PhD Defence Pim van den Berg

Integrated Photoacoustic/ultrasound imaging: applications and new techniques.

How can an integrated photoacoustic/ultrasound (PA/US) imaging probe be used? This thesis presented various examples of this. PAI is a unique combination of optical sensitivity to hemoglobin and other tissue chromophores and ultrasonic resolution, as we saw in chapter 1.

Next, we went into more depth on the probe’s design considerations and its performance in chapter 2. The combination of PA and US imaging modalities is a very suitable one. Both generate pressure waves in the ultrasound regime and can as such be detected using the same type of ultrasound transducers. They image complementary information as well: PA imaging (PAI) visualizes vasculature and functional information, whereas US imaging provides structural information and provides clinicians landmarks with which to interpret the PA data. However, for the best translation into the clinic, a PA system should be practical and cost-effective, something that cannot be said about current imaging systems. This can change with the imaging probe proposed in this chapter. It relies on a compact pulsed diode laser, which is integrated into the probe’s housing together with additional optics and the ultrasound detectors. Using a diffractive optical element that shapes the otherwise uneven light from the diode laser, the probe outputs a rectangular beam with a pulse energy of 0.56 mJ. The laser can be fired for short bursts of up to 10 kHz, which is important to minimize motion artefacts. The actual imaging rate is lower as multiple laser pulses are combined to improve the signal-to-noise ratio of the image. We achieved for instance an imaging rate of 10 PA images per second, while averaging over 20 laser pulses per image. In experiments on a tissue mimicking phantom we found the imaging depth can be improved from 5 down to 15 mm, but with a lower imaging rate. The resolution was determined, in phantom experiments on a sub-resolution hair, to be 0.28 mm axially (in depth) and between 0.4-0.6 mm laterally (sideways).

Pim van den Berg is a PhD student in the research group Biomedical Photonic Imaging, directed by Wiendelt Steenbergen.

In part I of this thesis we focused on the imaging applications of the integrated PA/US imaging. First, in chapter 3 we moved to a more powerful prototype that provides 1.3 mJ pulses. The chapter is a pre-clinical study on liver fibrosis in mice, a disease involving scar formation as a result of repeated liver damage. As no cure exists, currently animal studies are used in order to find one. In these studies, PA/US imaging can be a non-invasive alternative to histopathology, allowing to follow an animal over a longer period of time without sacrificing it at a certain stage of disease progression. In this study, CCl4 was administered repeatedly to the mice to induce fibrosis. After 6 weeks the mice (n = 6) were sacrificed to perform histology and compare this with PA/US imaging. We saw significantly more PA image features (p < 0.001) in the livers of fibrotic mice than a group of healthy mice (n = 5). A frequency analysis was performed on the radiofrequency PA signals. This showed that PA signals in fibrotic mice have a higher frequency content, and that this can be linked to fibrotic nodule formation. Histology confirmed fibrosis in the respective mice livers, showing significant increases in collagen I deposition and angiogenesis (p < 0.001 and p < 0.05 respectively). These histological markers of fibrosis furthermore correlated well with the amount of PA response from the livers, as well as with a quantification of the PA frequency content. This work shows that PAI using an integrated system is capable of assessing a fibrotic state, and may therefore have an important role in longitudinal animal studies.

Continuing with a clinical application in chapter 4, we studied detection of synovitis in finger joints of rheumatoid arthritis (RA) patients. Synovitis is an inflammation of the membrane around joints, and imaging of this inflammation can be used to predict how RA progresses. However, current modalities like MRI and ultrasound power Doppler (US-PD) have their drawbacks, mainly high cost and low reproducibility respectively. PAI may be a solution, with its high sensitivity to hemoglobin and its excellent ultrasound-based resolution. The goal of this feasibility study was to determine whether PAI is capable of detecting synovitis in joints with clinically evident inflammation. We made PA scans of clinically inflamed proximal interphalangeal joints of 10 patients, which were compared to non-inflamed joints from the same patients. In addition, joint scans were recorded of 7 healthy subjects, providing an added control group. When comparing these joint types, we found a significant increase in PA image features, on average 10 times more, near the joint membrane for the inflamed joints (p < 0.001). Each joint was also scanned using US-PD, with resulting images semi-quantitatively scored (0-3) by experienced rheumatologists who were blinded to the joint type. The US-PD score and the PA quantification agreed well (Spearman’s r = 0.64 and p < 0.001). This shows that we are capable of detecting clinically evident synovitis, but future research is required to investigate if PAI also has extra value over clinical examination. PAI may for example be capable of predicting how the disease would progress.

In part II of this thesis we focused on flow imaging using photoacoustics, a tool that may help future assessment of inflammatory, as well as cardiovascular diseases. Chapter 5 reviews first the state of the PA flow imaging field, which is quite broad due to the large number of ways to implement it. The main advantage of photoacoustics for flow imaging is that its sensitivity to hemoglobin in red blood cells (RBCs) allows a direct measurement of their flow with low tissue background amplitude. PA flow imaging techniques include those based on the Doppler effect, using a modulated continuous-wave laser and estimating the frequency shift in the returning PA response. When using pulsed light instead, RBCs generate a PA fingerprint that can be directly imaged. The ‘flow’ of the fingerprint can then be tracked to determine the velocity. Another approach is labelling of a part of the flow and tracking its subsequent motion is another good option, for instance by heating with high-intensity focused ultrasound it to increase its PA amplitude.

Chapter 6 investigates one of the flow imaging approaches, pulsed PA flow imaging, using the PA/US prototype probe, with the high pulse repetition frequency (PRF) of the integrated diode laser principally allowing continuous tracking of the PA fingerprint. The linear array of the system presents furthermore the opportunity to create a 2D map of flow velocities. The fingerprint signal was tracked over time using a cross-correlation to estimate the shift between laser pulses; this was done in a sliding window to obtain a flow estimation for each image pixel. PA flow imaging was studied in phantoms, both transparent and with tissue mimicking properties, with sub-resolution micro-particles flowing through tubing embedded in the phantoms. Results show flow imaging of velocities from 12 to 75 mm/s, with errors of 7% and 40% in transparent and realistic phantoms respectively. However, imaging of blood directly proved difficult due to the long pulse length of the diode laser, which limited the amplitude higher frequency PA signals.

Chapter 7 then goes on to show that PA flow imaging using the PA fingerprint of blood is feasible, though with a different PA system. To improve the detection of the PA response from RBCs, a high-frequency US probe is used: a 15 MHz linear array. The array was combined with a PIV-laser, which emits two timed pulses that are used to acquire PA image pairs a specified time apart. The shift in fingerprint signal during this specified time is computed using a windowed cross-correlation similar to chapter 6. Flow imaging was performed on both 40% RBCs in phosphate-buffered saline and on whole blood, with flow speeds ranging from 3 mm/s up to 25 mm/s. The relative errors were below 50%, down to 15%. The precision can likely be improved significantly by improving the setup, ensuring less distortions from floating particles, vibrations in the setup and differences in beam profile between laser pulse pairs.

In conclusion, chapter 8 summarizes the main findings of the thesis and the strength of the integrated imaging approach. It also highlights how single-wavelength PAI provides a treasure of information. The chapter also discusses the limitations of the work in the thesis, and how this can be improved in the future. For handheld imaging the contact mechanism between the probe and the skin should be improved and the imaging depth should be increased. PA flow imaging should demonstrate its advantage over ultrasound, ideally by combining it with multispectral PAI for estimating the rate of oxygen delivery.