Electrically tunable lenses – eliminating mechanical axial movements during high-speed 3D live imaging

Several key advances in cell biology are linked to significant innovations in microscopy (Morris and Payne, 2019). Over the past twenty years, high-resolution live-cell imaging methods have steadily improved. While it has become routine to acquire high-resolution three-dimensional (3D) time-lapse movies for hours without inflicting damage to DNA, cells or tissues (Draviam et al., 2006; Hart et al., 2021; Yue et al., 2020), it is clear that 3D movies can miss out on fast intracellular events – such as the contribution of astral microtubule dynamics to 3D spindle movements – due to the time lost during axial scanning (Follain et al., 2017; Liu et al., 2015; Yamashita et al., 2020). In such cases, high-speed image acquisition in the axial direction is required to fully encapsulate the 3D movement of intracellular structures. However, incorporating the third dimension (i.e. the z-axis), without losing out on temporal resolution, remains a significant challenge in high-resolution live-cell microscopy.

High-speed 3D imaging in the millisecond regime is a much-needed tool to tackle the challenge of imaging across scales. For example, high-speed axial scanning would allow robust 3D tracking of cytoskeletal dynamics near the immunological synapse in non-adherent and floating immune cells (Allam et al., 2021; Kumari et al., 2020) (Fig. 1A). Another example of a use for dynamic imaging across scales is the interrogation of mechanosensing by nanoscale adhesomes in cells that migrate within tissues (reviewed in Michael and Parsons, 2020) (Fig. 1A). High-speed 3D imaging can also support the visualisation of two or more dynamic structures near-simultaneously, for example, exploring how actin clouds and astral microtubules rotate the bulky mitotic spindle to position newborn cells within tissues (reviewed in Chin et al., 2014; Kulukian and Fuchs, 2013) (Fig. 1A). The overarching requirement to study the above examples is the ability to perform both high-speed and high-resolution live imaging while maintaining a delicate balance between exposure time, illumination intensity and duration of imaging, since the live samples are at constant risk of experiencing phototoxicity and photobleaching (Skylaki et al., 2016).

In order to generate the 3D structure of a sample, light microscopes heavily rely on the sequential capture of two-dimensional (2D) images at various focal planes and the subsequent utilisation of computational tools to reconstruct the corresponding 3D structure (Agard, 1984; Girkin, 2015). These techniques have been improving in terms of both axial scanning speed and reconstruction algorithms (O'Holleran and Shaw, 2014; Schniete et al., 2018). However, these techniques still fundamentally rely on mechanical movements of the microscope objective and stage in order to enable axial scanning. These mechanical movements, which are necessary for sequential switching between the focal planes of the sample, not only introduce inertia, hence affecting focusing speed, but can also damage the sample of interest.

In the case of a dividing cell, microtubule dynamics can reach speeds of up to 250 nm/s (Tamura et al., 2015). To image these highly dynamic events in 3D, high-speed imaging methods that allow fast transitions of the focal plane within the millisecond regime are crucial (Nakai et al., 2015). Currently, the state-of-the-art approach to combine high-speed and high-resolution imaging is lattice light-sheet microscopy, where a spatial light modulator projects extremely thin 2D lattice-shaped light sheets to scan the sample plane-by-plane at extremely high speeds and with negligible photobleaching to ultimately generate a 3D image (Chen et al., 2014). This technique has been applied to track microtubule growth dynamics in 3D within the mitotic spindle at high spatial and temporal resolutions (single-colour images collected every 0.755 s; Yamashita et al., 2015). More recently, lattice light-sheet microscopy has been used to investigate chromosome segregation dynamics in the spindle midzone of HeLa and RPE cells; here, whole-cell dual-colour volumes consisting of 101 frames with 300 nm step sizes were imaged at a rate of 20 volumes/min (Pamula et al., 2019). However, assembling and operating lattice light-sheet microscope components requires a high level of expertise. One of the latest iterations of the technique employs a complex setup of components, including two lasers, three objectives, four cameras, seven galvanometers, twenty-nine mirrors and thirty-five lenses (Liu et al., 2018), all of which require stable alignment for the generation of lattice-shaped light sheets. Commercially available lattice light-sheet microscopes, even if they are very expensive, have helped researchers circumvent the complex assembly process. However, the operation of these sophisticated microscopes lacks flexibility due to the highly restrictive manufacturer-configured orientation of light.

Here, we review an innovative and cost-effective approach that is increasingly being explored in the biosciences to achieve high-speed axial scans – the electrically tunable lens (ETL). ETLs are liquid lenses characterised by their on-demand shape-changing capability, which is regulated through electrical current (Blum et al., 2011). Notably, a micrometre change in their radius can generate the same optical effect as moving a traditional lens by several centimetres (Blum et al., 2011). We begin by explaining the core principles of an ETL and subsequently describe its benefits over ‘traditional’ microscopy setups, highlighting recent use of ETLs to achieve 3D imaging at high temporal and spatial resolutions. We also present some of the solutions being developed to offset the challenges an ETL brings to microscopy setups – mainly the increased occurrence of aberrations that result in blurriness or image noise, consequently obstructing the accurate representation of the imaged object. We conclude by outlining areas of technology and discovery where the future impact of ETL technology is likely to be particularly beneficial.

ETLs, also known as liquid lenses, allow microscopes to overcome the limitations imposed by the mechanical movements required for changing the focal plane. Since ETLs enable shifts in focus through changes in their shape – within milliseconds – in response to electrical current (Fig. 1B,C), they can be utilised for high-speed imaging. One of the first descriptions of a lens that could change its refractive index on demand came in the form of a plano-concave liquid crystal lens cell (Sato, 1979). This liquid crystal lens allowed the modulation of its refractive index into two different configurations through the application of voltage, based on the direction of polarisation of the incident light (Sato, 1979). However, this type of tunable lens suffers from a low response time because of the thickness of the liquid crystal layer (∼460 μm).

A second approach to creating shape-changing lenses involves utilisation of the electrowetting effect. Electrowetting refers to the change in contact angle between a metal solid and a conductive liquid (electrolyte) interface when sufficient voltage is applied (Mugele and Baret, 2005). The application of this phenomenon for the design of a variable focal lens was first demonstrated in 2000; here, the curvature of the lens was created at the interface of two equally dense non-mixing liquids with different refractive indices (oil and water) (Berge and Peseux, 2000). Change in the curvature of the lens was mediated by applying voltage to a metal electrode that was directly connected to the compartment containing the conductive liquid (water) (Berge and Peseux, 2000). Electrowetting-based tunable lenses can achieve faster response times, lower power consumption and a more compact design than their liquid crystal predecessors. However, this type of lens is fundamentally limited by a small aperture size (5 mm) – a consequence inherently created by the electrowetting effect, as the capillary forces generated are not sufficiently strong to support shape changes in a larger lens (Berge and Peseux, 2000). A small aperture size reduces both the spatial resolution and the depth of field of the system.

One of the latest iterations of a lens with shape-changing capabilities, widely known as an ETL (and the main focus of this Review), exhibits a central core consisting of a container filled with an optical fluid and sealed with an elastic thin polymer membrane (Blum et al., 2011) (Fig. 1B). Here, an electromagnetic actuator ring surrounding the lens exerts pressure on the highly elastic container, which is capable of changing the curvature of the lens. The electrical current flowing through the coil of the actuator ring determines the exact amount of pressure exerted, thus controlling the optical power of the lens (Blum et al., 2011).

The shape-changing capability of an ETL provides a significant advantage as it can produce a change in focal length much faster than non-ETL arrangements (Nakai et al., 2015). Due to their compact design, relatively low cost and wide availability, ETLs have been used in combination with a variety of systems, including those with an ophthalmological purpose (e.g. optical coherence tomography platforms; Grulkowski et al., 2018; Tao et al., 2014), endoscopes (Zou et al., 2015), laparoscopes (Qin and Hua, 2016; Volpi et al., 2017), profilometry systems utilising projectors to enable machine vision (Hu et al., 2020; Iwai et al., 2015; Zhong et al., 2020), satellite laser communications (Fogle et al., 2020), cameras (Guo et al., 2017; Miau et al., 2013), digital holography (Sanz et al., 2020; Wang et al., 2017), stealth laser wafer dicing (Lee et al., 2020) and head-mounted 3D displays for virtual- or augmented-reality applications (Chang et al., 2019; Konrad et al., 2016, 2017; Llull et al., 2015; Rathinavel et al., 2018; Shen and Javidi, 2018). The broad applicability and commercial availability of ETLs have favoured their incorporation into microscope systems for high-speed 3D imaging of light-sensitive living specimens, as discussed next.

The utilisation of ETLs has seen an overall rapid increase within the past two decades (Fig. 2). For bioscience research specifically, the use of ETLs has begun to steadily increase in the past decade, supporting a variety of microscopy techniques (Fig. 2).

The ETL saw its debut into bioscience research through its integration into a custom two-photon microscope for the study of Ca²⁺ activity in neuronal cell populations of the mouse neocortex (Grewe et al., 2011). Here, an ETL was used as a z-scanning device, allowing rapid axial shifts between different focal planes within 15 ms along an axial scan range of 700 μm. By using repeated axial shifts between two planes at different focal depths and subsequent fast refocusing, the authors were able to collate the activity signals of 40 neurons at a sampling rate of 21.6 Hz (Grewe et al., 2011). Therefore, the integration of an ETL to their setup allowed the elimination of inertia caused by the moving objective, in turn permitting 3D video-rate monitoring of distinct neuronal regions of interest within a large volume (Grewe et al., 2011). As the ETL model used in the above study could not reach a concave shape, an offset lens was necessary for the resulting setup to also reach negative focal lengths. Here, a sufficient signal-to-noise ratio was maintained for the axial focusing range through a combination of an ETL, an offset lens and a microscope objective. The fast focusing speed obtained was achievable due to the rapid refocusing response time (15 ms) and relatively large aperture size of the ETL, which conferred high-speed 3D scanning capabilities to a two-photon microscope setup (Grewe et al., 2011). The demonstration of the above configuration paved the way for rapid in vivo and ex vivo volumetric imaging of the activity of neuronal populations within the mouse visual cortex and retina (Table 1) (Han et al., 2019; Zhao et al., 2020).

ETLs have also been integrated into a selective-plane illumination microscope (SPIM) – i.e. light-sheet microscopy – to investigate the beating heart of zebrafish (Fahrbach et al., 2013). This system allowed the rapid imaging of 17 z-planes within the zebrafish beating heart at a rate of 510 frames per second, thereby enabling the tracking of shape changes and movements of multiple cells across a large volume. The ETL enabled volumetric imaging without the requirement of mechanical movement, which limits acquisition speed and is harmful to the sample. The biggest advantage of using an ETL was its high degree of flexibility, as the number of images, their spacing, size and acquisition speed could all be varied during the experiment. The study also documented the introduction of minor aberrations, specifically astigmatism and coma (Fig. 3; discussed further below), close to the borders of the images (Fahrbach et al., 2013). Depending on the image quality required for a specific imaging application, increasing aberrations at the borders of the image could limit the relatively usable field-of-view and scan range (Fahrbach et al., 2013). In this case, the authors resolved the above problem by under-sampling the point spread function (PSF), resulting in an increase in usable field-of-view (Fahrbach et al., 2013).

Moreover, integrating an ETL into a reflectance confocal microscope (RCM) has enabled the axial scanning of ex vivo oral epithelial tissue and in vivo skin tissue (Jabbour et al., 2014). An RCM allows a horizontal view of the skin and can be used in dermatoscopic diagnosis to distinguish between benign and malignant neoplasms (Shahriari et al., 2021). The authors were able to characterise axial and lateral resolutions over a scanning range of 255 μm at a scan rate of 0.2 Hz, which could potentially increase to 100 Hz as both response and settling times of the ETL can be less than 10 ms (Jabbour et al., 2014). This study highlights the flexibility the use of an ETL provides in terms of controlling the axial focal position and the scanning speed.

An ETL has also been incorporated into a widefield microscope for the study of microtubules and microtubule-end-associated protein complexes as part of the dynamic mitotic spindle within HeLa cells (Nakai et al., 2015). Here, a spatial resolution of less than 40 nm (versus 286 nm without the ETL) and a temporal resolution within 10 ms was achieved, enabling the visualisation and tracking of microtubule ends that move at an approximate rate of 250 nm/s (Nakai et al., 2015). By utilising three low-pass filters, the authors optimised the frequency of the driving signal, subsequently shortening the stabilisation time of the ETL and enabling faster focal transitions (see Table 1 for further examples of ETL use in the biosciences).

In summary, as an innovative 3D-imaging technology introduced to live-cell microscopy a decade ago, ETLs have undergone multiple improvements to solidify their place in enabling high-speed 3D studies of several dynamic cellular processes. ETLs appear to be mainly implemented in two-photon microscopy systems due to the inherent nonlinearity of the excitation process and relatively low resolution demands (Table 1). However, the ability of ETLs to rapidly shift in focal length expands high-speed imaging opportunities, while their compatibility with a variety of microscopes permits use outside of sophisticated research facilities, including in ecologically diverse environments, teaching laboratories and hospitals.

As noted above, the utilisation of ETLs may introduce aberrations as a result of the shape-changing capability of the lens, which is controlled by varying electrical currents and is a procedure necessary for achieving rapid changes in the focal plane (Fahrbach et al., 2013). Aberration is a term used to describe the deviation of light rays away from the ideal focus, leading to the inability of a microscope to form an exact representation of all the point sources of light within the imaged object (Fig. 3A,B) (Booth and Patton, 2014; Sanderson, 2019). The image of a point-like object, as seen from a microscope, is described by the PSF – a concentric geometrical 3D pattern of diffracted light that is projected through the lens (Fig. 3B) (Airy, 1835). Named after the scientist who first explored the PSF, Sir George Biddell Airy, the concentric geometrical pattern is called the ‘Airy pattern’, while its centre is known as the ‘Airy disk’ (Fig. 3B).

In the context of an ETL, increased aberrations that are generated from the shape-changing ability of the lens can degrade the PSF and therefore decrease the quality of the image. Astigmatism, comatic and spherical aberrations have been observed when ETLs are used (Fig. 3A) (Fahrbach et al., 2013; Fuh et al., 2015; Yan et al., 2019; Ye et al., 2020). In astigmatism, light rays propagating on two perpendicular planes focus at different distances from the lens (i.e. they are misaligned), resulting in blurriness and elongation of the imaged object across the z-axis (Goodwin, 2013) (Fig. 3A,B). Comatic aberrations occur when light rays reach the lens at oblique angles, subsequently focusing at different lateral positions within the focal plane (Goodwin, 2013). The object within the image appears skewed or trailing, thereby displaying a characteristic tail-like blurriness (Fig. 3A,B). Spherical aberration occurs when light rays entering the centre of the spherical lens are not in the same focal plane as the light rays entering at the lens periphery (Goodwin, 2013). Images in this case appear defocused with a severely degraded signal-to-noise ratio (Fig. 3A,B). Spherical aberrations can be induced when imaging the sample through interfaces with a refractive index mismatch, but they can also be induced by the internal elements of the lens itself (Diel et al., 2020). Possessing a shape-changing capability makes ETLs inherently prone to spherical aberrations, consequently limiting performance, and in turn image quality, in high-numerical-aperture (NA) applications. Some of these aberrations are beginning to be resolved through optoelectronic (Vogt, 2020) and computational (Cumming and Gu, 2020) solutions (see below for further details).

Characterisation of the different aberrations that can arise while using an ETL has led to the proposal to use adaptive optics (AO) to reduce these errors (Fig. 3C) (Fuh et al., 2015). AO can correct wavefront distortions depending on the measurements obtained from wavefront sensors (typically a Shack–Hartmann sensor); some examples of AO devices include spatial light modulators and deformable mirrors (Booth, 2014). Wavefronts are the waves emitted and propagated from a distinct point source of the specimen. The concept of AO was first realised in terrestrial telescopes, as complex optical aberrations typically arise due to the difference in refractive index between the Earth's atmosphere and the lens of the telescope, consequently distorting the ideal spherical form of the emitted wavefronts (reviewed in Ji, 2017). AO have been increasingly used in biology for improving imaging depth, while maintaining a high spatial resolution through the correction of aberrations (Marx, 2017).

Accordingly, integrating an AO device, specifically a micro-electro-mechanical system (MEMS) deformable mirror (Bifano, 2011), into an ETL setup has been shown to result in an overall reduction of aberrations (Fuh et al., 2015). AO has also been recently integrated into ETL systems for biological imaging. With a custom-built two-photon microscope that combines an ETL for fast axial scanning, multi-frame super-resolution reconstruction (a super-resolution algorithm that generates a higher resolution output by combining many spatially-related low-resolution images) and sensorless AO, an improvement in axial resolution and an overall higher signal-to-noise ratio than conventional two-photon microscopy has been achieved in the imaging of mouse cerebral nerve cells in vitro and in vivo (Ye et al., 2020). Sensorless AO is an indirect wavefront-sensing technique where, despite the lack of a wavefront sensor, the aberrations can still be quantified (Hampson et al., 2020). Aberration correction by sensorless AO is guided by the quality of the acquired images and mathematical algorithms that characterise the different types of aberrations (Zernike polynomials; reviewed in Hampson et al., 2020). Any calculations obtained inform the adjustments that need to be taken by the adaptive device in order to fully optimise image quality. The adaptive device most often used in sensorless AO approaches is a spatial light modulator (Jesacher and Booth, 2013; Ye et al., 2020). By combining an ETL and sensorless AO, previously indistinguishable features, such as the processes of microglial cells along the z-axis, can be successfully resolved (Ye et al., 2020). In addition, this custom microscope system is able to track the 3D morphological changes of microglial cells over time, specifically the collapses of microglial processes along the axial direction (Ye et al., 2020).

Another recent advance has helped to tackle a different source of optical aberrations in ETL setups: focus drift during long periods of imaging. Focus drift is a term used to describe the failure of a microscope to retain focus of the same plane over long periods of time, whereby this lack of focus occurs independently of the sample (Kreft et al., 2005). Although this is a common problem for most microscopes, ETL setups primarily suffer from thermal drift (Blum et al., 2011). This type of drift is caused by excessive heating of the lens, specifically triggered by long periods of exposure of the ETL to the electrical current that is required to change its curvature. Excessive heating of the lens can result in thermal expansion, which has an unwanted effect on the lens radius. To avoid this issue, the use of temperature sensors placed close to the ETL has been suggested, in order to constantly monitor the lens temperature (Blum et al., 2011). Recent models of commercially available ETLs now employ built-in temperature sensors, which can be used in combination with specialised software drivers to control the focal length based on the ongoing temperature of the lens. Such compensations are possible because this type of thermal drift is reproducible (Blum et al., 2011).

A further challenge that has been recently addressed is the position of the ETL within the microscope system. In an example mentioned above, the ETL was mounted in combination with an offset lens on top of the microscope's objective lens (Grewe et al., 2011). The offset lens was required to achieve a tuning range that covered both negative and positive focal lengths. This straightforward implementation of the ETL conferred an extensive axial scanning range (700 μm), but the overall microscope system was not telecentric, owing to the ETL not coinciding with the aperture stop of the microscope objective (Grewe et al., 2011). Non-telecentric conditions can affect the magnification or field-of-view size as the focal plane is shifted across the z-axis (Grewe et al., 2011). This issue can be solved by positioning the combined ETL and offset lens assembly between two relay lenses at a conjugate of the back focal plane of the detection lens (Fahrbach et al., 2013). The conjugate of the back focal plane corresponds to the position where the image is formed within the optical path (Qu and Hu, 2019) – similar to the position of the retina within the human eye. This successfully creates a telecentric configuration (i.e. light rays are parallel to the optical axis); hence magnification does not change as axial shifts in the focal plane are performed. However, it should be noted that at higher magnifications, both the axial scanning range and the NA of the detection lens are limited by this relay configuration – which is also difficult to implement in practice (Fahrbach et al., 2013). Here, the primary challenge is to place the additional relay lenses into a commercial system, but this challenge can be readily overcome with a customisable setup that allows user configuration of the light path.

Theoretical analysis and optical simulations of different configurations have further characterised the consequences of ETL positions in a microscopy system (Qu and Hu, 2019): (1) placing the ETL at the back focal plane yields a large scanning range but a varying magnification; (2) positioning the ETL between two relay lenses retains constant magnification but as a consequence decreases the axial scanning range; (3) placing the ETL behind two relay lenses maintains the magnification and produces a large axial scanning range, but constitutes a configuration that can only be achieved through customisation; and lastly (4) positioning the ETL at the imaging detector plane allows a constant magnification without any axial scanning capabilities, therefore compromising a key strength of the ETL. Subsequently, these optical simulations have been confirmed practically by imaging a dot checkerboard target using an ETL setup at driving currents of 0 and 250 mA (Qu and Hu, 2019). Configuration 1 yields a 20.6% difference in magnification across an axial scanning range of 3.540 mm when driving the ETL with 0 and 250 mA, while also exhibiting a degraded image resolution. Configuration 2 leads to a 0.97% change in magnification across an axial scanning range of 82 μm, while dots close to the borders of the imaged checkerboard appear distorted. Configuration 3 has not been tested experimentally due to the requirement for a specifically customised setup; however, optical simulation has shown a 0.23% variation in magnification with an axial scanning range of 2.020 mm. Lastly, configuration 4 results in a 14.5% magnification difference across a 31 μm axial scanning range (Qu and Hu, 2019). These results indicate that the ETL should be positioned based on the specific biological investigation; here, the key factors to consider are the desired axial scanning range and whether a constant magnification is required. The easiest implementation of the ETL is seen in configuration 1, as the ETL is simply placed on top of the objective.

In summary, the largest advantage of the ETL is its flexibility, which is conferred by its on-demand shape-changing capabilities. As a micrometre change in the radius of an ETL can produce the same optical effect as moving a traditional lens by several centimetres (Blum et al., 2011), liquid lenses are beginning to be exploited for rapidly shifting the focal plane within milliseconds and without any mechanical movement, allowing for compact optical setups.

In an era where blueprints for high-performance bespoke microscopes are emerging (Pinto et al., 2021 preprint), ETLs are an attractive approach to overcome the 3D imaging limitations of conventional microscopy systems. In the future, we expect imaging systems to fully exploit the strength of ETLs in reducing motion artifacts and mechanical complexity, which is particularly advantageous for use in unpredictable new environments. For example, ETLs are being investigated by NASA for use in space (Fogle et al., 2020). ETLs are also being miniaturised by Huawei for implementation in upcoming products, including those in the defence sector (Lide et al., 2020), taking advantage of their compact design, which is ideal for use in smaller spaces, their autofocus and their depth detection, as well as image stabilisation without requiring elaborate hardware. These engineering efforts are likely to give rise to robust and miniaturised setups that could stimulate new microscopy applications.

A cost-effective integration of ETLs into existing imaging systems is expected to vastly expand the opportunities for high-resolution and high-speed 3D live-imaging studies. Accordingly, the low cost of ETLs is expected to bring high-speed 3D-imaging capabilities to diverse cell biology communities. Some examples of applications where ETLs have been implemented beyond academic research laboratories include in dentistry (Eom et al., 2020), endomicroscopy (Jabbour et al., 2014) and capsule endoscopes (Seo et al., 2009), as well as in underwater microscopy of marine ecosystems (Mullen et al., 2016).

While there is ample evidence for increased axial scanning ranges and scan rates obtained by using an ETL, the effective management of ETL-induced aberrations requires individualised solutions. Spherical aberrations are the main performance-limiting type of aberration, especially when refocusing is attempted for high-NA imaging applications. Accordingly, optical aberrations at the edge of images may not be relevant for applications in which a smaller usable field-of-view is sufficient.

Alternative imaging modalities can also achieve high-speed volumetric biological imaging, such as lattice light-sheet microscopy, despite complex setup configurations (Chen et al., 2014). In addition to lattice light-sheet microscopy, oblique plane microscopy (OPM), a light-sheet microscopy technique that obliquely illuminates and captures fluorescence through a single objective lens (referred to as the primary objective; Dunsby, 2008), has been endorsed for its aberration-free high-speed 3D-imaging capability (Botcherby et al., 2008, 2012). Any fluorescence captured by the primary objective is relayed to a pair of remotely mounted objectives (referred to as secondary and tertiary objectives) before finally reaching the camera. While OPM setups require a specialised remote scanning feature, without excessive reflective elements, it can be a relatively convenient setup allowing environmental control and imaging of a variety of specimens, including samples in multi-well plates (Maioli et al., 2016; Yang et al., 2019). However, the main drawback of OPM has been its inability to fully exploit the NA of the primary objective during fluorescence detection, resulting in a low optical resolution and sensitivity. The latest OPM setups have attempted to solve this issue by employing a pair of refractive-index-mismatched objectives for remote focusing instead; however, full NA detection imaging has still been limited to a relatively small field-of-view (∼200 μm) at high magnifications (60× and 100×) (Sapoznik et al., 2020; Yang et al., 2019). Other commercially available tunable lenses, in addition to ETLs, are also being used for bioscience microscopy applications (Table 2). For a broader characterisation of tunable optical devices, we direct readers to Kang et al. (2020) and Chen et al. (2021).

Nevertheless, accessible optoelectronic tools to overcome any ETL-induced image aberrations, for example adding an aspherical correction lens within the optical path of an ETL configuration (Vogt, 2020), could help with their widespread implementation. AO is being explored in multiple microscopy modalities (reviewed in Booth et al., 2015; Hussain et al., 2020; Žurauskas et al., 2019) and has been shown to be a promising approach for reducing aberrations in different ETL setups (Fuh et al., 2015; Han et al., 2019; Qin et al., 2020; Ye et al., 2020); but these have to be custom-built for individual systems. AO devices, such as optofluidic refractive wavefront modulators and multiactuator adaptive lenses, are becoming commercially available, suggesting that their integration into microscope setups will eventually become simpler (Banerjee et al., 2018; Bueno et al., 2018; Pozzi et al., 2020; Rajaeipour et al., 2020a,b).

The next natural step is to explore the extent to which deep-learning methods can help to overcome ETL-induced aberrations. The incorporation of deep learning to sense and correct aberrations in bioscience imaging (Cumming and Gu, 2020; Krishnan et al., 2020 preprint; Zhang et al., 2021) is expected to improve the precision and reliability needed for a wider integration of ETLs into live-cell microscopy.

We thank Professor Noriko Hiroi (Keio University) and Dr Naoka Tamura (Draviam group) for providing live-cell imaging data with the ETL and without the ETL, respectively. We also thank Professor Hiromasa Oku (Gunma University), Dr Fabrice Caudron (Queen Mary University of London), Dr Peter Thorpe (Queen Mary University of London), Dr Sam Swift (Image Solutions Ltd) and David Dang (Draviam group) for comments on the manuscript, and Draviam group members for useful discussions.