Towards Intraoperative Tissue Characterisation with Industrial Precision LiDAR

Noan Le Renard, Steven S. Wong, stamatia Giannarou, Daniele Faccio, Timothy G. Constandinou, Jinendra Ekanayake

Keywords: Tissue deformation, Machine Learning, Medical Imaging, Point Cloud Data, Laser Line Probes, Viscoelasticity

Abstract:

There is a recognised need for real-time, point-of-procedure tissue identification during resective tumour surgery; this is made more significant by the need to account for the tissue shifting during tumour surgery in areas such as the liver and brain. This challenge with tissue mobility, deformation and ‘shift’ leads to the preoperative imaging which is currently used to localise tumours, such as MRI or CT scans, being rendered inaccurate or misleading.

In this work, we explored the use of an industrial precision handheld Laser Line Probe (LLP) with 25-micron accuracy to extract tissue viscoelastic information, with the goal of identifying healthy and cancerous tissue in real-time. This is anticipated to contribute to significantly improved surgical outcomes, with scalability to resource limited and technology sparse environments.

Simulation of intraoperative palpation was robustly paired with the LLP scanning and during direct probing of high-fidelity tissue models. We obtained point cloud scans which emulated time-series data, with the scan line characterising tissue deformation in 3D. By extracting physical and 3D point cloud features, we trained a Random Forest model capable of classifying and differentiating biophantom and nonorganic samples with a 96\% 10-fold cross-validation accuracy.

Track: 11. General Track

Registration Id: 23NKBKV9VHK

Submission Number: 177