Improving spatial resolution and contrast in ultrasound modulated optical tomography

Improving spatial resolution and contrast in ultrasound modulated optical tomography NT Huynh, H Ruan, ML Mather, BR Hayes-Gill, SP Morgan* Electrical Systems and Optics Research Division, Faculty of Engineering, University of Nottingham, University Park, Nottingham, NG7 2RD, UK. Steve.morgan@nottingham.ac.uk ABSTRACT Ultrasound imaging has benefited from non-linear approaches to improve image resolution and reduce the effects of side-lobes. A system for performing second harmonic ultrasound modulated optical tomography is demonstrated which incorporates both pulsed optical illumination and acoustic excitation. A pulse acoustic inversion scheme is employed which allows the second harmonic ultrasound modulated optical signal to be obtained while still maintaining a short pulse length of the acoustic excitation. For the experiments carried out the method demonstrates a reduction in the effective line spread function from 4mm for the fundamental to 2.4mm for the second harmonic. The first use of pulsed ultrasound modulated optical tomography in imaging fluorescent targets is also discussed. Simple experiments show that by changing the length of the acoustic pulse the image contrast can be optimized. The modulation depth of the detected signal is greatest when the length of the object along the acoustic axis is an odd number of half wavelengths and is weakest when the object is an integer multiple of an acoustic wavelength. Preliminary ultrasound modulated imaging results are also presented where the target generates light within the medium without the use of an external light source. Although signal to noise ratio is likely to be a major challenge, this result highlights a potentially useful application of ultrasound modulated optical tomography in bio- or chemi-luminescence imaging. Keywords: ultrasound modulated optical tomography, pulse inversion, fluorescence, bioluminescence, chemiluminescence 1. INTRODUCTION Ultrasound modulated optical tomography (USMOT) is a hybrid imaging technique that can reduce the effects of light scattering and improve the resolution of optical imaging systems by tagging light that passes through the ultrasound (US) column. The technique has potential for imaging and spectroscopy of tissue based on the absorption of coherent light by chromophores such as oxy- and deoxy-hemoglobin. It also has the potential to image fluorescence, although in this case incoherent light is used and the signal to noise ratio (SNR) is 2-3 orders of magnitude 1 lower than in the case when coherent light is detected. It is therefore of interest to investigate techniques to improve both imaging resolution and contrast. US imaging has benefited from non-linear approaches to improve image resolution and reduce side-lobes 2. Second harmonic imaging can also be applied in USMOT, for example, Selb et al 3 demonstrated a technique based on a lock-in CCD method to extract the second harmonic signal and improve imaging resolution. To improve axial resolution pulsed US can be employed, however, short pulses correspond to broad spectral bands and results in overlapping fundamental and second harmonic spectra. In conventional US, this can be overcome by employing a pulse inversion technique 4. In this paper we describe how pulse inversion can be employed to perform second harmonic USMOT. This results in high resolution pulsed USMOT imaging with ideally no overlapping fundamental and second harmonic spectra. The use of USMOT in imaging fluorescent targets is also discussed. This is challenging because the modulated light signal is much smaller than when coherent light is detected. A fundamental understanding of the factors that affect image contrast is therefore important. A system is described that is based upon, to our knowledge, the first use of pulsed acoustic excitation in an ultrasound modulated fluorescence tomography (USMFT) system. Simple experiments and

simulations show that image contrast depends upon the relative size of the fluorescent target and the acoustic wavelength. Finally this system is used to generate a line-scan of a chemi-luminescent object which generates light without requiring an external light source. This is an application where photoacoustic imaging cannot be applied and so offers an interesting area of research for USMOT in bio- and chemi-luminescence imaging. 2.1 Pulse Inversion 2. THEORY Pulse inversion involves exciting an ultrasound transducer consecutively with a pulse and then an inverted pulse. Summing the detected pulses allows the second harmonic signal to be extracted. Figure 1 shows the principle of the pulse inversion technique. The top row of the figure shows the case where no harmonics are generated (in practice this occurs when low acoustic intensities are applied). In the absence of harmonic distortion, the summation of a pulse followed by an inverted pulse results in zero output. In the presence of harmonic distortion (second row) the summation of two distorted pulses results in the cancellation of the fundamental while maintaining the second harmonic component. As discussed, the advantage of this approach is that the second harmonic signal can be obtained while still maintaining a short pulse length of the acoustic excitation. Figure 1 With no harmonic distortion (top row) a pulse (column 1) summed with an inverted pulse (column 2) results in zero output (column 3). In the presence of harmonic distortion (bottom row) the summation of a pulse and an inverted pulse removes the fundamental component but retains the second harmonic component. A speckle detection algorithm has been developed to perform pulse inversion USMOT based on previous algorithms Error! Reference source not found., Error! Reference source not found.. Pairs of ultrasound pulses (in this case tone bursts at 0 and ) are applied to a scattering sample and at a time corresponding to the propagation of the pulse to the ultrasound focus, strobe illumination extracts the contribution of a portion of the tone burst to the detected phase modulated speckle pattern. To allow a phase stepping algorithm to be applied in order to obtain the amplitude of the second harmonic, a phase shift is applied to the pairs of ultrasound pulses in the next detector frame (in this case tone bursts at /2 and 3 /2). Similar to previous algorithms 5,6, multiple pairs (~1000) of optical pulses are summed at the detector over the image acquisition time of the camera. The following derivation assumes that the fundamental frequency is completely cancelled by adding tone bursts of 0 and phase, the second harmonic frequency signal remains; and the modulated light intensity on the th pixel that is sampled by a short laser pulse (assumed to be a delta function in this derivation) is;, (1)

where is the second harmonic ultrasound modulated laser light intensity; is the fundamental ultrasound frequency; is the time delay due to the ultrasound pulse travelling from the transducer to its focal point; is the initial phase of an individual speckle. Summing tone bursts of phase /2 and 3 /2 within the next camera frame, the second harmonic frequency modulated speckle intensity sampled by the short pulse laser at is; The difference of these two pixels from each image;. (2). (3) As is random over the whole speckle pattern, taking the average of the square of the sin function is 6 ;, (4), (5) where denotes the averaging over the detected speckle pattern. The modulation depth is defined as;, (6) where is the mean intensity of the speckle pattern image, therefore;. (7) is the detected signal used to obtain images. 2.2 Ultrasound modulated fluorescence tomography We have previously demonstrated 7 that the shape of the temporal signal produced in a pulsed USMOT experiment can be simply modelled by a convolution of the optical profile along the optical axis with the acoustic pulse that propagates along the axis (figure 2). As an ultrasonic pulse propagates through the sample, at a particular time, it introduces a pressure change (compression or rarefaction) at a particular volumetric element of the medium, which contains the ultrasonic pressure at a given point in time (defined as a layer ). This modulates the sample s optical properties (scattering coefficient, absorption coefficient and refractive index) within that layer. Light emerging from the layer is phase modulated by the ultrasound which generates an optical pulse. When the ultrasonic pulse reaches the next layer of the sample, it produces another optical pulse, which is similar to the temporal pulse from the previous layer but with a phase delay due to the time taken for the ultrasonic pulse to propagate between the two consecutive layers. The speed of sound v a in water and in gel phantoms is approximately 1500m/s at room temperature. The speed of light v in such v c n media is expressed as, where c 3x10 8 m/s is the light velocity in vacuum, and n 1.33 is the refractive index of water or tissue. We assume that the US is focused immediately in front of the fluorescent target, on the same side as the laser, as this is consistent with our experiment (section 3) and other work 8 which demonstrates that the largest modulation depth is obtained with this configuration. As the speed of light is much higher than the speed of sound, the time taken for the modulated light to reach the fluorescent region (placed next to the US focal region) is neglected. Hence, the phase difference between the pulses from each layer depends only on the transit time of the ultrasonic pulse. Pulsed excitation light (to the fluorescent target) may be expressed as a summation of many phase shifted optical pulses.

Given that the US column is composed of many such layers, each of width z=v a t, the detected pulsed fluorescence light can be written as, I fluor ( t) m j 1 P( z). O( t j t) where z = v a jδt, Δt is a time delay of the acoustic field, related to the number of steps m along the US column. O(t) is an optical pulse from a given layer whose temporal profile is imposed by the ultrasonic excitation pulse. The profile P(z) represents the optical intensity distribution along the acoustic axis as a result of the combined acoustical and optical characteristics along the ultrasonic column. In a pulsed/tone burst USMFT experiment, as the US is focussed immediately in front of the target, the length of the target along the acoustic axis can be considered as an aperture placed at the same position. One can therefore consider the profile P(z) that contributes to the detected fluorescent signal as a combination of the acoustic and optical profiles and the size of the fluorescent target. We propose a simple expression relating the optical and acoustic properties to the profile P(z) which can be expressed as, (8) P( z) P ( z). P ( z). P ( z) us ex where P us (z) is the axial pressure profile of the US, P ex (z) the scattered light intensity profile along the ultrasonic column, and P fluor (z) is the fluorescent profile. In this simple model, we assume that the profiles can be treated separately although inevitably there will be some dependence. To demonstrate the trends of experiments, such as the relationship between object size and acoustic wavelength, this model has been shown to be reliable 7. In a simplified form, P us (z) and P ex (z) can be represented as a Gaussian distribution as a focused US transducer is usually used, and a narrow light beam illuminates the scattering medium. Profile P fluor (z) is related to the fluorophore distribution in the target. For example, in figure 2, the fluorescent target is represented as a top-hat profile. fluor (9) U/S column Fluorescent target and its profile Photodetector Pulsed USMFT signal Scattered light 1 U/S pulse travelling direction 0 Figure 2 A pulsed USMFT model showing a target with a flat fluorescent profile 3. EXPERIMENT SET UP 3.1 Pulse Inversion A system for performing second harmonic ultrasound modulated optical tomography is demonstrated which incorporates both pulsed optical illumination and acoustic excitation (figure 3). A function generator (Tektronix AFG3252) generates

the signals to drive both the ultrasound transducer (Olympus A304S, 2.25 MHz) via an RF amplifier (ENI A300) and the laser diode (λ = 638 nm, P = 40 mw). A 10 mm diameter aperture is placed 35 mm behind the scattering medium and 80 mm in front of the CCD camera (Hamamatsu ORCA C4742-95-12ERG) to control the speckle size. The ultrasound signal is a sinusoid pulse tone burst with interleaved inversed pulses. After the delay due to the ultrasound pulse propagation time to the focus ( in this case), the laser is strobed for a time ( ) which is much shorter than the ultrasound period (440ns). The exposure time of the camera = 204 ms to ensure that sufficient light is detected. With this configuration, each camera frame detects thousands of pairs of optical signals modulated by inverted and non-inverted acoustic pulses (interval time between pulses ). Figure 3 Pulse inversion USMOT system. A function generator drives both the ultrasound transducer (via a power amplifier) and the laser. The synchronization of the light and ultrasound drive signals means that lock-in detection is implemented at each pixel of the camera. The scattering sample is a 90 mm wide (x), 50 mm high (y) and 16 mm thick (z) agarose gel mixed with 1.6µm diameter microspheres (scattering coefficient, anisotropy factor. An absorbing half plane is embedded at the mid-plane of the gel to obtain the edge response function. The gel is placed inside a water tank which is scanned in the x direction with a 0.2 mm step size. At each step, the fundamental frequency, second harmonic modulated and average light intensity are averaged four times. 3.2 Ultrasound modulated fluorescence tomography The pulsed USFMT experimental setup is shown in figure 4. A collimated laser (λ = 632.8nm, P = 20mW) illuminates the sample and a photomultiplier tube (PMT, Hamamatsu H5783-20) detects the scattered light emerging from the sample. Filters from a fluorescent filter kit (Edmunds Optics NT67-010) are used as an excitation filter (604nm 644nm) and an emission filter (672nm-712nm). A signal generator (Tektronix AFG3022B) and an RF power amplifier (Amplifier Research 150A100B) drive a focussed 1MHz US transducer (Olympus Panametrics V314 NDT). The US is focused at a position close to, but not at the fluorescent object. A 15cm x 10cm x 12cm (XYZ) water tank sits on a computer controlled XYZ motorized stage (Standa 8MT175-50). The signal from the PMT is fed into an amplifier, before going to an oscilloscope (Tektronix TDS2024B 8-bit ADC) and subsequent storage on a PC. The fluorescent target contains Alexa633 fluorophore (Invitrogen concanavalin A, Alexa Fluor 633 conjugate).

Laser Excitation filter and 2D adjustable slits Water tank Fluorescent target Driver U/S transducer Emission filter Motorized stages Photodetector Figure 4 USMFT Experimental setup z x Amplifier + Oscilloscope y 6mm 16mm z Laser source US focus x y Fluorescent tube Figure 5 The fluorescent target is a 10mm long conical tube with the longest diameter of 1.8mm and the shortest diameter of 0.4mm which is embedded in a (22mm x 16mm x 30mm) (XYZ) scattering agarose gel containing polystyrene microspheres (1.6µm diameter, µ s ~12cm -1 ). The system is used to investigate the effect that object length along the acoustic axis has on the modulation depth of the detected signal. The fluorescent target (figure 5) is a conical tube with the largest diameter of 1.8mm and the smallest diameter of 0.4mm which is embedded in a (X=22mm, Y=16mm, Z=30mm) scattering agarose gel containing polystyrene microspheres (1.6µm diameter, µ s ~15cm -1 ). Each detected pulse is averaged by the maximum x128 setting on the oscilloscope and further averaging (x400) is carried out on the PC. The same detection system is also used to carry out an experiment where the light is generated within the medium without requiring an external light source. A commercially available chemi-luminescent target (a glow stick ) is embedded within a relatively weakly scattering sample (figure 6). The object is masked with a 2mm square aperture to produce a relatively small light source within the sample.

20mm Glow stick with 2mm mask 40mm US focus PMT Scattering medium (µ s ~4cm -1 ) Figure 6 Imaging configuration (no external light source) for imaging a chemi-luminescent target 4.1 Pulse Inversion 4. RESULTS X (mm) Figure 7 Line scans of an absorbing edge embedded at the mid-plane of a scattering medium using DC, 1 st and 2 nd harmonic ultrasound modulated light The edge response function shown in figure 7 was obtained in the experiment and then used to estimate the line spread function by a least squares fitting 9. The full width half maximum of these line spread functions are 4.02 mm for the fundamental, 2.43 mm for the second harmonic and 9.26 mm for the DC light. The line spread functions for the US alone have also been obtained by scanning a needle hydrophone (Precision Acoustics, 0.2 mm diameter) through the US focus. The full width half maximum of these scans (not shown here) are 1.83 mm for the fundamental and 1.12 mm for the second harmonic which are lower than those obtained by USMOT. 4.2 Ultrasound modulated fluorescence tomography For the scan of the conical fluorescent object (figure 8), the DC light level gradually increases with increasing x and then decreases as the US focus reaches the edge of the object and moves away. The US modulated fluorescence signal depends on the size of the fluorescent object and the modulation frequency of the US. The number and position of the peaks show good agreement between experiment and simulation.

Normalized magnitude Normalized signal 1 0.8 AC DC 0.6 0.4 0.2 0 2 4 6 8 10 12 14 16 18 x (mm) (a) 1 0.8 AC DC 0.6 0.4 0.2 0 2 4 6 8 10 12 14 16 18 x (mm) (b) Figure 8 Ultrasound modulated and DC light line scans of a fluorescent target embedded in a scattering medium a) simulated b) experiment 4.3 Ultrasound modulated chemi- and bio-luminescence tomography A line-scan of an edge obtained using by modulating the light emerging from the chemi-luminescent target is shown in figure 9. In comparison to the DC light response the modulated light response is much sharper, demonstrating an improvement in resolution obtained through ultrasound modulated optical tomography.

normalized signal 1 0.8 0.6 0.4 0.2 DC AC 0 2 4 6 8 10 12 x (mm) Figure 9 Edge response from a chemi-luminescent target obtained by modulating the light emerging from the target using ultrasound. 5. DISCUSSION AND CONCLUSIONS 5.1 Pulse Inversion One of the possible causes of the lower resolution in USMOT is due to multiple scattering of light in the region of the US focus. This reduces the effective USMOT focal zone and also produces second harmonic signals due to interference effects as suggested in 3. Another potential problem is that the algorithm, like all speckle based methods is susceptible to the effects of speckle decorrelation between consecutive frames. Although this is reduced for the cases of individual pulses as they are summed within the same frame, the method still requires a shift in the US phase and detection in the next frame. The axial resolution can be improved further by reducing the length of the US pulse as any spectral overlap between fundamental and second harmonic can be overcome using the pulse inversion method. Such improvements cannot be achieved using a filtering approach such as lock-in detection. Future work will involve shortening the pulse length to improve axial resolution and the use of microbubbles which have been used to improve image contrast in conventional non-linear acoustic imaging. Pulse inversion parallel detection USMOT is introduced in this paper which demonstrates a reduction in the effective line spread function from 4.02mm to 2.43mm. Previous second harmonic USMOT imaging [Error! Reference source not found.] has used continuous wave US to achieve a comparable improvement in lateral resolution. However, the use of pulsed US in this case allows axial resolution to be achieved by time gating along the ultrasound propagation axis and is also more compatible with clinical diagnostic US. 5.2 Ultrasound modulated fluorescence tomography The effect of fluorescent object size and US frequency on the images obtained in pulsed USFMT has been investigated. All systems to date have implemented CW US and so the effect of using pulsed US has not yet been considered. Pulsed

US offers advantages in terms of allowing high peak pressures while remaining within safety limits and also enabling axial resolution along the acoustic axis to be obtained using time gating. A range of results (to be published at a later date) show that the USMFT signal depends upon the size of the fluorescent target and the frequency of the US. If the object is of the order of an acoustic wavelength then the US modulated fluorescent pulses that propagate to the detector are likely to cancel at the detector plane and produce a signal of relatively low amplitude. If the object is an odd number of half wavelengths wide then the pulses will produce a larger detected signal due to more constructive interference occurring. The simple convolution model employed is able predict the detected signal. The simulation is simple because it neglects the effects of optical speckle and treats the fluorescent object as a planar object with different widths. More accurate quantitative results could be obtained by implementing a Monte Carlo simulation of light propagation 1, however, the simulation allows the trends of USMFT to be predicted. SNR is still very low and the acquisition time is high. This is a disadvantage as the safety threshold cannot be exceeded in practical applications and also photo bleaching of the fluorophore needs to be considered. In order to increase the SNR, micro-bubbles could be used to produce a bigger change in the optical properties within the insonified region. Nevertheless, the first pulsed USMFT system has been demonstrated and the effects of fluorescent target size and acoustic frequency on the detected signals have been investigated. As the acoustic pulse propagates through the medium, fluorescent pulses are generated which propagate to the detector. Depending on the size of the object and the acoustic frequency, these pulses can sum either constructively or destructively at the detector. When the object is an integer number of acoustic wavelengths wide the pulses sum destructively. When the object is an odd number of half wavelengths wide the pulses sum constructively and produce a comparatively higher signal. This has been demonstrated experimentally and using a simple simulation of the pulsed USMFT process. This effect needs to be taken into account as it will produce image artefacts. 5.3 Ultrasound modulated chemi- and bio-luminescence tomography Photoacoustic tomography has provided superior performance to USMOT in absorption imaging. There have also been applications of photoacoustic tomography in fluorescence imaging 10 although this relates the absorption of light at the fluorescence event to the fluorescent signal. Ultrasound modulated optical tomography offers the potential to directly modulate the fluorescence signal. The application of ultrasound modulated optical tomography to imaging chemi- and bio-luminescent targets offers an interesting challenge for USMOT as no external light source is required and photoacoustic tomography cannot be applied. We have demonstrated a simple experiment in a relatively weakly scattering medium which shows that the imaging resolution can be improved by modulating the light emitted from the target. In practical situations both the modulation depth and DC light levels will be much lower than in this experiment which will provide a significant challenge. Low noise detection and novel contrast particles are likely to play a role in its successful implementation. ACKNOWLEDGEMENTS This work was supported by the Biotechnology and Biological Sciences Research Council (BBSRC) UK (BB/F004826/1 and BB/F004923/1). HR was supported by China Scholarship Council. We thank Dr D He for useful discussions. REFERENCES [1] Q. Liu, S. Norton, and T. Vo-Dinh, "Modeling of nonphase mechanisms in ultrasonic modulation of light propagation," Appl. Opt. 47, 3619-3630 (2008) [2] F.A. Duck, Nonlinear Acoustics In Diagnostic Ultrasound, Ultrasound in Med. & Biol. 28, 1-18 (2002). [3] J. Selb, L. Pottier, and A. C. Boccara, Nonlinear effects in acousto-optic imaging, Opt. Lett. 27, 918-920 (2002). [4] D. H. Simpson, C. T. Chin, and P. N. Burns, Pulse Inversion Doppler: A New Method for Detecting Nonlinear Echoes from Microbubbles Contrast Agents, IEEE Transaction On Ultrasound, Ferroelectrics and Frequency Control, 46, 372-382 (1999).

[5] S. Leveque, A. C. Boccara, M. Lebec, and H. Saint-Jalmes, Ultrasonic tagging of photon paths in scattering media: parallel speckle modulation processing, Opt. Lett. 24, 181 183 (1999). [6] J. Li and L. V. Wang, Methods for parallel detection based ultrasound modulated optical tomography, Appl. Opt. 41, 2079-2084 (2002). [7] NT Huynh, D He, B R Hayes-Gill, J A Crowe, J G Walker, M L Mather, F RAJ Rose, N G Parker, M JW Povey, S P Morgan, Application of a maximum likelihood algorithm to ultrasound modulated optical tomography, accepted for publication in Journal of Biomedical Optics (2012). [8] B. Yuan, J. Gamelin and Q. Zhu, Mechanisms of the ultrasonic modulation of fluorescence in turbid media, Appl. Phys.104, 103102 (2008) [9] S. M. Bentzen, Evaluation of the spatial resolution of a CT scanner by direct analysis of the edge response function, Med. Phys. 10, 579-581 (1983). [10] D. Razansky, M. Distel, C. Vinegoni, R. Ma, N. Perrimon, R.W. Köster, V. Ntziachristos, Multispectral opto-acoustic tomography of deep-seated fluorescent proteins in vivo, Nature Photonics 3, 412-417 (2009).