Automatic Liver Segmentation from 2D CT Images Using an Approximate Contour Model

Abstract

Due to its precision, computed tomography (CT) is now generally used to image the liver and diagnose its diseases. Computer-assisted methods aimed at facilitating the extraction of organ shapes from medical images and helping to diagnose disease entities are rapidly developing. This study presents a new method of automatically segmenting the shape of the liver, both for cases free of lesions and those showing certain disease units, i.e. focused lesions like hemangiomas and hepatomas. For the 1,330 2D CT images of the abdominal cavity and liver analysed, the Dices similarity coefficient amounted to 81.3 %. The method proposed is to be used in an IT system supporting liver diagnostics.

Overview of Technical Elements of Liver Segmentation

“…Pictorial representation of liver lesion is not represented here. Marcin Ciecholewski [18] present novel method which automatically segments the liver shape. In CT scan images lumber section of the spine is utilized as seed point.…”

Advanced techniques in medical image segmentation of the liver

“…Heimann, 2007;Kainmüller et al, 2007) use the same sliver07 dataset (10 test studies) than in this work, so the results are directly comparable. The last 5 authors of this Table 3.4 (Wang et al, 2013;Casciaro et al, 2012;Ciecholewski, 2014;Campadelli et al, 2009;Ruskó et al, 2009) use not publicly available datasets and the comparison is not direct. It is observable that methods with better accuracy than ours required a high user interaction (Yang et al, 2014;Peng et al, 2014) or a high computational cost (Ji et al, 2013;Wang et al, 2013;Casciaro et al, 2012;Campadelli et al, 2009;T.…”

“…Liver segmentation algorithms are applied to MRI or CT images. In the literature, these methods are more commonly used in CT (Li et al, 2013;Heimann et al, 2009;Mharib et al, 2012;Punia and Singh, 2013;Ji et al, 2013;Park et al, 2003;Wang et al, 2013;Yang et al, 2014;Casciaro et al, 2012;Peng et al, 2014;Ciecholewski, 2014;Campadelli et al, 2009;T. Heimann, 2007;Kainmüller et al, 2007;Ling et al, 2008;Ruskó et al, 2009;Chen et al, 2009b;Soler et al, 2001) than in MRI (Punia and Singh, 2013;Gloger et al, 2010) for different reasons: CT has a better gradient response than MRI, it has less artefact effects because the movement is less (a CT study requires an acquisition time of 2-3 minutes whereas a MR analysis requires 6-8 minutes) and, consequently, CT involves less cost than MR.…”

“…These characteristics can be inappropriate in a clinical environment. Probabilistic atlases, level-sets, deformable models, statistical shape models are algorithms used in liver segmentation which are based on contour properties (Li et al, 2013;Heimann et al, 2009;Mharib et al, 2012;Punia and Singh, 2013;Ji et al, 2013;Park et al, 2003;Wang et al, 2013;Yang et al, 2014;Casciaro et al, 2012;Peng et al, 2014;Ciecholewski, 2014;Campadelli et al, 2009;T. Heimann, 2007;Kainmüller et al, 2007).…”

“…These algorithms also need training (Wang et al, 2013) or an initial iteration to form the initial curve, for example with more than 30 seeds (Yang et al, 2014) or drawing this initial curve (Peng et al, 2014). Some efforts are focused on providing this initial curve in an automatic way but the accuracy of the algorithm decreases (Ciecholewski, 2014) or computational cost increases more than 10 minutes for a typical dataset of 100-120 slices (Casciaro et al, 2012). In the case of statistical or deformable models, a hard training is also required in a similar way to the probabilistic atlases (T. Heimann, 2007;Kainmüller et al, 2007).…”