Chapter 7 because in surgery it can be highly relevant to be able to selectively enhance the contrast of one specific tissue type against all other tissues grouped together. For example, in laparoscopic colorectal surgery it can be very useful to enhance ureteral visualization to prevent iatrogenic injury11. In general the comparison has been made between one type of tissue, say ‘vessel’, and ‘all other tissue types’ aggregated in a pool. The corresponding labeling (Y vector, explained in the Supplementary section) is set to a “digital” one or zero meaning presence of the tissue or not. In Figure 7.4B the result of the above described method on two different tissue sets is depicted for tissue type “vessel” versus “all other tissue types”. A normal distribution is assumed as described in the following equation: Where: f(x) = Probability Density Function (PDF, see Supplementary section) σ = Standard Deviation (SD) μ = Label value on the x‐axis around which the PDF envelope curve is centered Two envelope curves, according to the assumed normal distribution, are plotted in the figure. They represent the assumed probability distribution of the algorithm results regarding the labeling of two groups of measurements. Figure 7.4B plots the individual results for ‘vessel’ centered around μvessel = 1, whereas ‘all other tissue types’ are centered around μother = 0 (colored asterixes indicate both groups). Figure 7.4C – 7.4F are likewise but for fat, bowel, tumor and ureter. Both envelope curves intersect at the crossover value of 0.5 on the X‐axis, which represents the decision border between both groups. The cumulative distribution function result12 (CDF, see Supplementary section) obtained at this cross‐over point of 0.5 now indicates the chance that the algorithm indeed correctly classified the tissue type under consideration. The CDF (0.5) thus can serve as a confidence indicating factor for tissue type recognition (i.e. the closer to 1, the better). For all considered tissue types the corresponding CDF (0.5) results and the associated standard deviation are listed in Table 7.2. The Supplementary section contains some more details on our data analysis. 98
proefschrift_Schols_SLV
To see the actual publication please follow the link above