Complete tumor resection with tumor-free margins (R0 resection) is a key prerequisite for therapeutic success in oncologic surgery, for example in head and neck tumors. In current clinical practice, however, the intraoperative determination of tumor margins is still largely subjective and depends heavily on the experience of the surgical team. Objective, image-based methods for intraoperative tumor margin detection are therefore only available to a limited extent.

The ARBOR project aims to support this process through an automated, marker-free, and data-driven approach. To this end, untreated and unsectioned tumor tissue is analyzed using nonlinear multimodal spectroscopic imaging. The resulting image data provide structure- and material-specific information about the tissue without the need for contrast agents or additional preparation steps.

For objective data evaluation, methods of machine learning and deep learning are employed. The imaging datasets are subsequently correlated with and annotated using histological sections in order to train robust algorithms for the automatic detection of tumor margins. On this basis, tissue regions are to be reliably classified and tumor boundaries precisely identified.

The information obtained in this way is then used to control a femtosecond laser, enabling highly precise and minimally invasive ablation of tumor tissue. In the long term, ARBOR aims to integrate these components into a robot-assisted surgical system. This is intended to enable a more objective and partially automated form of oncologic surgery, combining improved resection quality with maximal preservation of healthy tissue.

The project is funded by Deutsche Krebshilfe within the funding priority program “Surgically oriented operative procedures” (grant numbers 70116061, 70116281).