Automatic image-based segmentation of the heart from CT scans

Abstract

The segmentation of the heart is usually demanded in the clinical practice for computing functional parameters in patients, such as ejection fraction, cardiac output, peak ejection rate, or filling rate. Because of the time required, the manual delineation is typically limited to the left ventricle at the end-diastolic and end-systolic phases, which is insufficient for computing some of these parameters (e.g., peak ejection rate or filling rate). Common computer-aided (semi-)automated approaches for the segmentation task are computationally demanding, and an initialization step is frequently needed. This work is intended to address the aforementioned problems by providing an image-driven method for the accurate segmentation of the heart from computed tomography scans. The resulting algorithm is fast and fully automatic (even the region of interest is delimited without human intervention). The proposed methodology relies on image processing and analysis techniques (such as multi-thresholding based on statistical local and global parameters, mathematical morphology, and image filtering) and also on prior knowledge about the cardiac structures involved. Segmentation results are validated through the comparison with manually delineated ground truth, both qualitatively (no noticeable errors found after visual inspection) and quantitatively (mass overlapping over 90%).

Semantic Cardiac Segmentation in Chest CT Images Using K-Means Clustering and the Mathematical Morphology Method

“…To overcome this burden, fully automatic methods have been proposed by applying computer-aided technologies [1,7-10]. These techniques have been built based on earlier approaches [8] such as graph-based segmenting [11,12], mean-thresholding [13][14][15][16][17] and fuzzy clustering methods [18][19][20][21]. Later, the deep learning approach showed promising successful performance [7,9,10].…”

“…Inspired by the performance of the mean-thresholding method [13][14][15][16][17], we adapted it for this current study by utilizing the K-Means clustering method as a threshold criterion. The K-Means clustering method [31,32] is a simple unsupervised method, which exploits Euclidean distances to compute the mean of all given pixels and assigns pixels into k different clusters based on the nearest mean.…”

Dynamic updating atlas for heart segmentation with a nonlinear field‐based model

“…To address the above difficulties, a large number of methods for heart segmentation have been recently proposed . Traditional algorithms include thresholding, region growing and fuzzy c‐means clustering . Traditional methods often require many manual interventions.…”

Segmentation and Visualization of Human Coronary Artery Trees from CTA Datasets

“…It provides three-dimensional information [2], allows for a better understanding of cardiac three-dimensional anatomy [4], [5], medical diagnosis [6], and ongoing investigations of acute and chronic coronary heart diseases [7]. The noninvasive [9] manner empowered with the visualization of vessels attracts cardiologists to apply it in clinical environments.…”