A paper from Anton Kontunen investigating the feasibility of DMS in intraoperative tissue identification during breast cancer surgery was published in IEEE access!
The prototype DMS-based tissue identification system was used intraoperatively in 20 breast cancer surgeries in the autumn of 2019. Tissues were annotated based on recordings from a head-worn camera system to enable supervised learning. Statistical differences in the DMS spectra were found between tissue types, but the 4-class classification accuracy between skin, fat, glandular tissue and connective tissue was left at 44 %. The low accuracy can be attributed to multiple factors, including high within-class variances and delays in the real-time analysis, along with the uncertainty of the visual annotations.
Since the execution of the study, we have built a purpose-built DMS sensor better suited for real-time measurements. Also, the system didn’t add complications or prolong surgeries in the study. There’s yet work to be done and we’re up to the task!
You can read the open access article on IEEE Xplore.