NA-MIC Project Weeks

Skin Segmentation on MR to facilitate NousNav Registration

Key Investigators

• Reuben Dorent (King’s College London, UK)
• Tina Kapur (Brigham and Women’s Hospital, USA)
• Sarah Frisken (Brigham and Women’s Hospital, USA)
• Mohammad Jafari (Brigham and Women’s Hospital, USA)
• Samantha Horvath (Kitware)
• Jean-Christophe Fillion-Robin (Kitware)
• Harneet Cheema (Brigham and Women’s Hospital, USA)
• Fryderyk Kögl (BWH, TUM)

Project Description

Neuronavigation systems allow for visualization of pre-operative images and planning information to estimate the precise location of target surgical areas. Patient-to-image mapping is a key step in the workflow of these neuronavigation systems. Registration approaches typically rely on landmarks on pre-operative images as well as on the patient in the operating room (OR) [1]. An alternative approach would be to directly map the patient’s skin surface [2]. While extracting the skin surface of the patient in the OR can be performed using existing technologies (e.g., a pointer, a laser) [3], automatic skin surface extraction on scans remains an open problem. This project aims at developing an automated skin segmentation tool for pre-operative scans.

Objective

1. Create a database with manual annotations of the skin on T1w scans.
2. Develop a basic deep learning approach to perform skin segmentation.
3. Integrate the framework in Slicer.

Approach and Plan

1. Develop a method that is: 1/ robust to different acquisition protocols (e.g., scanner manufacturer, slice thickness, acquisition orientation); 2/ robust to different imaging modalities (e.g. CT, T1w MR, T2-FLAIR MR); relatively fast (~1-2min max).
2. Assess the performance of the proposed technique.
3. Develop a Slicer module with the pre-trained model.

Progress and Next Steps

1. The algorithm has been developed.
2. The Slicer module is now available here
3. The algorithm still needs to be benchmarked and qualitatively assessed

Illustrations

Background and References

1. Gerard, Ian J et al. “New Protocol for Skin Landmark Registration in Image-Guided Neurosurgery: Technical Note.” Neurosurgery vol. 11 Suppl 3 (2015): 376-80; discussion 380-1. doi:10.1227/NEU.0000000000000868
2. Shamir, R. R., Freiman, M., Joskowicz, L., Spektor, S., & Shoshan, Y. (2009). Surface-based facial scan registration in neuronavigation procedures: a clinical study, Journal of Neurosurgery JNS, 111(6), 1201-1206. Retrieved Jan 3, 2022, Link
3. BrainLab Softouch®