Page 161

proefschrift_Schols_SLV

Parathyroid differentiation during thyroid and parathyroid surgery by spectroscopy The quantitative results of LOO CV classification performance are listed in Table 10.4. LOO CV is based on the whole wide‐band spectral range covering both Si‐sensor and InGaAs‐sensor range data from thyroid and parathyroid surgery for train and test purposes. Table 10.4 Classification performance of combined Si‐sensor and InGaAs‐sensor features LOO CV TP TN Sensitivity Specificity PPV NPV Accuracy Parathyroid – Adipose 21/21 17/18 100 (81‐100) 94 (71‐100) 95 (75‐100) 100 (77‐100) 97 Parathyroid – Thyroid 18/21 18/23 86 (63‐96) 78 (56‐92) 78 (56‐92) 86 (63‐96) 82 TP = true positive; TN = true negative  numbers indicate identified tissue spots. A positive test is defined as the tissue observed being parathyroid gland; a negative test is defined as the tissue observed being adipose tissue / thyroid. Sensitivity; specificity; PPV = positive predictive value; NPV = negative predictive value; accuracy  numbers are percentages; numbers in parentheses indicate 95% confidence interval. LOO CV = leave‐one‐out cross–validation. 159 Discussion   The wide band (350 – 1830 nm) diffuse optical reflectance fingerprints of human thyroid, parathyroid and surrounding adipose tissue were identified in this explorative study. These spectra covered silicon (Si) and indium gallium arsenide (InGaAs) detector ranges, thereby exceeding wavelength boundaries (1600 nm or 1700 nm) reported by preceding work21,22,25,35. Spectroscopic measurements on human skin samples in the wavelength range of 1000 – 2200 nm have been reported previously36, but not regarding any of the tissues included by our study. Although the presented reflectance spectra (Figure 10.3) show great similarity between the three different tissue types, performance of automated tissue‐specific classification was fairly accurate in both sensor detection ranges (i.e. Si and InGaAs) to discriminate parathyroid tissue from either surrounding adipose tissue or adjacent thyroid tissue. Based on the classification accuracies (Table 10.2 and Table 10.3) we can conclude that Si and InGaAs sensors are equally suited for automated discrimination between parathyroid glands and surrounding adipose tissue or adjacent thyroid tissue. As tissue‐classification accuracies of around 80% are far from clinically relevant, we also investigated combining the two spectral sensor ranges, which resulted in improved classification performance regarding differentiation of parathyroid tissue from surrounding adipose tissue (Table 10.4).   Regarding the spectral signatures of the three tissue types (Figure 10.3) we observed a clear opposite course (i.e., ascending slope between 650 – 700 nm) for parathyroid compared to the spectra of adipose tissue and thyroid (i.e., descending slope). From


proefschrift_Schols_SLV
To see the actual publication please follow the link above