Bone tumor examination based on FCNN-4s and CRF fine segmentation fusion algorithm

• Published in: Journal of bone oncology Vol. 42; p. 100502
• Main Authors: Wu, Shiqiang, Bai, Xiaoming, Cai, Liquan, Wang, Liangming, Zhang, XiaoLu, Ke, Qingfeng, Huang, Jianlong
• Format: Journal Article
• Language: English
• Published: Elsevier GmbH 01.10.2023; Elsevier
• Subjects: Bone tumor; Computed tomography; Conditional random fields; Fully convolutional neural network; Image segmentation; VSI: MI Orthopedics; Image segmentation; Conditional random fields; Computed tomography; Bone tumor; Fully convolutional neural network
• ISSN: 2212-1374; 2212-1366; 2212-1374
• DOI: 10.1016/j.jbo.2023.100502

•Digital image segmentation is crucial for accurate quantitative analysis of medical images, including X-ray images of bone tumors.•Multi-level feature fusion and batch normalization were used to improve segmentation accuracy in image recognition with a convolutional neural network.•FCNN-4s algorithm uses fine feature fusion, BN layer, and data augmentation to improve bone tumor segmentation.•Adopts operations like Crop and Fuse, padding, ReLU activation, and SoftMax loss with optimized hyperparameters for better performance.•Improves bone tumor segmentation with refined structure and probability graph model, achieving higher accuracy and real-time performance. Bone tumor is a kind of harmful orthopedic disease, there are benign and malignant points. Aiming at the problem that the accuracy of the existing machine learning algorithm for bone tumor image segmentation is not high, a bone tumor image segmentation algorithm based on improved full convolutional neural network which consists fully convolutional neural network (FCNN-4s) and conditional random field (CRF). The improved fully convolutional neural network (FCNN-4s) was used to perform coarse segmentation on preprocessed images. Batch normalization layers were added after each convolutional layer to accelerate the convergence speed of network training and improve the accuracy of the trained model. Then, a fully connected conditional random field (CRF) was fused to refine the bone tumor boundary in the coarse segmentation results, achieving the fine segmentation effect. The experimental results show that compared with the traditional convolutional neural network bone tumor image segmentation algorithm, the algorithm has a great improvement in segmentation accuracy and stability, the average Dice can reach 91.56%, the real-time performance is better. Compared with the traditional convolutional neural network segmentation algorithm, the algorithm in this paper has a more refined structure, which can effectively solve the problem of over-segmentation and under-segmentation of bone tumors. The segmentation prediction has better real-time performance, strong stability, and can achieve higher segmentation accuracy.