2.1.

High-Speed Swept-Source Optical Coherence Tomography System

Figure 1 shows the schematic diagram of the high-speed swept-source OCT system used in this study. The system employed a double-buffered Fourier domain mode-locked (FDML) laser that enabled an effective A-scan rate of $\sim 240\mathrm{kHz}$ with an output optical power of $\sim 60\mathrm{mW}$, a sweep range of $\sim 140\mathrm{nm}$, and a bandwidth of $\sim 90\mathrm{nm}$ full width at half maximum (FWHM) centered at $1.3\text{-}\mu \mathrm{m}$ wavelength. The configuration of the FDML light source was similar to previously reported studies.^27,28

Fig. 1

Schematic diagram of the high-speed swept-source OCT system (blue: optics; green: electronics). Inset (left): swept light source spectrum showing a near Gaussian spectral shape. Inset (top right): imaging setup showing volumetric (3-D) OCT imaging, where a cover glass was pressed over the specimen with a controlled pressure measured by the scale below. Inset (lower right): imaging setup for M-mode OCT, where the beam was scanned through the ablation device. MZI: Mach–Zehnder interferometer; TRG: trigger signal; DA: differential amplifier; RM: reference mirror; GS: galvanometer scanner; C: circulator; OBJ: objective; f1, f2: scan lens and tube lens of the relay optics; and DAQ: data acquisition.

Light emitted from the FDML laser was split and coupled into a Mach–Zehnder interferometer (MZI) and an OCT interferometer. The interference fringes from the MZI were detected by a dual-balanced detector (PDB430C, Thorlabs, Inc., New Jersey) with a 350-MHz detection bandwidth and were acquired once before the OCT imaging session to recalibrate the OCT interference signals. The OCT interferometer consisted of a $90/10$ fiber-optic coupler, an optical circulator (AC Photonics, California), a sample arm with a scanning microscope,²⁹ and a single-pass reference arm. In the microscope, a pair of closely spaced galvanometer scanners with 5-mm mirrors (6215 H, Cambridge Technology, Inc., Massachusetts) was used to provide two-dimensional (2-D) beam scanning of the specimen. Light from the FDML was collimated onto the scanner by using a near-infrared (NIR) achromatic lens ($f=18\mathrm{mm}$) and then relayed to the objective by two NIR achromatic doublets ($f1=f2=75\mathrm{mm}$). A long working distance NIR objective ($\mathrm{WD}=37.5\mathrm{mm}$, M Plan Apo NIR $5\times$, Mitutoyo, Japan) was used to provide sufficient space for the RFA catheter and tissue mounting as well as to achieve high transverse image resolution.

Light from the sample and reference arms was interfered in the $50/50$ fiber-optic coupler (AC Photonics, California) and detected by using a dual-balanced detector with identical specifications to the MZI detector. Both the calibration (MZI) and OCT signals were digitized by using an A/D card (ATS9350, Alazar Tech, Canada) at $500\mathrm{MS}/\mathrm{s}$. The measured system sensitivity was 107 dB with an incident power of $\sim 24\mathrm{mW}$ onto the sample, and the 6-dB sensitivity roll-off depth was $\sim 2.5\mathrm{mm}$ (in air). The axial image resolution was $7.4\mu \mathrm{m}$ (FWHM in tissue, assuming a tissue refractive index of 1.38 at $\sim 1310\text{-}\mathrm{nm}$ wavelength), and a transverse resolution of $12.1\mu \mathrm{m}$ (FWHM in air) was characterized by using the knife-edge method. The relatively fine transverse resolution yielded a short depth of focus (DOF) of $\sim 500\mu \mathrm{m}$.

2.2.

Radiofrequency Ablation Setup

A commercial RFA instrument with a focal RFA catheter (Barrx Flex, Model 90-9000 and Barrx 90, Medtronic, Minneapolis) was used in this study. The instrument allowed adjustment of the RF energy settings between 12 and $15\mathrm{J}/{\mathrm{cm}}^2$ (power density: $40\mathrm{W}/{\mathrm{cm}}^2$, frequency: 460 kHz, duration: $\sim 0.3\mathrm{s}$) for the Barrx 90 catheter. The catheter ablation paddle was $13\times 20{\mathrm{mm}}^2$ and covered by an electrode array of 24 electrodes. Each electrode was $250\text{-}\mu \mathrm{m}$ wide with a separation of $250\mu \mathrm{m}$ from the adjacent electrodes. To improve light transmission and reduce aberration through the ablation paddle, a small region ($\sim 4\times 4{\mathrm{mm}}^2$) on the back of the paddle was machined to thin supporting plastic material without affecting the electrodes' integrity. The modified ablation paddle is shown in the lower right inset in Fig. 1, where the region (red arrow) allows OCT imaging during RFA application.

2.3.

Specimen Handling, Optical Coherence Tomography Imaging, and Radiofrequency Ablation Protocol

Fresh swine esophagi were excised and stored in Dulbecco’s Modified Eagle’s Media (Cellgro, Corning, Virginia) at 4°C. Before the RFA application, the swine esophagus was dissected from the distal end (stomach side) to proximal end (larynx side) into $\sim 2\times 3{\mathrm{cm}}^2$ specimens to ensure sufficient contact between the RFA catheter (ablation paddle) and specimen. RFA was performed with specimens at room temperature, and each specimen was then sequentially ablated with different RFA energy dosages in individual protocols in order to avoid biased measurements caused by tissue variations between individual animals. In this study, two OCT imaging protocols with different RFA settings were performed in order to investigate the coagulum formation due to RFA application. A total of 24 specimens were collected from six swine esophagi [four specimens per esophagus (mean, range: 3 to 6)]. Previous studies on ex vivo human colon tissues reported a decrease in the EP thickness measured in the OCT images with increasing pressure exerted by the imaging catheters.³⁰ Therefore, during the OCT imaging, controlled pressure was applied in order to reduce the variability of the thickness measurements. The force exerted by an endoscope to a distally mounted focal ablation catheter was measured as $\sim 100\mathrm{g}$ over the same surface area as the ablation paddle ($\sim 1.3\times 2{\mathrm{cm}}^2$), corresponding to a pressure of $\sim 38\mathrm{g}/{\mathrm{cm}}^2$. Details of the individual protocols are described below.

2.4.

Quantitative Analysis of Optical Coherence Tomography and Histology Data

Following the RFA application and OCT imaging, the specimens used in both protocols were inked to indicate the OCT imaged area, placed in standard tissue cassettes, and fixed in 10% formalin prior to histology processing. Standard hematoxylin and eosin (H&E) staining was used to assess the coagulum from the RFA application.^13,14 Digital pathology images were captured from the H&E histology slides by using a slide scanner (Aperio AT2, Leica Biosystems, Illinois) with a $20\times$ magnification. Both the OCT images and corresponding H&E histology images were analyzed to identify the coagulum characteristics, including the coagulum thickness.

Post-RFA OCT images were examined in order to segment the coagulum, residual EP, and underlying lamina propria (LP) layers manually, based on the difference in scattering properties. The thickness of the coagulum and coagulum + residual EP layer was measured in individual cross-sectional OCT images by assuming a tissue refractive index of 1.38 at $\sim 1310\text{-}\mathrm{nm}$ wavelength.³¹ A 2-D thickness map was generated from volumetric post-RFA OCT datasets. The average, as well as minimum and maximum thickness, was obtained from the OCT measurements of individual specimens (volumetric or single cross-sectional OCT images) in order to assess the effects of RFA. For the subset of specimens processed for histology, $\sim 4$ to 5 histological sections with a step interval of $200\mu \mathrm{m}$ were obtained from the OCT imaged region. The thickness of the coagulum and residual EP layer was manually measured at multiple locations using a digital pathology viewer (Aperio ImageScope v.11, Leica Biosystems, Illinois). Similar to the OCT measurement, the average, as well as minimum and maximum thickness, was measured and compared with the corresponding OCT measurement.

3.1.

Radiofrequency Ablation Coagulum Thickness Analysis

Figure 2 shows the cross-sectional OCT and representative H&E histology images from a swine esophagus specimen treated with a single RFA application of $14\mathrm{J}/{\mathrm{cm}}^2$. As shown in Fig. 2(a), a three-layer tissue architecture, a homogenous hyperscattering layer near the tissue surface corresponding to the coagulum (Co), a thin hyposcattering layer corresponding to the residual EP layer, and a hyperscattering layer corresponding to the LP layer were identified and manually segmented [Fig. 2(b)]. 2-D thickness maps of the coagulum and coagulum + residual EP layer [Figs. 2(c) and 2(d)] from the segmented volumetric OCT datasets are also shown. Varying thickness of the coagulum and coagulum + residual EP layer can be observed across the imaging field, thus suggesting nonuniformity of the coagulum. The comparison of the coagulum and coagulum + residual EP layer thickness between OCT and histology is presented in the subsequent section. This three-layer tissue architecture was consistent with the corresponding histology image [Fig. 2(e)]. In the zoomed-in histology image [Fig. 2(f)], the coagulum is visible as layers of squamous cells with intracellular voids, which may cause the associated hyperscattering of the coagulum in the OCT images.

Fig. 2

Coagulated tissue analysis. (a) Cross-sectional OCT image from an RFA-treated ex vivo swine esophagus specimen shows tissue architecture reminiscent of the coagulum (Co), the residual EP (Res. EP), and underlying LP layer. (b) Cross-sectional OCT image of (a) with the coagulum and residual EP layers manually segmented to generate 2-D thickness maps of the (c) coagulum and (d) coagulum + residual EP layer, respectively. (e) Representative H&E stained histology confirms the three-layer architecture observed in (a). (f) The zoomed-in histology image over the region of interest (red dashed box) marked in (e) shows characteristic features of the coagulum. Scale bars: (a–e): $500\mu \mathrm{m}$; (f): $200\mu \mathrm{m}$.

3.2.

Volumetric Optical Coherence Tomography Imaging of the Radiofrequency Ablation Coagulum for Different Radiofrequency Energy Settings

The aim of protocol (1) was to investigate the feasibility of using OCT to identify and measure the changes of coagulum thickness in individual specimens after single RFA applications with different RF energy settings. Figure 3 shows the (a, b) cross-sectional OCT images as well as (c, d) thickness maps of the coagulum layer from two representative specimens ablated with 12 and $15\mathrm{J}/{\mathrm{cm}}^2$, respectively. The fraction of the residual EP layer in the specimen treated with a higher energy RFA ($15\mathrm{J}/{\mathrm{cm}}^2$) was decreased when compared with one treated with a lower energy RFA ($12\mathrm{J}/{\mathrm{cm}}^2$). The coagulum thickness for different energy RFA applications can also be seen in the 2-D coagulum thickness maps [Figs. 3(c) and 3(d)]. Figures 3(e) and 3(f) show histology from the locations near the cross-sectional OCT images [Figs. 3(a) and 3(b)], where the boundary between the coagulum and residual EP layer was not as discernable in the specimen treated with $12\mathrm{J}/{\mathrm{cm}}^2$ [Fig. 3(e)] when compared with $15\mathrm{J}/{\mathrm{cm}}^2$ [Fig. 3(f)].

Fig. 3

Coagulum thickness analysis from specimens ablated with single RFA applications at different RF energies (12 and $15\mathrm{J}/{\mathrm{cm}}^2$). (a, b) Cross-sectional OCT images corresponding to a location indicated in the (c, d) coagulum thickness maps. (e, f) H&E stained histology images close to the locations of the respective (a, b) cross-sectional OCT images. Scale bars: (a–f): $500\mu \mathrm{m}$. Res. EP: residual epithelium; Co: coagulum; LP: lamina propria; MM: muscularis mucosa.

The average coagulum thickness in individual specimens was measured by using the segmented post-RFA volumetric OCT datasets. The coagulum thickness increased with increasing ablation energy (Fig. 4), as expected. A statistically significant difference ($p<0.01$, student t-test) was observed in the coagulum thickness at 15 versus $12\mathrm{J}/{\mathrm{cm}}^2$. Table 1 shows the thickness of the coagulum and coagulum + residual EP layer measured in the OCT images versus the corresponding H&E histology images by using a subset of specimens ($12\mathrm{J}/{\mathrm{cm}}^2$: $N=2$; $15\mathrm{J}/{\mathrm{cm}}^2$: $N=2$), including the two specimens shown in Fig. 3. A variation of the average thickness of the coagulum + residual EP layer, which may be attributed to the thickness variation of the EP layer prior to RFA application, was observed in the OCT measurements, because the specimens were collected from different swine and different longitudinal locations from the esophagi. In histology, one of the specimens treated with a higher energy RFA (specimen C) showed comparable coagulum thickness to one treated with a lower energy (specimen A). This may be related to specimen handling and fixation. Specimens shrink during fixation, which makes the comparison of the absolute thickness measurement between the OCT and histology images difficult. To account for shrinkage, we computed the ratio of the OCT measurement to the histology measurement for the coagulum or coagulum + residual EP layer thickness in order to compare OCT with histology. Table 1 shows that this ratio (OCT/H&E) for the coagulum is similar to the coagulum + residual EP layer among the four specimens evaluated. This suggests that, although shrinkage varied between specimens, it was consistent among different tissue types (e.g., the coagulum versus the residual EP layer).

Fig. 4

Coagulum thickness measured in the post-RFA volumetric OCT datasets of individual ex vivo swine esophagus specimens treated with single RFA applications at different RF energy settings (12 to $15\mathrm{J}/{\mathrm{cm}}^2$). The average coagulum thickness measured in the volumetric OCT datasets of individual specimens was used to represent the coagulum thickness due to RFA with the designated energies.

Table 1

Comparison of the coagulum versus coagulum + residual EP thickness measurements between the OCT and H&E histology images.

3.3.

Concurrent Optical Coherence Tomography Imaging of Radiofrequency Ablation

Protocol (2) was designed to demonstrate the use of OCT for real-time monitoring of tissue changes during RFA. Commercially available focal RFA catheters were modified, as described in Sec. 2.3, to enable concurrent OCT imaging during the RFA application. Two consecutive RFA applications were applied to individual specimens with energy settings of $12\mathrm{J}/{\mathrm{cm}}^2$ ($N=5$) or $15\mathrm{J}/{\mathrm{cm}}^2$ ($N=6$). Figure 5 shows the representative dynamics of tissue architectural changes during the first (a–f) and second (g–l) RFA application at $12\mathrm{J}/{\mathrm{cm}}^2$. During the first RFA application, at time $T=0.00\mathrm{s}$, regular esophageal structures [EP, LP, and part of the muscularis mucosa (MM)] were observed beneath the RFA ablation electrodes (AE). At time $T=0.10$ to 0.25 s, hyperscattering regions corresponding to the coagulum were created by the thermal energy delivery. This also caused water vaporization over or within the specimen and, subsequently, the change in the position of the AE layer and the underlying specimen [diamond arrows, Fig. 5(d)]. This was due to water vapor ventilation, as noticed in the M-mode OCT images (visualization 1). Blurring of the tissue speckle pattern was also related to the tissue motion induced by heating and water vaporization [Figs. 5(c)–5(e)]. After $T=0.30\mathrm{s}$ [Fig. 5(f)], after RFA energy deposition concluded (duration: $\sim 0.3\mathrm{s}$), the boundary of the hyperscattering layer becomes stationary (arrows) and is delineated clearly, thereby allowing measurement of coagulum thickness, similar to Sec. 3.2. Below the coagulum, a thin hyposcattering layer corresponding to the residual EP layer is observed [Fig. 5(f)]. The second RFA application was performed without removing the coagulum [Figs. 5(g)–5(l)]. At time $T=0.10$ to 0.25 s, there was a slight loss of electrode-tissue contact and blurring of the tissue speckle pattern, similar to the first RFA [Figs. 5(i)–5(k)]. The depth of the hyperscattering layer further increased and became more uniformly distributed after the second RFA application. In addition, the boundary of the hyperscattering layer (coagulum) nearly overlapped with the initial EP/LP boundary in two-thirds of the imaging field, thus suggesting the EP layer was fully ablated over these regions. However, distinct architectural changes in the LP layer could not be identified in the OCT image after the second RFA application, thereby making it difficult to assess the depth of the ablation.

Fig. 5

Dynamics of RFA with an energy density of $12\mathrm{J}/{\mathrm{cm}}^2$ during the first (visualization 1) and second RFA applications (visualization 2). (a) The distinctive layered structure of swine esophagus, including the EP, LP, MM, and the RFA catheter AE layer can be clearly observed at time $T=0.00\mathrm{s}$. (b–e) Changes in the tissue architecture due to the RFA application include generation of a hyperscattering layer (red arrows) and changes in the specimen position and tissue contact (diamond arrows) starting around time $T=0.10\mathrm{s}$ to 0.30 s. (f) At time $T=0.30\mathrm{s}$, the changes in tissue architecture stopped after the conclusion of the RFA energy deposition (duration: $\sim 0.3\mathrm{s}$), and residual nonablated EP layer (R-EP) can be observed. (g–k) During the second RFA application, the tissue architectural changes were similar to those in (a–e). (l) At time $T=0.30\mathrm{s}$, the tissue architectural changes stopped, as in the first RFA application, and the boundary of the hyperscattering layer overlapped with the initial EP/LP boundary in most of the OCT imaging field (red arrow). s: second. Scale bars: $500\mu \mathrm{m}$.

Figure 6 shows the H&E histology image (a) near the OCT imaged site of Fig. 5 and (b) from a location $\sim 5\mathrm{mm}$ away from (a), but still within the ablated region. Similar to Fig. 2(b), the coagulum was characterized by layers of squamous cells with intracellular voids. In addition, the superficial LP layer near the OCT imaged site exhibited a different architectural appearance [arrows, Fig. 6(a)] when compared with the LP layer away from the OCT imaged site [arrows, Fig. 6(b)]. This difference can also be identified by comparing the tissue architecture of the superficial part of the LP layer (arrows) and the deeper part of the LP layer (stars) in Fig. 6(a). Both observations confirmed that the ablation extended from the EP into the superficial LP layer over the OCT imaged site, where the coagulum boundary nearly overlapped with the initial EP/LP boundary in most of the OCT imaging field. While both histology images were acquired within the regions treated with two consecutive RFA applications, the difference in the appearance of the superficial LP layer suggests that the thermal energy delivery was nonuniform across the ablated regions. The coagulum thickness observed in Fig. 6(a) was thinner when compared with Fig. 6(b), which may be related to sloughing of the coagulum during removal of the ablation catheter or the fixation process.

Fig. 6

Representative H&E stained histology images (a) near the imaged site as Fig. 5 and (b) from a location $\sim 5\mathrm{mm}$ away from (a). (a, b) Layers of squamous cells with intracellular spaces corresponding to the RFA coagulum, after two RFA applications, can be identified. Distinctive architectural appearance of the superficial part of the LP layer (red arrows) was observed in (a) when compared with the deeper part of the LP layer (stars, a) and (b) the LP layer from a different location (red arrows, b), thus suggesting the thermal energy delivery extended into the LP layer in (a). Scale bars: $200\mu \mathrm{m}$.

Figure 7 shows the dynamics of tissue architecture changes from a specimen treated with two consecutive RFA applications of $15\mathrm{J}/{\mathrm{cm}}^2$. The changes observed during the (a–f) first and (g–l) second RFA applications were similar to those in Fig. 5. However, due to the higher energy, the hyperscattering layer corresponding to the coagulum was thicker after the first RFA application. The boundary of the coagulum nearly overlapped the initial EP/LP boundary in some portions of the OCT imaging field [red arrow, Fig. 7(f)], thus suggesting a complete EP ablation over these regions. After the second RFA application [Figs. 7(g)–7(l)], the boundary of the hyperscattering layer overlapped the initial EP/LP boundary in nearly the entire OCT imaging field. In addition, no significant architectural changes within the LP layer as a result of the RFA application could be identified in the OCT image even though a higher energy setting was used. This makes it difficult to measure the increase in coagulum thickness resulting from the second RFA application.

Fig. 7

Dynamics of RFA with an energy density of $15\mathrm{J}/{\mathrm{cm}}^2$ during the first (visualization 3) and second RFA applications (visualization 4). (a) The distinctive layered structure of swine esophagus, including the EP, LP, MM, and RFA catheter AE layer can be clearly differentiated. (b–e) Changes in the tissue architectures due to RFA, including the generation of a hyperscattering layer (red arrows) and changes in the specimen position and tissue contact (diamond arrows), were observed at time $T=0.10\mathrm{s}$ to 0.30 s. (f) At time $T=0.30\mathrm{s}$, the changes in tissue architecture stopped after RFA deposition concluded (duration: $\sim 0.3\mathrm{s}$), and only a limited fraction of residual nonablated EP (R-EP) was observed due to the higher RF energy. (g–k) The architectural changes during the second application were similar to (a–e). (l) At time $T=0.30\mathrm{s}$, the changes stopped, and the boundary of the hyperscattering layer overlapped the initial EP/LP boundary in most of the OCT imaging field (red arrow). s: second. Scale bars: $500\mu \mathrm{m}$.

Table 2 summarizes the coagulum thickness measured in the M-mode OCT images of individual specimens after the first RFA application. Similar to Fig. 4, the specimens treated with $15\mathrm{J}/{\mathrm{cm}}^2$ RFA applications showed thicker coagulum than those treated with $12\mathrm{J}/{\mathrm{cm}}^2$. In addition, although the coagulum thickness after the first RFA application listed in Table 2 is higher than those reported in the previous section (Fig. 4 and Table 1), it should be noted that only cross-sectional OCT images from a single location were used in the analysis, rather than a volumetric OCT dataset as in first imaging protocol. Lastly, as shown in Figs. 5 and 7, there were no discernible OCT changes in the tissue scattering properties within the LP layer resulting from the RFA applications. Therefore, using OCT to measure the increase in the coagulum thickness from the second RFA application after it reached the depth of the LP layer was difficult.

Table 2

Coagulum thickness from two different energy settings measured in M-mode OCT after the first RFA application.