Abstract

Background: The Zygomaticomaxillary Complex (ZMC) is essential for facial structure, function, and aesthetics. Due to its prominent anatomical position, the zygoma is highly prone to fractures. Accurate reconstruction of Computed Tomography (CT) images in such cases is crucial for fracture reduction, fragment repositioning, and the design of patient-specific implants. However, reconstructing defective regions with precision is often time-consuming and requires close collaboration between surgeons and medical modeling experts. Objective: This study aimed to develop an automated deep learning-based method for CT image reconstruction of zygomatic fractures to support surgical planning and navigation.
Material and Methods: In this computational experimental study, an automated deep learning approach was implemented for CT image reconstruction of zygomatic fractures, using a modified U-Net architecture with dilated convolution blocks. The method was evaluated quantitatively using several metrics, including the Dice Similarity Coefficient (DSC), Jaccard Index, Precision, Recall, Specificity, Hausdorff Distance, and surface analysis. Results: The proposed model demonstrated high accuracy, achieving DSC values of 0.98 and 0.96 for the border and surface regions of the zygoma, respectively, indicating its strong capability in reconstructing missing regions with high fidelity. Conclusion: This automated method provides a reliable and effective solution for reconstructing zygomatic defects, offering valuable support for preoperative planning and intraoperative navigation in zygoma reduction surgeries.

https://jbpe.sums.ac.ir/article_51848.html