2.1.

Animal Tissue Preparation

Fresh porcine small bowels were obtained from a local abattoir, cut into 20- to 30-cm long segments, moistened with physiological saline, and refrigerated at 4°C for up to 30 h (from the time of slaughter) until needed for fusion experiments. Prior to the fusion experiment, a segment of small bowel was selected and immediately dissected into 5-cm-long pieces for tissue-fusion experiment. Prepared 5 cm samples were kept hydrated in sealed plastic sample bags with saline and used within 30 min.

2.2.

RF Tissue Fusion

The RF tissue fusion setup is shown in Fig. 1. The RF generator is an energy research tool prototype (developed by Covidien, Boulder, Colorado) capable of delivering a programmable sinusoidal current from 0 to 7 A and a power from 0 to 350 W. An operating RF frequency of 472 kHz was chosen to avoid neuromuscular stimulation and electrocution. A bipolar anastomosis prototype was used as the tissue-sealing device in this experiment with jaws to clamp on the tissue sample, with RF energy supplied by the embedded electrodes in the jaws during the application of CP provided by an air compressor connected to the jaws. Prewritten RF energy control algorithms were loaded into the tissue fusion software written in LabVIEW [National Instruments (NI), Austin, Texas] in the PC to control the entire procedure. The algorithm was configured to control RF energy delivery to ensure a predetermined variation in tissue impedance profile,^2,3 i.e., to first raise the tissue impedance rapidly to a starting threshold and then to maintain a slowly rising impedance until the impedance finally reaches the preset end impedance.

Fig. 1

(a) Radiofrequency (RF) tissue fusion setup and (b) a typical tissue temperature profile during fusion.

During RF fusion, a piece of porcine small bowel sample was clamped between the fusion device jaws. The RF generator supplied the RF energy and also continuously monitored both the voltage and the current delivered to the tissue. The varying tissue impedance was then obtained by using the real-time voltage and current readouts. The air compressor was capable of supplying a variable CP from 0 to 0.5 MPa via a pneumatic system integrated with the anastomosis prototype device.

2.3.

Temperature Measurement

Tissue temperature was measured using a fine-tip (0.005 in.) Teflon-insulated J-type thermocouple (5TC-TT-J-36-36, Omega Engineering, Bridgeport, New Jersey). The thermocouple was inserted through slits made on the sealing device jaws and glued in place at the top of the slit, so that its tip emerged 0.25 mm above the electrode surface. In this way, the thermocouple was in contact with the tissue surface without piercing it and was insulated from the electrodes. The communication between the thermocouple and the computer was achieved through an NI PXI-6289 DAQ board and an NI SCC-68 terminal block. The latter hosted four NI SCC-TC02 Thermocouple Signal Conditioning Modules. Each SCC-TC02 could drive one thermocouple and had individual signal conditioning modules with a 2-Hz low-pass filter, which filtered out the RF signal and eliminated the RF interference from the thermocouple readout.

2.4.

Laser-Induced Tissue Fluorescence Measurement

The LIF system is described briefly in this section. The UV-laser fluorescence system employed two excitation laser diodes emitting at 375 and 405 nm. The rationale for the excitation-wavelength selection is given as follows. In existing large bowel ex vivo studies,^19,20,22 both 375- and 405-nm excitations were used. Furthermore, in one of the few small bowel studies,¹⁴ excitation wavelengths such as 760 and 820 nm were used for two-photon excitation. These correspond to 380- and 410-nm excitations in conventional fluorescence studies. In addition, as suggested by Zonios et al.,¹⁹ due to hemoglobin absorption, the collagen peak appears at 460 nm (red shift) instead for bowel tissues, and this collagen peak falls within our excitation and detection bands. A custom-made fiber-optic probe (Romack, Inc., Williamsburg, Virginia) formed a two-way laser delivery and fluorescence collection device, consisting of six hexagonally packed collection fibers surrounding an excitation fiber (Fig. 2). The diameter and numerical aperture of the fibers were 200 μm and 0.22, respectively, and the fiber material was chosen to have low autofluorescence and attenuation in the UV spectral region of interest. The probe distal tip was covered by a glass window, and the fibers and window were housed in a stainless steel tube with an external diameter of 2 mm. This could enable it to be inserted into small slits made on the sealing device jaws for future real-time analysis. The proximal end was divided into two arms: the first one contained the excitation fiber and was coupled to the laser optics using an SubMiniature version A (SMA) connector; the second arm included the six emission fibers arranged in a linear array inside another SMA connector. During LIF, the proximal end was made perpendicular to the sample plane in gentle contact with the sample without inducing any pressure.

Fig. 2

Experimental setup for the laser-induced tissue fluorescence system.

The laser outputs were collimated by 4.6-mm focal length lenses, and the beams were coupled into the excitation fiber using an 8-mm achromatic focusing lens. Two steering mirrors were used to optimize the alignment and to maximize the coupling efficiency. Two program-controlled beam shutters were placed in front of laser diodes to switch laser exposure on and off as well as to control the exposure durations. The output optical power at the distal end of the probe was 3.5 mW.

The fluorescence emission from the six collection fibers was focused onto the 80-μm-wide input slit of an imaging spectrograph (Specim Ltd.) using an achromatic lens doublet (60 and 30 mm). A 430-nm long-pass filter was inserted between these two lenses to block the excitation laser reflections from the sample. The light dispersed by the prism-grating-prism element of the spectrograph was then acquired with a sensitive-cooled CCD camera (Retiga EXI, QImaging, Surrey, Canada, $1392\times 1040\text{pixels}$). All optical components were assembled using a cage system and a breadboard so that the system was robust and portable. A LabVIEW program controlled the exposure time and beam shutters and enabled acquisition of fluorescent signals from the laser diode with an adjustable number of measurements.

2.5.

BP Measurement

The mechanical strength of the fused tissue was evaluated by a BP testing system custom built in the lab. This system consisted of a syringe pump, a pressure gauge, a sample injection needle, and a surgical clamp to close the small bowel tissue. The main arm of a Y-splitter tubing system was connected to a water-filled syringe controlled by the syringe pump. The other two split arms were connected to the pressure gauge and the sample injection needle, respectively. The surgical clamp sealed the other end of the piece of fused small bowel to make it a “tissue balloon.” The sample injection needle was used to pierce the small bowel tissue to allow water to be infused into the sealed bowel without damaging the seal. As the amount of water inside the tissue was increased at a rate of $20\mathrm{mL}/\min$ using the syringe pump, the pressure also increased until the fused tissue leaked or burst at the fusion line, and the highest value of water pressure recorded by the pressure gauge was the BP.

2.6.

Histopathology Analysis

To assess how the structural changes influence the tissue qualitatively, histological examination of RF-fused tissue was carried out. Samples of porcine small bowel were fused, and histological sections were taken before and after fusion. The samples were dissected and conserved in formaldehyde, stained with hematoxylin and eosin (H&E), sliced transversal to the seal, and were prepared on microscope slides.

3.1.

Tissue Temperature Evolution during Fusion

The tissue temperature variation was a result of the variation in RF energy supplied to the tissue as determined by the specific impedance-control algorithm used. The embedded thermocouple showed that the tissue temperature evolution during fusion had two stages [a typical tissue temperature profile during fusion is shown in Fig. 1(b)]: first the tissue temperature rose rapidly from the initial tissue temperature to a higher value usually within a few seconds and then the tissue temperature became relatively stable, where the tissue temperature variation was within the range of 60 to 90°C. Such temperatures ensure the necessary collagen denaturation^1,4 and are believed to be essential for a strong fusion. Temperatures that are too low ($<60°\mathrm{C}$) may not cause denaturation of collagen, and temperatures that are too high should be avoided in practice because it may lead to permanent damage of the tissue or necrosis.

3.2.

Tissue Fusion and BP Tests

Tissue samples from the same animal were used in the fusion experiment. Fusions were carried out by controlling the RF energy in order to achieve a predetermined tissue impedance variation with a rising slope of $0.01\mathrm{\Omega }/\mathrm{ms}$ to an end impedance of 200 Ω. Fifteen samples were fused with five each being fused at the following CPs: 0.05, 0.15, and 0.25 MPa. After fluorescence spectroscopy, the fusion strength of these samples was tested using the BP measurement device, and the results appear in Fig. 3. The BP results obtained were in line with the previous RF fusion strength study reported elsewhere.²³ Fusions made at higher CPs (0.15 and 0.25 MPa) had a higher mean BP of $\sim 40\mathrm{mmHg}$. Samples fused at 0.05 MPa of CP displayed an average BP of $<10\mathrm{mmHg}$. The histopathology and fluorescence spectroscopy results from these samples are given in the following sections to understand the difference in the fusion strength.

Fig. 3

Burst pressure (BP) test results versus fusion compressive pressure (CP).

3.3.

Histology

An additional six samples were selected from the same batch of tissues and fused (two for each CP: 0.05, 0.15, and 0.25 MPa). These samples were analyzed using fluorescence spectroscopy followed by histology. The fluorescence spectra for these samples are not shown here, because we note that they exhibit the same spectral characteristics and similar statistical and classification results as those presented in Fig. 5. The histology characteristics should remain the same for all these samples as they are from the same batch, as proven in another study.²³ Histological images for fused samples are shown in Fig. 4. At 0.05 MPa, some degree of thermal damage, as well as tissue compression, can be seen. Residual muscle can still be identified, and clear demarcation among tissue layers can be observed. A higher CP at 0.15 MPa resulted in a much thinner and homogeneous fused region with the disappearance of structural features. Serosal and muscle layers were highly compressed, the submucosa and mucosa were merged into one homogeneous layer, and a higher thermal damage could be seen. Finally, the observable features of fusions at 0.25 MPa CP is similar to that of 0.15 MPa fusions, although some degree of cracking can also be seen, which might be due to the formation of steam vacuoles during RF fusion. Samples fused at 0.05 MPa of CP were clearly underfused. Samples fused at 0.15 and 0.25 MPa of CP displayed similar architecture and features, being closer to a well-fused state. These histology results are consistent and reproducible among samples fused with equal fusion parameters.

Fig. 4

Hematoxylin and eosin (H&E)-stained histology images for samples fused at different CPs. Images on the left are the cross-sections cut transversally across the fusion line, and images on the right are the cross-sections cut parallel to the fusion line.

3.4.

Fluorescence Spectroscopy

In total, 804 fluorescence spectrum acquisitions were made on the 15 fused tissue samples with half made at each excitation wavelength (375 and 405 nm). Of these, 240 were “control” acquisitions on unfused tissue, 192 were “poorly fused” acquisitions at 0.05 MPa CP, and 372 were “well-fused” acquisitions at 0.15 and 0.25 MPa CPs. Figure 5 shows the mean fluorescence spectra and associated standard deviation error bars. For each excitation wavelength, both normalized (a.u., arbitrary units) and non-normalized (counts) spectra are displayed in order to show changes incurred to both waveform and overall intensity of the spectra.

Fig. 5

Mean emission spectra (440 to 700 nm) with standard deviation for “well-fused” versus “poorly fused” versus “control” samples: (a) 375-nm excitation and normalized to unit area, (b) 375-nm excitation and non-normalized, (c) 405-nm excitation and normalized to unit area, and (d) 405-nm excitation and non-normalized. The integration time used for collecting these spectra is 6 s.

The 375-nm laser excited mean fluorescence spectra displayed a main emission peak close to 455 nm, a slight shoulder around 500 to 515 nm, a secondary peak around 590 nm, and a narrow peak at 680 nm. The 405-nm laser excitation spectra showed a main peak in the 470- to 500-nm region and a secondary peak around 590 nm. In the normalized spectra, the steepness of the 375-nm excitation spectra seemed to increase in the 460- to 550-nm region from the “control” to “poorly fused” and then to “well-fused” samples. In the control cases, a shoulder was visible around 500 to 510 nm, which became less apparent for the increased fusion cases. For the normalized 405-nm excitation spectra, a slight blue shifting seemed to occur around 475 nm, and this primary peak had greater intensity. In the non-normalized spectra, overall intensity appeared to increase with fusion for both 375- and 405-nm excited spectra in the 440- to 470-nm and 470- to 490-nm bands, respectively. Additionally, not only did the average intensity increase with improved fusion, but also there was a greater variability in the magnitude of this intensity among the acquired spectra.

3.5.

Data Analysis

Kruskal–Wallis nonparametric testing was applied to the fluorescence spectra, and based on the features of the mean fluorescence spectra, two spectral parameters were chosen: (1) the ratio of normalized fluorescence emission intensity at two distinct wavelengths: $I_n(520\mathrm{nm})/I_n(490\mathrm{nm})$ for 375-nm excitation and $I_n(575\mathrm{nm})/I_n(475\mathrm{nm})$ for 405-nm excitation; (2) the average intensity between 440 and 470 nm for 375-nm excitation and between 470 and 490 nm for 405-nm excitation, symbolized as $I_t{(440\text{to}470\mathrm{nm})}_{\mathrm{av}}$ and $I_t{(470\text{to}490\mathrm{nm})}_{\mathrm{av}}$, respectively. Note that these two ratios were selected based on the direct observation of the prominent features in the mean fluorescence spectra in Fig. 5. Figure 6 shows the associated box plots for Kruskal–Wallis testing results. With the significant level $\alpha =0.05$, obtained $P$-values shows that the means for these two parameters are significantly different among the different classes.

Fig. 6

Box plots for the Kruskal–Wallis test data: (a) $I_n(520\mathrm{nm})/I_n(490\mathrm{nm})$ for 375-nm excitation, (b) $I_n(575\mathrm{nm})/I_n(475\mathrm{nm})$ for 405-nm excitation, (c) $I_t{(440\text{to}470\mathrm{nm})}_{\mathrm{av}}$ for 375-nm excitation, and (d) $I_t{(470\text{to}490\mathrm{nm})}_{\mathrm{av}}$ for 405-nm excitation. $P$ values for (a–d) are $4.24\times {10}^{-10}$, $2.30\times {10}^{-9}$, $1.12\times {10}^{-10}$, and $2.41\times {10}^{-9}$, respectively.

Sparse multinomial logistic regression (SMLR) was applied to the spectral dataset with the purpose of resorting to the most relevant spectral features, enabling classification of data into three classes: control, poorly fused, and well fused. Four hundred two input samples were used, where each one consisted of a concatenation of the normalized 375- and 405-nm excited spectra of a given acquisition, associated to its appropriate quality of fusion label. The fluorescence emission values in the 435- and 650-nm range were used as features in the SMLR test. To assess SMLR classification, a $K$-fold cross-validation strategy was adopted, where each subset contained the six nonindependent acquisitions made during each probing, yielding $K=67$.

The general accuracy in classifying the test data was 94.3%. For well-fused tissue (as defined by BP testing), it could be differentiated from poorly fused and control tissues via SMLR with 95.2% sensitivity, 95.4% specificity, 94.7% positive predictive value (PPV), and 95.8% negative predictive value (NPV). Poorly fused tissue could be distinguished from other tissue states with 89.5% sensitivity, 97.1% specificity, 90.5% PPV, and 96.7% NPV. Control tissue could be distinguished from any fused state with 100% sensitivity, 100% specificity, 100% PPV, and 100% NPV.