Automatic segmentation of Computed Tomography (CT) images of domestic pig skeleton using a 3D expansion of Dijkstra’s algorithm

• The major bones in CT scans of pigs were automatically segmented and identified.
• Semi-automatic methods might be implemented for reducing errors.
• The method will constitute an important basis for pig breeding and management.

A 3D expansion of Dijkstra’s algorithm used for automatic segmentation and identification of the bones in CT images of live pigs was developed and validated. The major bones in the skeletons of 208 out of 485 live pigs (43%) were segmented and identified from the images without major errors. The segmentation and identification is executed through 8 main operations: (1) identify the full bone structure by a threshold of Hounsfield units, (2) identify forelimbs by voxel connectivity and set landmarks, (3–8) segment out and identify the individual bones in different main parts of the bone structure by the 3D expansion of Dijkstra’s algorithm. The algorithms described will constitute an important basis for further work applying CT in pig breeding and management.

Figure optionsDownload as PowerPoint slide