An automatic segmentation algorithm for conglutinated bone fragments in 3D CT images of lower limb fractures based on morphology

The commonly used automatic segmentation algorithms often have over-segmentation and error-segmentation problems when segmenting conglutinated bone fragments in computed tomography (CT) images. And the end face details of the fragments are lost in this process. At present, bone fragments are often manually segmented by doctors. This increases the workload of doctors and takes low efficiency and poor accuracy in bone segmentation. We propose an automatic bone fragment segmentation algorithm based on morphology to segment the conglutinated and separated bone fragments and reconstruct them in 2D and 3D spaces. The pixel classification and clustering method based on morphology are proposed in in the algorithm to identify the pixels in the fracture images. And a concave points detection method is proposed to segment the conglutination area of the fractured bone. This algorithm and two other commonly used algorithms are used in the experiment to segment the three typical fracture images. The segmentation results are quantitatively analyzed to demonstrate the advantages of the morphology algorithm. The morphology algorithm performs well in the metrics of Accuracy (.93 ± .07), Sensitivity (.92 ± .08), Dice coefficient (.92 ± .08) and Mean intersection over union (MIoU) (.92 ± .07). The segmentation results of this algorithm are closest to the manual segmentation results of doctors compared to the other two algorithms. We have demonstrated that the algorithm proposed in this paper has good segmentation accuracy for separated and conglutinated bone fragments. This algorithm significantly improves the segmentation accuracy and efficiency of bone fragments under complex fracture conditions.