Multispectral characterization of tissues encountered during laparoscopic colorectal surgery 105 The result then is: Meaning that the new measured tissue most likely is of the first type, because the result value of 0.7348 is closer to 1 than to 0. And the observation was not likely to be of the second type because the result value of 0.13 is closer to 0 than to 1. However, it is clear there is an overlapping area between the label 0 and label 1, where it becomes difficult to decide which tissue type label is closest to the measured label. Therefore the statistical distribution of the trained labels is used to determine the threshold between 0 and 1 which separates one tissue type label from another. In the next paragraph the way this threshold is determined will be addressed. Important to mention is the case of more than two tissue types. More input labels may be defined in the label matrix Y. To calculate Figure 7.4A a larger matrix Yˆ with 5 rows is defined (plotted around numerical x‐values 0, 1, 2, 3, 4 for every tissue under consideration). PDF and CDF12 The cumulative distribution function (CDF) represents the ‘accumulated’ probability ‘up to x’ for a given Probability Density Function (PDF). CDF(x) is an anti‐derivative of PDF(x). In this study CDF has been used to determine the threshold values between the tissue type labels (0, 1, 2 …) which can be used to classify the calculated label of a new measured tissue type. Based on the calculated labels Yˆ on the training data X we can calculate the CDF of every tissue type’s calculated label. This actually is an indicator for the variation of the labels calculated by the TPCR method. Yˆ X B 0.7348 0.13
proefschrift_Schols_SLV
To see the actual publication please follow the link above