P. R., van den Berg, C. A., & Išgum, I. (2017, September). Deep MR to CT synthesis using unpaired data. In International workshop on simulation and synthesis in medical imaging (pp. 14-23). Springer, Cham.