Spring 2023 Newsletter

Spotlight

Spotlight

New advances in auto segmentation of clinical imaging to improve skull base surgical procedures.

The generation of segmented anatomic models from a clinical CT scan through the application of artificial intelligence algorithms.

Nikolas Blevins, MD, and his research team are developing innovative methods using deep learning technology to identify and label anatomic structures from clinical imaging studies.  This system incorporates state-of-the-art artificial intelligence algorithms to identify the highly variable and complex vital structures comprising the base of the skull.  Such an approach enables the generation of 3-dimensional virtual models of the geometry that is vital to making interventions safe, effective, and minimally invasive.  Both CT and MRI datasets can be used in this process, providing insights into soft-tissue and osseous structures.  The speed and consistency of the system far exceed that achievable using traditional manual segmentation methods, facilitating the integration of resulting patient-specific virtual anatomy into a broad spectrum of clinical applications.

The models derived from the autosegmentation system are being incorporated into developing surgical simulation platforms, allowing surgeons to rehearse complex procedures in an immersive virtual environment prior to undertaking the actual procedure in the operating room.  Similarly, geometric analysis of anatomic models can yield valuable information for improving cochlear implant selection and placement.  Derived models are also being integrated into an augmented reality surgical workstation to effectively guide surgeons during anatomically challenging approaches.  Similarly, the newly available geometric data from vestibular schwannomas is being used to better predict their clinical impact and optimize treatmentselection and timing.  The autosegmentation system and its applications will be presented at the Combined Otolaryngologic Spring Meeting in Boston. 

The use of an automatically generated cochlear model to predict the optimal path and distance that a cochlear implant array will take within a specific patient’s inner ear.

Virtual 3-dimensional models of the inner ear and tumor-derived superimposed on an MRI from a patient with a vestibular schwannoma.  Such automatically derived anatomic models will help develop optimal treatment protocols for affected patients.