2.1.

Integrated Catheter and System for Intravascular Imaging

We designed and built an integrated hybrid IVPA/IVUS catheter, see Fig. 1. Imaging of arteries was performed with the device described in earlier works.^21,26 A functionally similar probe was used for the PA spectroscopy measurements of cholesterol and cholesterol derivates, replacing the first one which broke after the experiment. Light is delivered to the vessel wall through 400-μm-diameter core optical fiber (Pioneer Optics, Bloomfield, Connecticut), and a lead magnesium niobate-lead titanate (PMN-PT) single crystal US transducer (custom built²⁷) was used to send and receive US waves. The tip of the fiber was polished at an angle of 34 deg and covered with a glued-on quartz cap to preserve an air-glass interface deflecting the beam by total reflection. The transducer had a center frequency of 44.5 MHz and a $-6\text{-}\mathrm{dB}$ fractional bandwidth of 45%. Fiber and transducer were mounted in an assembly with an outer diameter of 1 mm. The fiber tip and transducer center were separated by $\sim 1.1\mathrm{mm}$. The angle between the optical and acoustical beams was 22 deg; the beams overlapped between 0.5 and 4.5 mm from the transducer.

Fig. 1

Combined intravascular photoacoustics (IVPA) and intravascular ultrasound (IVUS) imaging. (a) The intravascular IVPA/IVUS catheter is inserted in an ex vivo human coronary artery and rotated in 1 deg steps. Laser pulses for IVPA imaging and electric pulses for pulse echo (IVUS) imaging are delivered to the vessel wall interleaved at every angular position. (b) Photograph of the catheter tip on the edge of a 10 eurocent coin. White arrow indicates 1 mm. (c) Photograph of the catheter tip, showing the light beam exiting the catheter. (d) Photograph of artery in setup, side view. On the left, the translation ($x$) and rotary stage are visible.

The catheter was connected to a spectroscopy/imaging system.²¹ A tunable laser (Vibrant B/355-II, OPOTEK, Santa Clara, California) provided the excitation light (pulse width 5 ns, repetition rate 10 Hz, pulse energy 1.2 mJ at catheter tip) for PA imaging. An arbitrary waveform generator (WW2571A, Tabor Electronics Tel Hanan, Israel) transmitted a Gaussian-modulated cosine wave for pulse echo imaging, which was transmitted to the probe through a custom-built expander and limiter. The received ultrasound and PA signals were band pass filtered (13 to 60 MHz 5th-order Butterworth, custom built), amplified by a 43-dB amplifier (AU1263, Miteq, Hauppauge, New York) and digitized at a sample frequency of $350\mathrm{MS}{\mathrm{s}}^{-1}$ by a 12-bit data acquisition card (Acqiris DP310, Agilent, Santa Clara, California). A schematic of the experiment and photos of the catheter are shown in Figs. 1(a)–1(c).

2.2.

Cadaver Coronary Artery Acquisition and Handling

We obtained coronary arteries from the Department of Pathology of the Erasmus MC at autopsy and imaged them within 24-h post mortem. Consent was obtained from the relatives and the protocol was sanctioned by the Medical Ethics Committee of the Erasmus MC (MEC-2007-081). The specimens were mounted on two cannulas in a tank containing saline solution. Large side branches were closed and arteries were pressurized to 100-mmHg to maintain an open lumen. A photograph of an artery mounted in the setup is shown in Fig. 1(d).

2.3.

Combined PA and US Imaging of Human Coronary Arteries

The cadaveric coronary arteries were imaged at room temperature within 24-h post mortem. We performed an IVUS pullback of the arteries (iLab system with Atlantis SR Pro catheters; Boston Scientific, Fremont, California) to find sites of interest, and marked them using suture needles. The combined IVPA/IVUS catheter was mounted in a rotation/translation stage [Steinmeyer, Albstadt, Germany; see Fig. 1(d)]. The suture needles were located and at these sites coregistered IVPA and IVUS cross-sectional scans were made.

Optical and US transmissions alternate for exactly coregistered data acquisition, as illustrated in Fig. 1(a). This procedure results in IVPA and IVUS images of the arterial cross section, as shown in Figs. 2(a) and 2(b). The acquisition time of one coregistered IVPA/IVUS image is 36 s. The image shown here was acquired at 1210 nm. The IVUS and IVPA data sets may be overlaid to create a compound display [Fig. 2(c)]. Spectral scans were made at selected angles. One scan results in a two-dimensional (2-D) data matrix, shown in Fig. 2(d), containing the PA spectra at the different distances from the catheter. The acquisition time of a 76 wavelengths, 32 times averaged data set is $\sim 5\min$. Local absorption spectra can be extracted, as shown in Fig. 2(e). Histology [Oil Red O stain; Fig. 2(f)] confirms the presence of lipids in the areas with a high PA signal at 1210 nm.

Fig. 2

Spectroscopic IVPA imaging. Coregistered IVPA and IVUS data are processed to create IVPA (a), IVUS (b) and combined IVPA/IVUS (c) images. At selected angular positions, indicated by the green line, 2D spectral IVPA (d) data are recorded by changing the wavelength of the light while keeping the catheter at rest. (e) Spectra at several locations in the vessel wall are extracted (f) IVPA and IVUS images are matched to histology stains for tissue characterization. Scale bars indicate 1 mm. Dynamic range for both ultrasound (grayscale) and Photoacoustic (PA, black-red-yellow) is 28 dB.

US images were created in conventional pulse-echo mode. We transmitted a 10-V peak to peak Gaussian-modulated cosine wave with a center frequency of 30 MHz and a 100% $-6\text{-}\mathrm{dB}$ bandwidth. 2-D images were obtained by rotating the catheter in 1-deg steps and repeating the US pulse-echo measurement in every direction.

IVPA images were acquired at several excitation wavelengths around 1200 nm. No averaging was applied for IVPA imaging; IVUS traces were averaged 10-fold. Digitized IVPA and IVUS data were band pass filtered (10 to 40 MHz 300th-order zero-phase forward and reverse finite impulse response filter), Tukey windowed, enveloped, and log compressed. The data were then converted to Cartesian coordinates and displayed using the “hot” and “gray” colormaps in MATLAB (The Mathworks, Natick, Massachusetts) for the IVPA and IVUS images, respectively. Combined IVPA/ IVUS images were created by overlaying the IVPA data on IVUS image using a nonlinear red-yellow-white color scale and a linear transparency scale. All data processing was done using MATLAB R2007b.

2.4.

PA Spectroscopy of Cholesterol and Cholesterol Derivates

A small volume of lipid material (cholesterol, cholesterol oleate, and cholesterol linoleate; all from Sigma Aldrich Chemie BV, Zwijndrecht, The Netherlands) was mixed with a minimal amount of water to eliminate all air from the sample. The densely packed lipids were then positioned directly in front of the catheter tip. They were contained in a holder, made of TPX (TPX® RT18; Goodfellow, Huntingdon, England) and designed to minimize acoustic reflections. We recorded spectral PA data by recording the signal trace generated by the lipids at wavelengths between 1125 and 1275 nm, spaced by 2 nm. This procedure resulted in a 2-D data set as in Fig. 2(d).

The magnitude of the locally generated PA signal is dependent on the local fluence, the absorption coefficient and the Grüneisen parameter of the tissue. The Grüneisen parameter, representing the thermo-acoustic efficiency, is dependent on the sound speed, the thermal expansion coefficient, and the specific heat at constant pressure. It therefore varies between different tissue types. Since the differences in local fluence are negligible within a small wavelength range, local variations in the PA signal amplitude with wavelength can be interpreted as changes in the absorption coefficient. We concentrated on the spectral range near 1200 nm where a prominent harmonic of the C─H stretch vibrational mode signals the presence of lipids.²⁸ Moreover, we take advantage of the fact that the absorption spectrum of connective tissue is markedly different from that of lipids in this wavelength range, while the absorption of water is relatively low. We analyzed the data and determined the 32 times averaged normalized PA spectra for the three lipids.

Robustness of the spectra was investigated by studying the individual variations in single-shot spectra, and monitoring the spectral shape for different depths in a sample. The stability of the spectrum was evaluated by repeating the measurement several times over 75 min.

2.5.

PA Spectroscopy of Human Coronary Arteries

At each imaging location, in addition to the IVPA/IVUS images acquired at several excitation wavelengths, full spectral scans were made at particular angular locations, which were selected based on IVUS morphology. Spectral PA data were recorded from 715 to 1800 nm with 5-nm steps. The data were then processed, as illustrated in Fig. 3. After the bandpass filtering operation described above, we selected regions containing the tissue and reference signals. The reference data contain nonsaturated signal close to the catheter that is due to light absorption in the catheter, which generates image artifacts. The amplitude of this signal was used to correct for variations in light intensity. Subsequently, the data were upsampled, jitter corrected, and downsampled to the original sample frequency. The data in the signal region was then Tukey windowed and corrected for variations in the light intensity using the jitter corrected reference data. We extracted the envelope of the tissue signal. Next, locations of high signal intensity were chosen automatically by selection of all peaks above a certain threshold in the 1195 nm enveloped signal trace. Spectra at those locations were noise filtered using a fourth order digital smoothing polynomial (Savitzky–Golay; SG) filter. Resulting spectra were compared to the PA spectra of cholesterol, cholesterol oleate, and cholesterol linoleate determined on the pure compounds, and to reference absorption spectra of collagen, elastin and a fatty acid mixture from literature.²⁹

Fig. 3

Spectroscopic intravascular photoacoustics (sIVPA) data processing to obtain depth-resolved IVPA spectra. (a) Band pass filtered PA traces at different wavelengths. Regions containing the tissue (bottom rectangle) and reference (top rectangle) signals are indicated. (b) Upsampling and jitter correction. (c) Light intensity-corrected data in the tissue signal region with colored circles marking peaks above a certain threshold in the signal envelope of the PA trace at 1195 nm. (d) Resulting spectra at those sites after noise filtering.

2.6.

Histopathology of the Atherosclerotic Human Arteries

After imaging, the arterial specimens were cut at the imaging plane and serially sectioned for histology staining. The segment proximal from the image plane was fixed in 4%-buffered formaldehyde and embedded in paraffin. The tissue block was serially sectioned (5-μm thickness) at the imaging plane and stained by Hematoxylin–Eosin (H&E) and Resorcin–Fuchsin (RF). The segment distal from the imaging plane was embedded in optimal cutting temperature compound (Tissue-Tek®, Sakura Finetek Europe B.V.), frozen in isopentane vapor at 77 K, and stored at $-80°\mathrm{C}$ until serial sectioning (20-μm thickness) for Oil Red O (ORO) and H&E. Tissue sections were viewed and digitized using a whole-slide scanner (Nanozoomer, Hamamatsu Photonics, K.K., Hamamatsu City, Japan). The ORO staining identifies lipids (stained red). The H&E stained sections were used to assess tissue morphology.

3.1.

PA Spectroscopy of Cholesterol and Cholesterol Derivates

Measurement of the PA signal of cholesterol, cholesterol oleate, and cholesterol linoleate in the 1200-nm range yielded three distinct PA spectra (Fig. 4). The absorption features in this wavelength range are caused by the second overtones of the vibrational modes of the C─H bonds in the molecules.^28,29 Wavelength shifts and intensity changes between the spectra of cholesterol, cholesterol oleate, and cholesterol linoleate originate from the differences in their molecular structures: the cholesterol esters have a fatty acid tail that consists mainly of CH2 groups. The two highest peaks at $\sim 1196$ (1192 nm for cholesterol) and 1210 nm are associated with the asymmetrical stretching modes of the methyl (CH3) and methylene (CH2) group vibrations, respectively, while the symmetrical stretching modes of the same groups are visible only as a shoulder or inflection points around 1151 and 1167 nm, respectively.

Fig. 4

Absorption spectra of cholesterol, cholesterol oleate, and cholesterol linoleate in the 1200-nm wavelength region as recorded by PA spectroscopy exhibit clear differences.

The robustness of the recorded spectra was evaluated by studying the reproducibility of spectra without averaging and by comparing averaged sets at different depths. Figure 5 summarizes the results of these experiments for one of the pure lipid compounds investigated. In Fig. 5(a), the difference between unaveraged spectra is shown to be very small, in the order of a few percent of the total signal. We attribute the remaining variation to measurement noise that is not adequately smoothed by the SG filter. Likewise, the spectra acquired at different depths are very similar [Fig. 5(b)].

Fig. 5

Robustness of lipid PA spectroscopy showing data from cholesterol linoleate as an example. (a) Comparison of “single-shot” spectra. (b) Spectra obtained at different depths. (c) Evolution of the spectral amplitude over time. (d) Data from panel (c) after normalization.

The stability of the absorption spectrum over time is demonstrated by the data in Figs. 5(c) and 5(d): we see a moderate increase in signal strength, but no change in spectral shape. The change in signal amplitude may be attributed to variations in packing density and water content in the lipid sample. Over time, the (unsaturated) lipid in the container is likely to be oxidized. We see that the spectrum appears to be robust against such variations.

3.2.

IVPA Spectroscopy of Human Coronary Arteries

Dual-modality imaging of a human coronary atherosclerotic artery (left anterior descending artery, male aged 56) ex vivo revealed a large eccentric plaque with high PA signal intensities [Fig. 6(a)] in the regions that were shown to be lipid-rich by histology [Fig. 6(b)]. The artery was embedded in peri-adventitial fat. An elaborate discussion of the separate IVUS and IVPA images at 1210 and 1230 nm may be found in Ref. 21.

Fig. 6

IVPA spectra of a human atherosclerotic coronary artery, ex vivo. (a) Dual-modality IVPA/IVUS image, arrowhead indicates suture needle. Along the white lines (# and *), spectral IVPA data were obtained. Normalized IVPA spectra at the locations of the dots are shown in (c) (#, blue dots), (d) (*, blue dots), (g) (*, green dot), and (h) (*, purple dots), compared to absorption spectra obtained in Fig. 4 and reference spectra for collagen, elastin, and fatty acids from literature.²⁹ (b) Histology section of the site imaged in (a), showing near circumferential intimal lipid accumulation (red) and presence of a calcification (Ca). The lipid rich core (#) and pathological intimal thickening (PIT; *) are magnified ($3\times$) in (e) and (f), respectively, showing the locations where the IVPA spectral data are recorded. (i) Hematoxylin–Eosin (H&E) histologic stain of the area #, showing reduced cellularity in the plaque. (j) H&E stain of the area *, showing PIT and an intact media. Spectra locations in (c), (d), (g), and (h) are stated in millimeter from the catheter tip. Oil Red O stain, $\mathrm{Lu}=\text{lumen}$, $\mathrm{PL}=\text{peri-adventitial lipid}$. Scale bars indicate 1 mm. Dynamic range of IVPA and IVUS image is 40 dB; color scales as in Fig. 2.

Positive confirmation of the presence and identification of lipids requires wavelength-dependent measurement. In the two directions indicated by the white lines, labeled as # and *, in the IVPA/IVUS image, we acquired spectroscopic IVPA data. Spectra in tissue were composed from one single IVPA acquisition per wavelength, creating a data set as in Fig. 2(d), and then plotting the signal strength as a function of wavelength at a single pixel in depth. We show the IVPA spectra of different tissues in Fig. 6.

The direction marked as # contained atherosclerotic plaque [Fig. 6(c)], while the line-of-sight marked as * intersected with the lipid-rich diseased intima [Fig. 6(d)], the adventitia [Fig. 6(g)]—the outermost layer of the vessel wall consisting of connective tissue—and a region of peri-adventitial tissue [Fig. 6(h)] composed mainly of adipocytes (fat cells) containing fatty acids.²⁹

We found that the lipid laden atherosclerotic plaque and diseased intima [Figs. 6(c) and 6(d), respectively] both had IVPA spectra that closely resemble the reference PA spectra of cholesterol oleate and cholesterol linoleate obtained on the pure compounds. This finding was in agreement with the assessment of the corresponding areas in the histological sections [Figs. 6(e), 6(i), 6(f), and 6(j), respectively], that showed lipid-rich foam cells and extracellular lipid droplets, but no cholesterol crystals. The H&E stain [Figs. 6(i) and 6(j)] shows differences in cell density and orientation of cells in the intima (cell nuclei are colored blue/purple). The lower cell density of the more central plaque areas [Fig. 6(i)] indicates a more advanced state of atherosclerosis with extracellular lipids (larger red spots in central region of ORO stain [Fig. 6(e)]. H&E [Fig. 6(j)] also illustrates the much larger adipocytes in the peri-adventitial region. The IVPA spectrum found in the adventitia, shown in Fig. 6(g), was very similar to both the absorption spectra of elastin and collagen, while the IVPA spectra of the lipid-rich peri-adventitial tissue [Fig. 6(h)] resembled the absorption spectrum of the reference fatty acid mixture.²⁹