Applying a Preoperative Planning Method Based on 3D Artificial Intelligence-generated Lumbar Models in Transforaminal Puncture: A Phantoms Study.

Background: Transforaminal puncture is a critical element of lumbar transforaminal epidural steroid injections used to manage lumbar radicular pain. Numerous challenges persist, owing to the intricate 3-dimensional (3D) anatomy of the spine and the delicate nature of the neurovascular structures involved. Consequently, performing the puncture expeditiously, precisely, and safely is imperative. Although numerous scholars have explored methods for reconstructing 3D lumbar models from patient data, the practical application of these models in puncture path planning for transpedicular procedures remains limited. Approaches based on artificial intelligence offer promising advantages for constructing patient-specific 3D models to facilitate puncture pathways planning.

Objective: In this experimental study, we proposed a preoperative planning method utilizing 3D artificial intelligence-generated lumbar models to improve the accuracy and efficiency of the transforaminal puncture process.

Study Design: A phantoms study.

Setting: The Fifth Affiliated Hospital of Sun Yat-sen University, Zhuhai, P.R. China.

Methods: A total of 24 puncture trials utilizing 12 phantom models were independently conducted by 2 surgeons, employing our developed preoperative planning method and conventional fluoroscopy. After one month, one of the surgeons repeated the procedure. Puncture error, characterized by the discrepancy between the preoperative planning puncture target and the actual postoperative needle puncture point (measured in millimeters), as well as puncture procedure duration (measured in minutes), were evaluated by comparing the newly developed preoperative planning method with the traditional fluoroscopy method employed in the transforaminal puncture process.

Results: The average puncture error associated with the preoperative planning method was significantly lower than the conventional fluoroscopy method (3.33 ± 0.73 mm vs 5.25 ± 0.92 mm, P < 0.001). Additionally, the average puncture time of the preoperative planning method was significantly shorter than the conventional fluoroscopy method (7.29 ± 0.95 minutes vs 11.48 ± 1.27 minutes, P < 0.001).

Limitations: Our study used a small number of models; additional clinical trials are required to validate our preoperative planning methods.

Conclusion: The preoperative planning method utilizing 3D artificial intelligence-generated lumbar models for transforaminal puncture demonstrated superior accuracy and efficiency in phantom trials over the traditional fluoroscopic method. This newly developed preoperative planning technique has the potential to significantly improve the accuracy and efficiency of the transforaminal puncture process.