Optical coherence tomography: an experimental validation for vascular imaging of saphenous vein bypass grafts

To measure the accuracy of OCT and iScan we had the following objectives: (1) to obtain OCT diameter measurements of the cross section of synthetic grafts and cadaveric vessels using both vendor software and iScan; (2) to validate the OCT vendor and iScan measurements using photomicroscopy; and (3) to compare the accuracy of iScan measurements with OCT vendor software measurements.

To allow for the validation of OCT, phantoms of known dimensions, which closely match that of a human coronary artery and mimic the optical properties of arterial tissue, were manufactured in house using precision moulds of specified dimensions that were custom manufactured in-house (NHS Medical Physics). The matrix material of the phantom was RTV silicone (Slygard 184) kit which was made in a ratio of resin:reactive 10:1 and alumina powder was added as the main scattering component. Carbon black was added in the wall for further attenuation of light. The silicone resin was mixed with 50 mg ml⁻¹ of alumina powder and cast into the mould, this combination was based on research previously published by Bisaillon et al [11]. To ensure a uniform distribution of scattering material, the mixture was fornicated in an ultrasonic bath for 5 h before adding the reactive. It is cured at room temperature for 48 h.

Phantom version one (figure 1(a)) had a diameter 3.5 mm and a wall thickness of 1 mm throughout. There were two phantoms made to version 1 specifications. Version 2 (figure 1(b)) introduced a tube with a variable diameter, giving three zones: 4.5, 3.5 and 2.5 mm diameters with wall thicknesses of 0.5 mm, 1.0 mm and 1.5 mm respectively. The purpose of the variable diameter was to test the sensitivity of iScan to minor changes more characteristic of human vessels by mimicking the presence of stenosis. Version 3 (figure 1(c)) included different diameters (4.5, 3.5 and 2.5 mm) and different wall thicknesses of 1.25, 1.75and 2.25 mm. There were two phantoms made to version 3 specifications (3i and 3ii). Version 4 (figure 1(d)) introduced a 'plaque like bump' of 0.1 mm thickness. Version 5 (figure 1(e)) was created with 2 'plaque like bumps' for further testing. These plaques were designed to give a stronger OCT signal than the surrounding phantom, mimicking the characteristic signal of real plaques. All versions above were imaged but to facilitate clear comparison of the images for the reader, only version 3i has been shown in the figures section.

Zoom In Zoom Out Reset image size

Eight saphenous veins were obtained from cadavers at the Anatomy Department, University of Glasgow, which were fixed in formalin and stored in a safe and cool place throughout the study period. All vessels were approximately 80–100 mm in length and diameters of approximately 2–4 mm.

Experimental OCT imaging was performed with the St. Jude Medical Ilumien system. The OCT Dragonfly catheter (2.7 French) was purged with 3 ml of H₂O and then 1 ml of omnipaque contrast (300 mg I ml⁻¹). The phantom or vessel was placed in the holding apparatus and submerged in a container filled with water. A six French sheath was placed 1 cm into the proximal end of the graft. A 0.014 inch guidewire was inserted into the end of the OCT catheter and then they were together passed through the sheath and advanced to the distal end of the graft. The graft was flushed with H₂O to remove air bubbles.

The light of the OCT probe was aligned at the proximal end of the phantom and the system calibrated. The automatic pullback was commenced (20 mm s⁻¹) at the same time as an 11 ml bolus of contrast was injected. The whole process was repeated a total of five times.

The St Jude OCT automatic detection tool measured the mean diameter measurements of OCT images.

The measurements of the customized phantoms were known before testing but because of the possibility that the dimensions may vary during or after the manufacturing process, the actual dimensions of each phantom were acquired by photomicroscopy shortly after OCT had been performed. The same approach was adopted for the cadaveric vessels to provide a gold standard for comparison. In this way, the actual dimensions of each phantom and blood vessel were used for comparison with the OCT findings.

Five corresponding 1 mm sections measured by OCT were sliced from each graft by hand using a scalpel with the grafts fixed to a measuring scale. Practice and care was taken not to damage the geometry of the graft during slicing. These sections were measured by a photomicroscope (Olympus SX10) at ×2 magnification, using Nikon software 2.3 (NIS-Elements). The microscopy image (figure 2) was superimposed on a circular target to guarantee the line of diameter crossed the centre of the section. The microscope was calibrated and a minimum of ten measurements of inner diameter were taken at adjacent points on the inner circumference and the mean calculated. It is possible there was interobserver error, however this was low. The average standard deviation of the ten measurements was 5.4 μm (minimum 2.08 μm and maximum 9.97 μm).

Zoom In Zoom Out Reset image size

In order to provide open access to the software for OCT research purposes and measure more parameters than the OCT vendor software, X.J of the Medical Devices Unit, NHS GGC, developed a novel imaging software suite (Medipass-iScan) [12] in-house. iScan was an ImageJ plugin using dynamic programming [13]. Dynamic programming is an optimization approach that simply stated, bypasses local minima which has been reported for the lumen and plaque detections for in vivo OCT images [14–16]. Similar to the approach proposed by Wang et al [16], the developed software detected the smoothest lumen boundary which was close to the available data and which at the same time preserve the discontinuities. The central circular regions of the OCT images were identical for all the pullback images. The central circular regions were removed by typing in the pixel coordinates of the sensor and the diameter of the circle to the software. The OCT polar images were transferred to rectangular images. The dynamic programming was applied to the rectangular images; firstly, the guide wire was detected and removed from the images; and then the lumen boundaries were detected [16]. All OCT pullback images were recorded and digitally archived. Images that corresponded with the cross section of the graft measured by photomicroscopy were measured offline using iScan software. The tube properties derived were: inner diameter, luminal area and wall thickness.

Figure 3(a) presents an OCT image of phantom version 3i with all three zones visible on the longitudinal view at the bottom. The 0.014 mm guidewire as labeled resulted in the long shadow through the wall of the phantom; the circular structure with turquoise borders was the OCT catheter. The uneven reflections in the wall were indicative of the alumina powder, which was dispersed throughout the wall to reflect the light. Figure 3(b) shows the OCT image analyzed by iScan. Measurements of vendor images and iScan images were then compared to our gold standard, photomicroscopy. Figure 2 shows photomicroscopy measuring of a phantom.

Zoom In Zoom Out Reset image size

Linear regression of iScan OCT diameters compared to the photomicroscopy measurements demonstrated that the iScan OCT measurements of microscopy were closely related with R² = 0.901.

The Bland–Altman plot (figure 4) represents the differences between the photomicroscopy and iScan OCT measurements, plotted against the average of photomicroscopy and iScan OCT (one sample t-test of the differences p < 0.001). There was evidence of limited bias (104 μm) and 95% limits of agreement: −100 and 308 μm.

Zoom In Zoom Out Reset image size

The mean variation of iScan OCT measurements from microscopy was 3.02% (min −26%, max 88%) with SD of 7.96%. The mean variation of OCT vendor measurements from microscopy was 3.03% (min −15.56%, max 15.62%) with SD of 3.94%.

Figure 5(a) presents an OCT image of a saphenous vein ex vivo. Figure 5(b) shows the OCT image analyzed by iScan. Figure 6 shows photomicroscopy of a section of saphenous vein.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 7 displays linear regression of iScan OCT diameters of saphenous veins ex vivo compared to the photomicroscopy measurements. It demonstrated that the image measurements of iScan OCT were positively related with R² = 0.704.

Zoom In Zoom Out Reset image size

The Bland–Altman plot of saphenous vein imaging (figure 8) represents the differences between the photomicroscopy and iScan OCT measurements, plotted against the average of photomicroscopy and iScan OCT (one sample t-test of the differences p = 0.051). There was evidence of limited bias (116.24 μm), similar to the phantom grafts. 95% limits of agreement were −1754 and 1987 μm.

Zoom In Zoom Out Reset image size

The mean variation of iScan OCT measurements of saphenous vein ex vivo from microscopy was 7.62% (min −65.02%, max 88.7%) with SD of 28.98%. The mean variation of OCT vendor measurements from microscopy was 16.74% (min −50.94%, max 79.63%) with SD of 25.73%.

The phantom was cut at e.g. 30 mm along the phantom as seen by the naked eye but this may have actually been at e.g. 30.1 mm. The phantom was fixed with tape to a measuring scale and cut at 2 mm intervals with an average slice thickness of 0.5 mm. An automated cutting device would have removed this limitation but was not available.

The specimen lumen was in most instances circular during photomicroscopy, as could be verified by the circular target superimposed. However due to the preserving solutions used on the saphenous veins making them firmer than in vivo, the circularity of some sections were distorted during cutting. We tried to minimize this error by readjusting the section to the circular target and also by taking an average of the ten diameter measurements to reduce possible bias.

There were instances where iScan did not detect the entire border of the images lumen as artefacts e.g. guidewire reflections, were detected instead of the inner border. As diameter measurements were an average of five frames, this will have minimized a measurement error if for instance, the guidewire was detected instead of the lumen wall. The algorithm to rule out the guidewire shadow is being constantly improved upon, and this should not be an issue in future versions.

The mean difference of measurements of saphenous veins from microscopy was smaller for iScan than the vendor software however, the 95% limits of agreement was smaller with the vendor software than with iScan. We believe the increased variances or 95% limits of agreement with iScan were mainly due to the section alignment. The OCT measured sections were perpendicular to the guide wire; whereas the photomicroscopy measured sections were perpendicular to the middle axis of the vessel. Since the guide wire was not necessary aligned with the middle axis, the larger the diameter, the bigger the discrepancy of the sections measured.

This research has demonstrated that OCT is accurate when measuring inner diameter and this has been validated by photomicroscopy. OCT has also been shown to detect even the earliest stages of intimal thickening; this can be seen on OCT as bright tissue with a homogeneous thin rim [4]. However, there is currently no established cut-off value for what categorizes neointimal growth as pathologic [9]. OCT's accuracy makes it most suitable to measure neointima and so potentially OCT may be used to assess the burden of disease in vein grafts for clinical and research purposes.