Robust near real-time estimation of physiological parameters from megapixel multispectral images with inverse Monte Carlo and random forest regression

• DOI: 10.1007/s11548-016-1376-5
• Publication Date: 2016-05-03T09:55:47Z
• Authors: Sebastian J. Wirkert; Hannes Kenngott; Benjamin Mayer; Patrick Mietkowski; Martin Wagner; Peter Sauer; Neil T. Clancy; Daniel S. Elson; Lena Maier-Hein
• Keywords: Robustness

Multispectral imaging can provide reflectance measurements at multiple spectral bands for each image pixel. These be used estimation of important physiological parameters, such as oxygenation, which indicators the success surgical treatment or presence abnormal tissue. The goal this work was to develop a method estimate parameters in an accurate and rapid manner suited modern high-resolution laparoscopic images.While previous methods oxygenation are based on either simple linear complex model-based approaches exclusively off-line processing, we propose new approach that combines high accuracy with speed robustness machine learning methods. Our concept is training random forest regressors using spectra generated Monte Carlo simulations.According extensive silico vivo experiments, features higher than state-of-the-art online orders magnitude faster other nonlinear regression methods.Our current implementation allows near real-time from megapixel multispectral images thus well tissue analysis.