ACaVaS – Automated Cardiac Valves Segmentation. TÜBİTAK ARDEB 1002. 2011-2012


The project aims at 3D fully-automatic segmentation and volumetric reconstruction of the aortic and mitral valves from contrast enhanced CT images.


Heart disease is the leading cause of death in the Western world. Mortality is close to 2% in primary valvular disease, where one or more of the 4 cardiac valves (particularly the aortic and mitral) is affected. Adult mitral disease due to history of rheumatic fever is almost eliminated in developed countries,but approximately 20% of the population over 55 years of age still develops mitral valve deficiency; and roughly 2.5% of the world population acquires some form of symptomatic aortic valve diseases regardless of age [1].

Surgical treatment for valvular disease consists of replacement of restoration of the affected valve(s).In either procedure, the insight acquired by the medical expert who will perform the surgical procedure is limited with the information contained in the static 2D/3D medical images acquired from conventional imaging devices. Because the human heart is a dynamic organ, there is an evident need for a patient-specific and dynamic valvular model. Such a model will provide the medical expert the possibility to carry out a detailed pre-operative planning of the surgical procedure.

Thereby, the purpose of the project is to generate a new method to fill this gap. State-of-the-art on segmentation of valvular structures consists of semi-automatic methods, where an expert has to manually mark the images first (initialization) and then the desired structures can be segmented automatically based on the expert’s markings. In this project, the objective is to minimize the need for expert markings by proposing a quasi-automatic segmentation method. For this purpose, 3D semi-automatic segmentation of cardiac CT images corresponding to the closure of the aortic and mitral valves is realized, and their volumetric reconstruction is actualized by point-cloud formation from the segmentation results. The proposed semi-automatic method only requires the slice level corresponding to the valves’ lower bound to be defined by the user. In the project proposal automated segmentation of the valvular structures would be repeated on the images acquired at every 10% timepoint of the cardiac cycle; and the resulting segmentations would be used to create a 4D animation of the aortic and mitral valves. However we encountered a delay in this goal due to reasons unforeseen in the beginning.

Nevertheless, the segmentation method proposed is expected to aid medical experts in making the most out of the imaging data of late-stage valvular disorder cases acquired by non-invasive techniques such as CT and MR, and in the pre-surgical planning stage of such cases.

[1] Lloyd-Jones, D., et al. Heart Disease and Stroke Statistics_2010 Update: A Report From the American Heart Association, Circulation 121, e46-e215, 2010.


  1. Unay, D., Harmankaya, İ., Öksüz, İ., Çubuk, R., Çelik, L., Kadıpaşaoğlu, KA., “Model-free Automatic Segmentation of the Aortic Valve in Multi-slice Computed Tomography Images” (under review).
  2. Harmankaya, I., Oksuz, I., Unay, D., Kadıpasaoglu, K., “Automated Aortic Supravalvular Sinus Detection in Conventional Computed Tomography Image ”, Proc. of IEEE 21st Signal Processing and Communications Applications Conference (SIU), Girne-KKTC, 2013.
  3. Harmankaya, I., Buyuk, K., Unay, D., Oguz, C.T., Acar, U., Kutluk, Y., Polat, N. Erk, S., Bayram, B., Kadıpasaoglu, K., “Four-dimensional modeling of the human heart”, Proc. 5th International Bioengineering Congress (BEC), Izmir-Turkey, 2010.

The Team

  • İbrahim Harmankaya – Bahçeşehir Univ.
  • Devrim Ünay, Asst. Prof. – Bahçeşehir Univ.