A novel approach for the automated segmentation and volume quantification of cardiac fats on computed tomography

Abstract

The deposits of fat on the surroundings of the heart are correlated to several health risk factors such as atherosclerosis, carotid stiffness, coronary artery calcification, atrial fibrillation and many others. These deposits vary unrelated to obesity, which reinforces its direct segmentation for further quantification. However, manual segmentation of these fats has not been widely deployed in clinical practice due to the required human workload and consequential high cost of physicians and technicians. In this work, we propose a unified method for an autonomous segmentation and quantification of two types of cardiac fats. The segmented fats are termed epicardial and mediastinal, and stand apart from each other by the pericardium. Much effort was devoted to achieve minimal user intervention. The proposed methodology mainly comprises registration and classification algorithms to perform the desired segmentation. We compare the performance of several classification algorithms on this task, including neural networks, probabilistic models and decision tree algorithms. Experimental results of the proposed methodology have shown that the mean accuracy regarding both epicardial and mediastinal fats is 98.5% (99.5% if the features are normalized), with a mean true positive rate of 98.0%. In average, the Dice similarity index was equal to 97.6%.

Automatic Epicardial Fat Segmentation and Quantification of CT Scans Using Dual U-Nets With a Morphological Processing Layer

“…In this work, we used the cardiac CT dataset from [22]. To properly evaluate the cardiac fat of the CT scans, all of the raw images in this dataset were transferred to (-200, -30) the HU range to keep the cardiac fat more obvious.…”

“…Rodrigues et al investigated the feature extraction method comprising of classification and registration algorithms for automatic segmentation and quantification of epicardial fat. This method can obtain an ideal performance in the Dice similarity index [22], where the procedures are some what complicated. For the results of cross-validation on another patient, it can be observed that the epicardial fat is oversegmented on the cardiac CT image, which affects the effectiveness of segmentation.…”

Fractal triangular search: a metaheuristic for image content search

“…In this subsection, a practical experiment using computed tomography (CT) images is performed. In a previous work [19], we proposed a method that automatically segments two types of cardiac fats: epicardial and mediastinal.…”

“…The goal is to find an optimal or an appropriate near optimal transformation of one image causing it to resemble the whole or part of the other [1]. In regards to pattern recognition, the goal is to align landmarks between two images [17] or to use similarity measures to find the most suitable position of a template [18, 19]. 3D reconstruction and geometry estimation can also involve the use of search for specific areas and patterns [20].…”

CoreSlicer: a web toolkit for analytic morphomics

“…–TAATotal abdominal fat attenuation: mean attenuation value within the abdominal fat, in HU (ND).–TAA_HUFat(thoracic)− 190 to − 30 HUEpicardial fat area: fat located between the heart and the pericardium, in mm 2 [94–100]. Random forest [101, 102], geodesic active contours [103], fuzzy C-means [104], other [105]EFAEpicardial fat attenuation: mean attenuation value within the epicardial fat, in HU [106]–EFA_HU…”