All-optical ultrasound is an emerging imaging modality that shows great promise for biomedical imaging. With this modality, optically absorbing structures convert pulsed or modulated light into ultrasound via the photoacoustic effect. Compared with conventional electronic systems comprising piezoelectric or capacitive ultrasound transducer elements, optical transducers have been shown to yield similiar or higher pressures and bandwidths.

The researchers present a new paradigm for all-optical ultrasound imaging that enables real-time, videorate 2D imaging at a frame rate of 15Hz.

Whereas electronic transducer elements are defined through electrode patterning or physical separation, in this work optical ultrasound sources are shaped and positioned by means of optical confinement achieved using lenses and scanning optics.

This allowed the synthesis of dynamically reconfigurable source arrays of arbitrary geometry, thus enabling 1D source arrays exhibiting non-uniform source spacing and reduced image artefacts. In addition, a combination of innovations was used to achieve a sensitivity suitable for imaging weak reflections from deep within biological tissue in real-time. Among these innovations were a centimetre-scale ultrasound generation surface comprising carbon nanotubes and an elastomeric polymer that was suspended in free space, eccentric optical excitation achieved through the use of cylindrical optics and a high-finesse Fabry-Pérot cavity for ultrasound reception. The sensitivity of the system was demonstrated on different biological tissues, with its real-time dynamic capabilities demonstrated through the imaging of pulsating blood vessel sections and acoustic contrast agent injections.

Potential to overcome all possible limitations

In this work, an all-optical ultrasound imaging paradigm is presented that allows real-time, video-rate 2D imaging of soft tissue. This was achieved through the use of an optical ultrasound generator, a highly sensitive fibre-optic receiver, eccentric illumination resulting in acoustic sources exhibiting constrained acoustic radiation patterns, a modest excitation pulse repetition rate and pulse energy, fast sequential scanning of a source aperture using a galvo mirror exhibiting low inertia, source arrays exhibiting nonuniform source spacing and partially overlapping sources.

A wire phantom was used to determine the resolution of the system (axial: 75μm, lateral: 100μm), and the penetration depth (up to at least 6mm) and dynamic range (up to 30dB) were determined using tissue samples. In addition, the high frame rate (15Hz) was used to capture the dynamics of a pulsating vessel phantom and the injection of an acoustic contrast agent.

All-optical ultrasound has the potential to overcome several fundamental limitations of its electronic counterparts. Electronic ultrasound transducers typically derive their sensitivity from mechanical resonance at a fixed frequency and bandwidth, resulting in a fixed spatial resolution and penetration depth. In addition, electronic ultrasound imaging probes typically comprise transducer elements that are spatially fixed and arranged in a periodic array, which can result in image artefacts.

As a result, electronic imaging probes are optimised for specific applications, and typically multiple probes are required for versatility. In contrast, optical ultrasound sources do not rely on resonance, and hence can be tuned to either highresolution imaging or large imaging depth through temporal modulation of the excitation light. In addition, the position and spatial confinement of optically generated ultrasound sources can be varied dynamically using optical methods to avoid imaging artefacts and tailor the source aperture to a wider range of applications.

Compared with conventional ultrasound imaging arrays comprising piezoelectric or capacitive transducers, the all-optical ultrasound set-up achieved a lower dynamic range of 30dB that was limited by comparatively high image artefact levels and relatively low signal-to-noise ratios.

Improve the dynamic range

The image artefact levels were due to the use of a low number of acoustic sources and a single receiver, combined with the delay-and-sum image reconstruction algorithm, and could be reduced in various ways. First, multiple fibre-optic ultrasound receivers can be distributed across the aperture.

In this work, an all-optical ultrasound imaging paradigm is presented that allows real-time, video-rate 2D imaging of soft tissue.

This is similiar to conventional electronic linear arrays comprising a multitude of electronic transducers that can each transmit and receive acoustic signals. This way, the back-scattered acoustic field can be recorded in multiple locations in parallel, resulting in the detection of a larger fraction of the back-scattered energy at the same frame rate. This will decrease the artefact levels and alleviate the observed limitedview artefacts, at the expense of a significant increase in complexity.

Second, subsequent ultrasound sources can exhibit spatial overlap without exhibiting cross-talk, and an arbitrary number of sources can be positioned within the aperture. Consequently, the image quality could be improved by using more source locations. However, currently the image acquisition and reconstruction rate is limited by the computational cost of the reconstruction and the inertia of the galvo mirror, and hence faster mirrors (for instance resonant galvo mirrors) and parallelised reconstruction algorithms would be required to accommodate the higher pulse repetition rate required to maintain high frame rates.

Third, the artefact levels could be further reduced by implementing different source density apodisation (SDA) schemes. While the schemes studied in this work achieved a significant improvement in image quality, alternative SDA schemes might yield even further improvements. Further research is required to determine the most effective SDA schemes.

Finally, different image reconstruction algorithms that better exploit the spatial coherence of consecutive A-scans, such as delay, multiply and sum (DMAS) or short-lag spatial coherence (SLSC), might result in reduced artefact levels.

The dynamic range of the images could be improved further by increasing the signal-to-noise ratio of the data.

It has previously been shown that the presence of a rigid backing can improve the efficiency of optical ultrasound sources, thus improving the A-scan signal-to-noise ratio. However, the presence of a rigid backing introduced spurious ringing artefacts that decreased the acoustic bandwidth and the image quality; hence, in this work, an ultrasound-generating membrane was suspended in free-space.

Alternatively, increasing the optical fluence would result in larger pressures to be generated in the ultrasoundgenerating membrane. However, as the fluence used in this work (38mJ/cm2) approached the damage threshold of the membrane (approximately 175mJ/cm2), the pressure increase achievable using a pulsed light source is limited.

However, alternative optical excitation schemes, such as coded excitations or temporally modulated pulses, could be employed to deliver more optical energy over an extended period of time without exceeding the optical damage threshold. The optical ultrasound sources distributed across the source aperture were found to consistently generate a wide bandwidth ranging 4–31MHz. However, the lateral extent of the sources (224 ±53μm) was too large to fully exploit this bandwidth: the sources were highly directional for higher frequencies, and thus only the lower frequencies contributed to the images. Hence, in this work, signals were band-pass filtered 2–15MHz, which limited the achieved resolution. To use the generated bandwidth, acoustic sources of reduced lateral extent should be used.

These can be achieved through a change in optics, for instance using a lens with a shorter focal length. However, to avoid damaging the ultrasound-generating membrane, the pulse energy should be decreased when decreasing the optical spot size, thus reducing the generated pressure amplitude and signal-to-noise ratio.

In addition, in the set-up presented in this work, the lens focuses excitation light onto the ultrasound-generating membrane through the transparent wall of a water bath and a layer of water, resulting in a wall-reflection artefact in the images at a depth of approximately 3.5cm. Decreasing the focal length reduces the maximum distance between the water bath wall and the sound-generating membrane, and thus will limit the maximum imaging depth. Ultimately, optical and acoustical scattering within the nano-composite membrane will limit the minimum lateral extent of the acoustic sources.

In a 2D all-optical ultrasound imaging paradigm, the lateral field of view was limited by the lateral extent of the acoustical aperture (15.5mm), which in turn was limited by the exit pupil of the optics. Thus, through a change in optics a wider source aperture could be achieved. However, to fully exploit a wider aperture, the optical alignment needs to be improved to avoid elevational offset of the sources. In addition, the fibre-optic acoustic receiver was positioned within the image plane and placed in front of the sound-generating membrane. As a result, ultrasound reflected off the generator membrane resulted in additional ghost images. These spurious reflections could be limited by positioning the receiver in the plane of the sound-generating membrane, at the expense of a reduced acoustical aperture due to optical shadowing. Alternatively, the ringing could be removed through source deconvolution.

The galvo mirror used in this work actually comprised two orthogonal angle-adjustable mirrors, of which one was kept fixed. However, in principle this galvo mirror allows arbitrary scanning of excitation light in 2D. Using the presented centimetre-scale nanocomposite membrane, a 2D acoustic source aperture is hence readily generated that could be used for 3D imaging. If the cylindrical lens were replaced with a spherical lens, an isotropic resolution of approximately 100μm could be achieved using the same imaging paradigm, at a frame rate of up to 1Hz. However, as the resulting optical focus is much smaller, the pulse energy of the excitation light would need to be reduced to avoid optical damage to the ultrasound-generating membrane.

For future configurations

In future, water surrounding the probe could therefore be replaced with acoustic coupling gel applied to only the surface of the imaging target, as is customary in routine clinical practice. Alternatively, a bundle of optical waveguides could be used to obtain a flexible imaging probe allowing free-hand operation similiar to current clinical ultrasound probes, or miniaturised for applications.

As the electrical and metal components can be spatially separated from the acoustic sources and receiver, an all-optical ultrasound probe can be made insensitive to electromagnetic interference. This will enable the application of 2D or 3D ultrasound imaging in electromagnetically harsh or sensitive contexts, such as real-time monitoring of radio frequency ablation, electrophysiology and neuromodulation.

In addition, the application of alloptical ultrasound imaging within an MRI scanner will provide clinicians with concurrent, multiscale imaging yielding supplementary information during MRIguided interventional procedures, such as high-precision instrument localisation, video-rate real-time monitoring of MRI-guided highintensity focussed ultrasound (HIFU) treatment and visualising microbubble mediated drug delivery.

Finally, the adaptability of the frequency, bandwidth and geometry of optical ultrasound sources will allow seamless tailoring of an imaging system to the imaging target to optimise the image quality, frame rate and resolution.