Fig. 1.

Overview of our approach. a Samples are drawn from our n-layered tissue model. Monte Carlo simulations are performed to evaluate the expected reflectance spectrum for each tissue sample. b The created spectra are adapted to fit the detection wavelengths of the multispectral imaging system. Noise is added, and the data are normalized and transformed to absorption. The normalized data and the physiological parameters coming from the tissue model are used to train a random forest regressor. c Our custom-built multispectral laparoscope is used to acquire multispectral images during interventions. Each pixel in these images corresponds to one reflectance measurement. d The regressor trained in (b) is used to estimate, e.g., oxygenation and blood volume fraction for each pixel in the multispectral image