Naslov
IMAGINE HEART

Autori
Galić, Irena ; Leventić, Hrvoje ; Habijan, Marija ; Romić, Krešimir

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, znanstveni

Izvornik
Thesis of reports SCIENTIFIC-PRACTICAL CONFERENCE OF RUSSIAN AND CROATIAN SCIENTISTS / Korotčenko N. A. - Moskva, 2019, 147-148

ISBN
978-5-907226-34-0

Skup
Scientific - practical conference of Russian and Croatian scientists

Mjesto i datum
Moskva, Ruska Federacija, 04.10.2019. - 05.10.2019

Vrsta sudjelovanja
Plenarno

Vrsta recenzije
Podatak o recenziji nije dostupan

Ključne riječi
Cardiovascular diseases analysis ; Medical image processing

Sažetak
Cardiovascular diseases (CVDs) are one of the most dangerous threats to human health. According to the last available statistics from European Heart Network, CVDs are responsible for 45% of all deaths in Europe, while around 85 million (11.2% of the European population) are still living with CVDs. The evolution of noninvasive cardiac imaging enables the inspection of a human body without the need for surgical interventions. Such imaging, in collaboration with sophisticated computer-aided medical software, can provide multi-dimensional cardiac images through-out a single cardiac cycle. The implementation of such medical software is a complex task, mostly because it involves the processing of a massive amount of highly-dimensional data. To overcome this obstacle, an initial step in the development of such software is the isolation of the cardiac structures from surrounding tissues and organs, i.e., removal of unnecessary information. Usually, this task is performed by the segmentation methods and algorithms, which have tremendously improved throughout the last few decades. Our previous research on this subject involves the development of novel semi-automatic left atrial appendage (LAA) segmentation and analysis methods, used for the planning of the LAA closure procedure. LAA closure is a relatively novel percutaneous procedure for stroke prevention in patients suffering from atrial fibrillation. The method segments LAA with dice coefficient overlap of 92.52% and 91.63% against the ground truth segmentations. After LAA segmentation, the method proposes the location for the occluder placement and calculates measurements needed for closure procedure planning. While in this case the necessary measurements can be determined from the segmentation, in other medical procedures, such as cerebral arteriovenous malformation (AVM) embolization, the segmentation alone is not sufficient. Through the combination of AVM localization and vessel delineation methods, our proposed solution allows the decomposition of the AVM into veins, arteries, and the nidus, thus helping with the planning of the procedure. Moreover, our current research involves multi-labeled whole heart segmentation method that obtained an overall average score for the entire heart segmentation with the dice coefficient overlap of 89% against the ground truth. After receiving satisfying segmentation, the next necessary step for obtaining full cardiac health overview is the calculationof the essential cardiac indicators, as well as the determination of its geometry. The process of determination of these indicators is referred to as cardiac quantification. Therefore, our focus is also aimed at automatic determination of the hearts’ functional parameters such as ejection fraction, chamber volumes, cardiac output, stroke volume, myocardial mass, wall thickness, and thickening. The complete cardiovascular health analysis can be provided by the determination of multiple cardiovascular quantifications of heart chambers, combined with the quantification of essential arteries (aorta, coronary arteries, carotid). Although the methods for individual quantifications already exist, our aim is to improve the diagnostics and patient treatment through the development of advanced methods, which will be able to provide diagnostics and patient treatment with the development of advanced methods that will provide the highest number of heart quantifications from one medical image. The final result will be able to assist medical professionals with CVD diagnosis, as well as provide researchers with options for testing and improving developed methods.

Izvorni jezik
Rus

Znanstvena područja
Računarstvo

POVEZANOST RADA