2.1.

Ureter Preparation and Application of Electrothermal Energy

This study was conducted according to an ex-vivo imaging protocol approved by the Institutional Animal Care and Use Committee at the University of Arizona. Twelve ureters were explanted from six female domestic Yorkshire swine within 2 h of euthanasia. Tissues were stored and flushed in 0.9% saline to prevent drying. Two ureters were set aside as untreated controls. Of the 10 remaining ureters, one was used to test optimal energy parameters, and nine were treated by the following protocol. A conventional bipolar high-frequency electrosurgical generator (Model 26021, Karl Storz, Tuttlingen, Germany) and a bipolar cautery tip (part no. 67-1065, Gemini Cautery System, Gaithersburg, Maryland, United States) were used for the external application of electrothermal energy. A weighted lever was used to elicit a controlled contact pressure of 105 g over $\sim 4{\mathrm{mm}}^2$ of the external ureteral surface.

Three treatment endpoints were selected based on externally visible signs of injury with 5 s of energy application (Fig. 1). Low-power lesions had mild to absent visible signs of injury. Medium-power lesions showed subtle coagulative injury (whitening and dehydration). High-power lesions had obvious signs of injury, including charring, contraction, desiccation, or evaporation. Three segments of each ureter were treated at low (16 W), medium (26 W), and high (36 W) powers with $\sim 3\mathrm{cm}$ between regions ($n=27$ segments). The order of power treatments was randomized among ureters.

Fig. 1

Treatment endpoints as determined by external visual inspection following 5 s of treatment. (a) Porcine ureter segment treated with 16 W of energy (low power), characterized by minimal to no grossly visible signs of injury. Normal appearing external connective tissue (ct), underlying smooth muscle (sm), and bright red blood vessels (bv) are appreciated. (b) Medium-power lesion treated with 26 W of energy (white box), characterized by subtle coagulative injury with tissue whitening, smooth muscle contraction (asterisk), and dark, coagulated blood vessels (arrowhead). (c) High-power lesion applied with 36 W (white box), characterized by obvious signs of coagulation (whitening) and, in some cases, gross ablation with darkening and charring of tissue. $\text{Scale bar}=2\mathrm{mm}$.

2.2.

OCT Image Acquisition, Processing, and Analysis

Immediately following treatment, ureters were mounted onto a dedicated 7F low-profile optical probe connected to a swept-source OCT volumetric laser endomicroscopy imaging system (95501-LP, Ninepoint NVisionVLE), described previously.²³ In short, this technology uses a near-infrared swept source laser with a wavelength range from 1260 to 1360 nm and a 50-kHz repetition rate for the continuous acquisition of two-dimensional (2D) cross-sectional OCT images. The axial and lateral resolutions of the system were 7 and $40\mu \mathrm{m}$, respectively. An automatic spiral-scanning pullback mechanism was used to acquire a series of up to 1200 stacked 2D images (6 cm) in $\sim 90\mathrm{s}$. Ureters were imaged in their entirety to include both normal and treated areas.

Image processing and analysis were conducted using ImageJ (National Institutes of Health, Bethesda, Maryland, United States) by an experienced observer. In the grayscale images, darker areas corresponded to higher signal reflectivity. The same observer qualitatively compared OCT images and the corresponding histology to characterize features of normal and injured ureters. Following confirmation of injury on histology, definitive criteria for injury were developed and used to label OCT images as normal or injured. The lesion size (mm) was determined by multiplying the number of consecutive slices with the injury by an image slice thickness of $50\mu \mathrm{m}$. Given the observer’s prior exposure to histology and adjacent images in OCT volumes, image labeling was considered unblinded. The extent of ureteral wall involvement and the severity of damage were assessed and compared among power groups.

2.3.

Histology Processing and Analysis

The proximal, middle, and distal segments of two untreated control ureters ($n=6$ segments) and segments of injured areas with adjacent untreated tissue ($n=27$ segments) were processed for histology using routine protocols. Longitudinal sections of $6\text{-}\mu \mathrm{m}$ thickness were acquired at $125\text{-}\mu \mathrm{m}$ intervals. Hematoxylin and eosin (H&E) and Masson’s trichrome-stained slides were analyzed using a brightfield microscope (BX41, Olympus, Japan). Slides with the maximal lesion size were formally reviewed and measured by a board-certified pathologist.

2.4.

Physician Detection Study

Three board-certified physicians specializing in Interventional Radiology, Pathology, and Urology, with no prior OCT experience, were trained to differentiate normal and injured ureters using 14 example OCT images. Physician detection accuracy was evaluated using a dataset of 126 OCT images with equal representation of normal ($n=63$ images) and injured ($n=63$ images) areas. The injured image dataset encompassed all lesions with confirmed injury on ground-truth histology. Based on the labeled lesion start and end slice number, three images of each lesion were acquired at 12.5%, 50%, and 75% longitudinal intervals. Images with significant distortion or unique artifacts, which were not discussed in training were excluded from the test dataset. For each ureter imaged, four to six OCT images of normal (untreated) areas were selected to match the number of injured images. We assured a minimum of 0.5-cm (100 images) distance between images and from injured areas.

2.5.

Statistical Analysis

The average lesion size (mm) for samples that met the criteria for injury on both OCT and histology was compared among power groups, presented as $\text{mean}\pm \text{standard deviation}$ for low ($n=5$ segments), medium ($n=8$ segments), and high ($n=9$ segments) power groups. Pearson’s correlation coefficient was used to assess the linear relationship among measurements of lesion size on OCT ($n=27$ segments) and histology ($n=27$ segments), including false positives and false negative measurements. A one-way analysis of variance (ANOVA) with post-hoc Tukey’s honest significant difference (HSD) test was used to compare lesion size as measured with OCT among power groups. Lesions identified on OCT that did not meet the criteria for injury on ground-truth histology were considered false positives, whereas lesions identified on histology that did not meet the criteria for injury on OCT were considered false negatives (see Sec. 3.2 for a description of the criteria). $p$-Values less than 0.05 were considered statistically significant. Statistical analysis and figures were generated using Rstudio (Version 2023.03.1+446).

3.1.

Comparison of Untreated Ureters on OCT and Histology

Untreated areas on OCT images demonstrated spatial changes in contrast and reflectivity that corresponded with the normal histological features of the urothelium, lamina propria, smooth muscle, and adventitia (Fig. 2). The multilayered transitional cell urothelium was identifiable on histology in all samples. Regions of isolated epithelial denudation $<10\mathrm{mm}$ in length were identified in $\sim 10\%$ of untreated controls and in normal areas distant from treated areas. On OCT images, the urothelium was only within the visible resolution in larger, relaxed ureters, where it appeared as an inner hyporeflective (light) band in regions with mucosal folds. The underlying lamina propria, characterized as a well-organized layer of dense, fibrillary collagens on histology, consistently appeared as a thin, uniform, hyperreflective (dark) band on OCT images. The smooth muscle layer, consisting of alternating circular and longitudinal muscular fibers on histology, correlated with the large heterogeneous layer with a striated texture on OCT images. The adventitia was identified as the outer hyporeflective layer with a diffuse hyperreflective speckled pattern.

Fig. 2

Comparison of untreated control porcine ureters on OCT and histology. (a), (b) Axial representations of an H&E and trichrome-stained ureter, respectively. (c) Example of an untreated porcine ureter on OCT images. Darker grayscale areas correspond to increased signal reflectivity and backscattering, which was more prominent in areas with higher collagen composition as shown in panel (b) (blue). U, urothelium; LP, lamina propria; SM, smooth muscle; A, adventitia.

OCT image analysis was not possible in areas with two types of artifacts. First, areas with folds showed increased scattering and signal attenuation across all layers beyond the urothelium. Second, probe slippage or non-concentric misalignment during acquisition could cause severe apparent distortion in the imaged tissue.

3.2.

Comparison of Injured Ureters on OCT and Histology

Electrothermal injury was identified in 23/27 (85%) treated segments on ground-truth histology. Four areas treated at low power (16 W) did not meet histological criteria for injury. On histology, injury was characterized by a focal transmural lesion with coagulative denaturation of collagen bundles and swelling and fragmentation of smooth muscle fibers (Fig. 3). On OCT images, electrothermally injured areas were identified by signal homogenization and attenuation, which corresponded to decreased signal contrast and reflectivity, respectively (Fig. 4). An image stack that demonstrates one high-power (36-W) lesion among normal areas of tissue is provided in Video 1 (see Appendix). In general, tissue swelling, or an increase in wall thickness compared with adjacent normal areas, was appreciated in injured areas. However, swelling was excluded from formal injury criteria due to the variability in ureter sizes across animals and among proximal, middle, and distal ureteral segments.

Fig. 3

Characterization of electrothermal injury on histology ($20\times$ magnification). The dashed vertical line indicates an approximate boundary between normal (left) and injured (right) areas. The horizontal dotted lines represent the approximate boundary among tissue layers. Electrothermal injury was identified in areas with denuded uroepithelium (U) and sloughing of epithelial cells (arrow); coagulative denaturation of collagen fibers in the lamina propria (LP), which is best appreciated at the lesion center (circle); and fragmentation and swelling of smooth muscle (SM) fibers (arrowhead). The adventitia (A) with blood vessels is appreciated but was not used for the evaluation of injury.

Fig. 4

Comparison of the morphological appearance of an injured porcine ureter (a) near the boundary and (b) at the center of a lesion treated with high-power (36-W) electrothermal energy. (a) Compared with normal areas, injured areas showed lightening of the lamina propria (LP*), decreased contrast and heterogeneity in the smooth muscle (SM*), and increased wall thickness (WT*). The dotted lines show the approximate boundaries of the lesion, which fulfills both injury criteria. (b) Toward the lesion center, the circumferential extent of injury and the wall thickness (WT*) increase, and the tissue appears more homogeneous. Characteristic features of normal tissue are faint to absent.

Two consistent features were observed on OCT images and were used as criteria for injury. Criteria 1 was characterized by a reduction in signal reflectivity in the lamina propria layer, resulting in visible lightening of the inner dark band. Criteria 2 was defined as the loss of microstructural detail (i.e., decreased contrast and loss of heterogeneous striations) in the smooth muscle layer. These criteria were exclusively assessed in artifact-free regions and had to demonstrate continuity for at least 1/8 (45 deg) of the total circumference.

Using these criteria, electrothermal injury was identified on OCT images in all but one histology-confirmed lesion (22/23, 95.7%). This medium-power (26-W) lesion measured 4.1 mm on histology and was suspicious for injury on OCT images, but it did not meet the OCT criteria due to its presence on a fold and was therefore considered a false negative. Among the four low-power lesions that did not meet the criteria for injury on histology, three were absent and one false positive was identified as a lesion of 0.9-mm length on OCT images. The average lesion size for lesions that met the criteria for injury on both OCT and histology is shown in Fig. 5. Comparison of OCT to histology measurements revealed a moderate positive correlation (Pearson correlation coefficient, $r=0.65$, $p=0.00087$).

Fig. 5

Average lesion $\text{size}\pm \text{standard deviation}$ for each lesion as measured with OCT and histology. OCT measurements of lesions treated with high power (36 W) were significantly larger than those treated with low power (16 W) ($p=0.038$).

3.3.

Comparison of Power Levels on OCT Images

Morphological features of lesions treated at all power levels on OCT images did not noticeably differ. The extent of damage observed circumferentially (Fig. 4) and measured longitudinally (Table 1) was the only characteristic that correlated with the power level. The lesion size as measured on OCT differed among power groups ($p=0.024$, ANOVA). Only the difference between low- and high-power lesions as measured on OCT was statistically significant ($p=0.038$, Tukey’s HSD).

Table 1

Comparison of lesion size among power groups as measured with OCT and histology, presented as mean ± standard deviation.

3.4.

Physician Detection Study

The results of the physician detection accuracy study are summarized in Table 2. The average time to complete the test was $28\pm 7\min$. An average of 112.3/126 (89%) normal and injured images were correctly interpreted by three physicians. The average sensitivity and specificity were 82% and 96%, respectively. Five images with injury consistently received false negative interpretations by all three physicians. An example is shown in Fig. 7. These images demonstrated decreased reflectivity in the lamina propria layer (criteria 1), whereas the microstructural detail in the smooth muscle layer remained unaffected or displayed a slight decrease in contrast (criteria 2).

Table 2

Results of the physician lesion detection accuracy study.