Researchers at GW have developed a novel solution to identify nerve tissues that can allow an accurate effective performance of a nerve-related surgery. The novel solution is based on a deep-learning network framework that utilizes a U-Net architecture with a Transformer block based fusion module at the bottleneck. The identification of nerve tissues may occur from a multi-modal optical imaging system. The novel solution also affords a heightened situational awareness during surgery thus allowing a respective surgeon to better perform associated duties. The novel solution also affords a respective surgery or the like to be greatly cost-effective owing to the utilization of a deep learning architecture with multi-modal inputs.
The disclosed invention can be implemented as either an apparatus, a device, a system, or a method as can be appreciated. The disclosed invention can include various aspects as follows: (i) an optical imaging nerve identification module that uses Mueller polarimetric imaging that can calculate intrinsic birefringence patterns from fibrous nerve structures; (ii) a U-Net Architecture module to assist in associated computer vision and medical imaging tasks, especially with respect to related semantic segmentation tasks as can be appreciated; (iii) a transformer module associated with the U-Net Architecture module, that can be utilized for fusion for varied modalities as can be appreciated.
Fig. 1 – One example of an aspect of the disclosed invention
Applications:
- Prediction of nerve tissues during/for nerve-related surgery
- Non-invasive intraoperative nerve identification
Advantages:
- Accurate and effective performance of surgery
- Increased situational awareness as to nerves during surgery
- Cost-effective