Automated spectroscopic tissue classification in colorectal surgery 115 Results In 10 consecutive patients (5 male, 5 female) undergoing open colorectal surgery 253 in vivo tissue spectra were recorded on 53 tissue sites (colon n=12, adipose n=9, muscle n=7, artery n=10, vein n=8, ureter n=7). Table 8.1 summarizes patient characteristics and the number of measured sites and spectra per tissue type. Figure 8.3 shows mean diffuse reflectance spectra for colon, adipose, muscle, artery, vein and ureter. In respectively Figure 8.4A and 8.4B mean spectra, with corresponding standard deviations, for ureter and artery are plotted paired with adipose tissue. Classification of spectral data in Si‐detector response range From the 36 extracted features (see Figure 8.2) 11 were located within the silicon detection range. Given the study sample of an average of 10 spots per tissue type, inclusion of maximum 1 feature is allowed. Binary logistic regression identified amplitude difference Ft6 (B6 – B3) as best distinctive feature for differentiation of ureter from adipose tissue. Gradient Ft4 (B6 – B1) was found to be best distinctive for differentiation of artery within mesenteric adipose tissue. Figure 8.5A and 8.5B show box plots for these Si‐detector based features, extracted for respectively ureter versus adipose tissue and artery versus adipose tissue. The quantitative results of classification performance are listed in Table 8.2. LOO cross‐validation is based on Si‐detector range data acquired during colorectal surgery for train and test purposes. Classification of spectral data in InGaAs‐detector response range From the 36 extracted features 25 were located within the spectral detection range of InGaAs (see Figure 8.2). After binary logistic regression, amplitude difference Ft35 (F3 ‐ F9) was selected as the most promising feature for differentiation of ureter from adipose tissue. For differentiation of artery from adipose tissue, gradient Ft13 (W2 – F4) was identified as best distinctive feature. Figure 8.6A and 8.6B show box plots for these InGaAs‐detector based features, extracted for respectively ureter versus adipose tissue and artery versus adipose tissue. The quantitative results of LOO cross‐validation classification performance are listed in Table 8.3. LOO cross‐validation is solely based on InGaAs‐detector range data acquired during colorectal surgery for train and test purposes.
proefschrift_Schols_SLV
To see the actual publication please follow the link above