A conditional statistical shape model with integrated error estimation of the conditions; Application to liver segmentation in non-contrast CT images

• A novel relaxed conditional SSM with integrated error estimation is created.
• The conditional SSM is flexible and data-driven.
• The model can bridge between non-conditional and conventional conditional SSMs.
• The proposed method outperforms state of the art algorithms in liver segmentation.
• Improved performance is mainly in difficult to segment and oddly shaped cases.

This paper presents a novel conditional statistical shape model in which the condition can be relaxed instead of being treated as a hard constraint. The major contribution of this paper is the integration of an error model that estimates the reliability of the observed conditional features and subsequently relaxes the conditional statistical shape model accordingly. A three-step pipeline consisting of (1) conditional feature extraction from a maximum a posteriori estimation, (2) shape prior estimation through the novel level set based conditional statistical shape model with integrated error model and (3) subsequent graph cuts segmentation based on the estimated shape prior is applied to automatic liver segmentation from non-contrast abdominal CT volumes. Comparison with three other state of the art methods shows the superior performance of the proposed algorithm.

Figure optionsDownload high-quality image (315 K)Download as PowerPoint slide