Introduction: Deep drilling at the lateral skull base with a linear drill holds great potential to reduce invasiveness but poses challenges due to the proximity of the drill channel to critical structures, such as the facial nerve, which can be mechanically or thermally damaged. The development of a drill system capable of real-time temperature monitoring and prediction is crucial to minimize the risk of thermal damage and enhance the safety of procedures, like implantation, taking samples or drug delivery.

Materials: A novel drill system has been developed with an integrated temperature sensor at the drill tip and a microphone on the drill shaft, enabling the collection of process-parallel temperature and sound data. This data is used to train machine learning algorithms for predicting the future temperature development during drilling. Drillings were performed on bone equivalent material stacks mimicking the structure of the lateral skull base.

Results: During calibration, the drill with the integrated temperature sensor exhibits promising behavior, with an uncertainty of 0.02K compared to a calibrated thermometer. Using this drill, drillings (n=193) with different parameters on eight bone equivalent material stacks are performed. Six machine learning algorithms are trained, with the linear gradient boosting machine (LGBM) showing the most promising prediction accuracy (around 1K mean absolute error).

Discussion: The presented drill system and predictive model mark significant progress in ensuring safety during minimal invasive procedures at the lateral skull base.

Die Autorinnen/Autoren geben an, dass kein Interessenkonflikt besteht.