A fully automatic method for 3D composite material CT image segmentation and reconstruction

Abstract

This paper proposes a method for fully automatic segmentation and reconstruction of 3D woven composite material CT images. This method redesigns the yarn segmentation task for 3D woven composite material images. By combining semantic segmentation techniques with gradient flow-based instance segmentation techniques, it can accurately segment the boundaries of adhered yarns in low-contrast CT images. Ultimately, test results demonstrate that this method can effectively segment yarns of various shapes and sizes, achieving clear yarn segmentation boundaries in segmented samples and significantly improving the reliability of yarn segmentation.

Claims

1. A method for fully automatic 3D composite material CT image segmentation and reconstruction, characterized in that: The steps include: a) Perform preliminary semantic segmentation of yarns in various types of woven materials in CT images, where yarns include warp and weft yarns; b) Based on the semantic segmentation results of YARN, an input dataset for the instance segmentation pre-training model is constructed, and a gradient flow-based segmentation method is used for pre-training on the dataset; c) Based on the semantic segmentation results of woven materials, cases where yarn adhesion exists are selected for manual modification; d) Based on the manual modification results, the instance segmentation pre-trained model is fine-tuned on a dataset of similar types of woven materials. During fine-tuning, the semantic segmentation results are used as the original images, and the manual modification results are used as labels to input into the instance segmentation pre-trained model; e) Design a post-processing module to correct the instance segmentation results; f) reconstructing the divided yarn; In the step (b), the input data set of the instance segmentation pre-training model is constructed by the semantic segmentation result of the yarn, that is, the warp semantic segmentation result is used as the original image and label data at the same time; the gradient flow-based segmentation method includes the Cellpose framework, specifically, the network predicts the final segmentation label by learning the direction field and probability map of each pixel to obtain the flow field of the yarn, the direction field represents the vector of each pixel pointing to the center of the yarn, the probability map represents the probability of each pixel belonging to the yarn, and the flow field is used to describe the shrinkage of the pixel from the yarn boundary to the center; the network architecture includes the U-Net architecture, the input is the image data set constructed based on the semantic segmentation result, the output is the direction field and probability map of each pixel, and the training loss function of the network includes the direction field. loss and cross entropy CE loss for probability maps; The loss is calculated as follows: ； ； in is the total number of voxels, and At voxel points The predicted directional gradient and the true directional gradient on is the CE loss function term, is the overall loss function of the network; The manually modified annotations in step (c) focus on the situation where yarns are stuck together after semantic segmentation.

2. The method for fully automatic 3D composite material CT image segmentation and reconstruction according to claim 1, characterized in that: The semantic segmentation network used in step (a) includes a nnUNet network, and the loss function for training the nnUNet network includes a Dice + cross entropy CE loss function, which is specifically calculated as follows: ； ； ； in is the Dice loss function term, is the CE loss function term, is the overall loss function of the network, is the total number of voxels, and At voxel points The segmentation results and labels on for The SoftMax function output relative to all voxels.

3. The method for fully automatic 3D composite material CT image segmentation and reconstruction according to claim 1, characterized in that: In the step (d), similar types of woven materials refer to materials with similar yarn cross-sectional areas and yarn shapes.

4. The method for fully automatic 3D composite material CT image segmentation and reconstruction according to claim 1, characterized in that: In the step (e), the post-processing module is used to correct the warp yarns that were not correctly identified in the instance segmentation by differentiating the semantic segmentation results of the warp yarns from the instance segmentation results, detecting the unidentified yarns through a threshold method, and adding them to the instance segmentation results, while also supplementing the semantic segmentation results of the weft yarns.

5. The method for fully automatic 3D composite material CT image segmentation and reconstruction according to claim 1, characterized in that: In the step (f), the two-dimensional slice segmentation results of the cross-sections of the warp and weft yarns are reconstructed into three-dimensional visualization.

Images