AUTHOR=Wang Jichuan , Zhao Zhiqing , Liang Haijie , Zhang Ranxin , Liu Xingyu , Zhang Jing , Singh Swapnil , Guo Wei , Yan Taiqiang , Hoang Bang H. , Geller David S. , Tang Xiaodong , Yang Rui TITLE=Artificial intelligence assisted preoperative planning and 3D-printing guiding frame for percutaneous screw reconstruction in periacetabular metastatic cancer patients JOURNAL=Frontiers in Bioengineering and Biotechnology VOLUME=12 YEAR=2024 URL=https://www.frontiersin.org/journals/bioengineering-and-biotechnology/articles/10.3389/fbioe.2024.1404937 DOI=10.3389/fbioe.2024.1404937 ISSN=2296-4185 ABSTRACT=Background

The percutaneous screw reconstruction technique, known as the “Tripod Technique,” has demonstrated favorable clinical outcomes in the management of metastatic periacetabular lesions, as evidenced by our prior investigations and corroborated by independent studies. Nevertheless, there is a steep learning curve in handling this technique, with possible complications such as intraarticular screw placement.

Methods

Preoperative pelvic CT scans were acquired before surgery and utilized for the guiding frame design. A convolutional neural network model was trained with annotated data to identify the starting point and trajectory of each potential screw. A model boundary intersection detection technology was used to determine the optimal diameter and length of each screw. A non-rigid registration technology was matched with a prefabricated model of the body surface to design personalized anchoring skin pads. Finally, a polylactic acid-based guiding frame for intraoperative was custom-made with a 3D printer.

Results

12 patients underwent a guiding frame-assisted Tripod procedure for treatment of periacetabular metastatic lesions. An intraoperative CT scan was performed in all cases to confirm screw trajectories. Among 36 screws that were implanted, 26 screws were implanted as designed. The remaining ten screws drifted, but all remained within the intra-osseous conduit without any complications. The mean surgical time was 1.22 h with the guiding frame compared with 2.3 h without the guiding frame. Following the surgical procedure, a noteworthy enhancement in pain management, as evidenced by a reduction in scores on the visual analog scale (p < 0.01), and an improvement in functional status, as assessed through the Eastern Cooperative Oncology Group score (p < 0.01), were observed when compared to the patient’s pre-operative condition.

Conclusion

This proof-of-concept investigation demonstrates that the amalgamation of AI-assisted surgical planning and additive manufacturing can improve surgical accuracy and shorten surgical duration. While access to this technology is currently constrained during its early stages of development, it is anticipated that these limitations will diminish as the potential of AI and additive manufacturing in facilitating complex orthopedic procedures becomes more evident, leading to a surge in interest and adoption of this approach.