Mieke Bus

7 107 During the development of the subsequent study by Freund et al, several methods were adapted compared to the study described above. (12) As discussed in chapter 4 and chap- ter 5, it is hypothesized that lesion grade correlates with µ OCT . In chapter 4 and chapter 5, µ OCT  was assessed based on circular OCT grey-scale images and used µ OCT as a cut-off for UTUC grading. As further development revealed, this method of calculating µ OCT did not accurately account for the a) confocal point spread function of the OCT probe and b) the sensitivity roll-off of the OCT system, since both were treated as a single offset (µ CAL ) to the attenuation coeffi- cient. Originally, the µ OCT  value was determined by adding µ CAL  to the fitted decay constant. This approach assumes a fixed value of the system induced attenuation and can be cali- brated on samples with known µ OCT  as described by Almasian et al. (13) Critically, it is assumed that the focus is located at the position where the probe is in contact with the tissue boundary. For Dragonfly probes applied in the upper urinary tract however, this assumption is invalid. The apparent Rayleigh length Z R , focus position z f  and roll-off parameters were, therefore, calibrated independently and used as fixed parameters in the µ OCT  analysis of the present study. Furthermore, analysis was done on the circular TIF images as provided by the used OCT system. These circular TIF images are presented in a 10 LOG greyscale format and an addi- tional adjustment for the log-transformation should have been applied to these images for correct µ OCT  analysis.  As a result, the µ OCT  was potentially underestimated. To correct for this possible underesti- mation, additional software was developed in which the circular OCT images were directly coupled to the raw amplitude OCT data used, which was used for analysis. Secondly, the fitting algorithm was adapted to correct for the sensitivity roll-off and confocal point spread function. A new µ OCT analysis with the new additional in-house software (LabVIEW 17.0) was per- formed. In the developed software, both circular scans and linear raw data were matched on a frame-to-frame basis (e.g. a single index of both multi-frame formats yields the same scan in circular and linear representation). Using these improved definitions for sensitivity roll-off and confocal point spread, the µ OCT  values are more accurate compared to the method described in chapters 4 and 5 and resulted in substantially higher µ OCT  values for low-grade (median µ OCT  3.3 mm -1  (IQR 2.7 – 3.7 mm -1 )) and high-grade UTUC (4.9 mm -1  (IQR 4.3 – 6.1 mm -1 )) (p=0.004). Performing ROC