Optimal segmentation of microcomputed tomographic images of porous tissue-engineering scaffolds.

The morphometric properties of the porous tissue-engineering scaffolds play a dominant role in the initial cell attachment and subsequent tissue regeneration. These properties can be derived nondestructively with the use of quantitative analysis of high-resolution microcomputed tomography (microCT) imaging of scaffolds. Accurate segmentation of these acquired images into solid and porous subspaces is critical to the integrity of morphometric analysis. The absence of a single image-processing technique to provide such accurate separability immune to all the intricacies of the acquired data makes this seemingly simple task significantly error prone. Consequently, an optimal segmentation has to be selected by ranking the segmentations produced by a multiplicity of methods. This article proposes a robust, easy-to-implement, unambiguous, signal-processing-based, ground-truth-free, segmentation rating metric that correlates with visual acuity. With the use of this metric it is possible, for the first time, to threshold the data with a wide range of techniques and select automatically the technique that best delineates the acquired image. The proposed solution has been extensively tested on microCT images of scaffolds fabricated with biodegradable poly (propylene fumarate) (PPF) with the use of a solvent casting particulate leaching process. The approaches proposed and the results obtained may have profound implications for accurate image-based characterization of tissue-engineering scaffolds.