Text

Objectives and questions: Intramedullary nailing remains a primary surgical method for stabilizing femoral fractures, with distal interlocking playing a pivotal role in ensuring both rotational and axial stability. Different screw insertion techniques for distal interlocking have evolved substantially, ranging from traditional freehand methods, through navigated approaches, to emerging AI-supported procedures.

In a cadaver study, we tested the AI-assisted surgery approach to facilitate the distal locking of long cephalomedullary nails and following we also present first clinical cases.

Material and methods: The AI-Assisted surgery System (AIAS) is, among other things, designed to support distal locking of cephalomedullary nails. It does not need any markers or reference bodies. It is wired to the C-arm and automatically recognizes and processes each newly acquired 2D C-arm image. Based on an AP and a lateral C-arm image and a statistical model of the femur stored in the system, a 3D reconstruction of the distal femur can be computed. Based on a single oblique lateral image, the relative 3D position between a drill bit and a nail implanted in the distal femur can also be computed. This static 3D navigation information can then be used to advise on adjusting the angle of the power tool. When hooked up to an augmented reality device worn by the surgeon (Magic Leap 2, optional), there is also real time feedback by virtually augmented arrows at the power tool indicating if and in which direction the power tool needs to be tilted in order to hit the locking hole. This works before the actual drilling starts as well as during drilling.

In addition and without any additional steps, the system provides the needed length of the interlocking screws.

Results: A total of 42 distal locking screw drillings were performed by five experienced surgeons and with additional AR-Information via HMD in 17/42 procedures.

There was a 100% success rate in the first attempt, e.g. no miss in any of the 42 drillings.

Moreover, the system was successful in predicting the length of the interlocking screws with a median error of <1.5 mm.

Discussion and conclusions: The presented AI-Assisted Surgery System (AIAS) demonstrated a high degree of precision in a cadaver study for distal interlocking of long cephalomedullary nails. Moreover, the screw length was predicted reliably, thus saving a surgical step by avoiding manual length measurement. Initial clinical experiences show high accuracy and provide recommendations for the intraoperative setup.