2.1.

Step 1: Understanding the Concept of the Imaging Modality

The first step in developing robust measures to test medical imaging systems is a comprehensive understanding of the underlying principles driving the imaging modality of interest and its source of contrast. For PAI, the signal is generated through the photoacoustic effect, which refers to the generation of ultrasound waves from the absorption of electromagnetic energy³ (Fig. 1): short light pulses, typically in the nanosecond range, illuminate a sample and the absorption of energy by the imaged object results in heat generation; the rise in temperature leads to an increase in pressure, which generates broadband ultrasound waves that propagate through the tissue and are detected by ultrasound transducers. The amplitude of the recorded pressure wave provides information about the local absorption of optical energy within the object. The time interval between the illumination pulse and the arrival of the ultrasound wave at the detector can be used to calculate the distance between the detector and the absorbers.

Fig. 1

Photoacoustic signal generation. A pulsed light source illuminates the object to be imaged (e.g., tumor tissue). As the light propagates through the tissue, it is scattered and absorbed by spatially varying endogenous or exogenous chromophores. The absorption and scattering coefficients, ${\mu }_a$ and ${\mu }_s$, determine the fluence distribution $\mathrm{\Phi }$ and consequently, the absorbed energy distribution $H$. The absorbed energy generates a pressure distribution $p_0$. The combined photoacoustic efficiency of conversion from heat into pressure is represented by the Grüneisen parameter $\mathrm{\Gamma }$. Due to the elastic nature of the tissue, the generated pressure propagates as an acoustic wave through the tissue and is detected by ultrasound sensors. This process is affected by the acoustic properties of the tissue and the sensor response. Finally, image reconstruction is performed to visualize the recorded data. Created with BioRender.

For illumination in tissues, light in the near-infrared (NIR, 650 to 1350 nm) window is often chosen to optimize penetration depth.⁴ Applying pulses at multiple wavelengths enables multispectral photoacoustic systems to distinguish between endogenous chromophores or exogenous contrast agents, such as nanoparticles or organic dyes.^5,6 The primary endogenous absorbers in tissue are oxy- and deoxy-hemoglobin (${\mathrm{HbO}}_2$ and RHb), lipids, melanin, collagen, and water (Fig. 2). The optical absorption coefficients vary with wavelength, and thus the relative concentration of each chromophore can be extracted through spectroscopic inversion. Functional parameters such as haemoglobin oxygen saturation (${\mathrm{sO}}_2$) can be calculated from the absorption difference between ${\mathrm{HbO}}_2$ and RHb (${\mathrm{sO}}_2={\mathrm{HbO}}_2/({\mathrm{HbO}}_2+\mathrm{RHb})$).^9,10 Using only endogenous contrast, PAI therefore has the potential to non-invasively visualize simultaneously various important tissue parameters within a single imaging session. Current state-of-the-art technology can achieve two-dimensional imaging in almost real-time, and three-dimensional (3D) images in time scales of seconds to minutes.^5,11,12 Importantly, the modality is also able to image across scales, from whole tumor volumes (macroscopic, cm depth) over to vascular networks (mesoscopic, mm depth) down to individual cells (microscopic, $\mu \mathrm{m}$ depth)¹¹ simply by scaling the system configurations. Providing 3D multi-parametric information with high temporal resolution, PAI has found a wide range of applications in both pre-clinical and clinical environments.¹³

Fig. 2

Absorption spectra of the main endogenous chromophores. Absorption spectra are displayed for melanin (brown); oxy- (red) and deoxy-hemoglobin (green; both $150{\mathrm{gl}}^{-1}$); water (blue; 80% by volume in tissue); lipids (yellow; 20% by volume in tissue); and collagen (black) in the wavelength range of 400 to 1400 nm. Data from Refs. 7 and 8.

2.2.

Step 2: Defining Relevant Parameters and Their Magnitude

Following a thorough comprehension of the imaging modality, the parameters that play a role in the signal generation process need to be precisely defined and their magnitude well understood to design and develop suitable phantoms for the modality (Table 2). For PAI, optical, acoustic, and (thermo)mechanical properties are paramount. To characterize light propagation, knowledge of the probability of a scattering or absorption event per unit path length is essential. These parameters are captured by the linear absorption coefficient ${\mu }_a(\lambda )$ and linear scattering coefficient ${\mu }_s(\lambda )$. For scattering events where the source-to-measurement distance is much larger than $1{/\mu }_s(\lambda )$, the reduced scattering coefficient ${{\mu }_s}^{\prime }(\lambda )$ is used to describe scattering behavior. The reduced scattering coefficient is defined by ${{\mu }_s}^{\prime }(\lambda )=(1-g){\mu }_s(\lambda )$ where $g$ is the anisotropy factor, a parameter that indicates how scattering intensity varies with angle. For PAI phantom applications, the anisotropy factor plays a role in microscopic or mesoscopic applications or interactions near a light source. In addition, the refractive index $n$ is relevant for light reflection and refraction at interfaces. Scattering and absorption effects are wavelength-dependent and determined by the constituents of the medium.

Table 2

Definition of relevant material properties parameters in optical and acoustic imaging applications. The soft tissue values are just representative values. The properties may vary with the experimental conditions (e.g., temperature). See literature for values that pertain to specific samples and controls.

When light is converted into heat and subsequently into a pressure wave, the efficiency of the conversion is represented by the Grüneisen parameter $\mathrm{\Gamma }$. In order to assess how the generated pressure wave propagates within a medium, a well-characterized broadband frequency description of the material-specific speed of sound $c$ and acoustic attenuation coefficient α is essential. Other acoustic parameters, such as the acoustic backscattering coefficient ${\mu }_{bs}$ or the ultrasound non-linearity parameter ($B/A$), are less reported in PAI, as the key interest lies in determining the amplitude of the acoustic losses than analyzing the fate of the encountered losses. The propagation of acoustic waves is determined by the mechanical properties of the medium, specifically by its density and elastic properties (quantified by the Young’s modulus/shear modulus).

Average values for the optical,^15,20 acoustic,^16,21 and thermoelastic¹⁴ properties of soft tissues are summarized in Table 3. It is worth noting that methods for measuring these parameters across different tissue types can often require complex equipment that must be independently calibrated, which is particularly challenging for some parameters, e.g., the acoustic backscattering coefficient or the Grüneisen parameter.

Table 3

Overview of representative acoustic and optical properties found in soft tissues. Optical properties cover a spectrum from 600 to 900 nm. For values outside this wavelength range, please refer to the literature.20

2.3.

Step 3: Defining the Purpose of the Phantom

For the creation of high-performing phantoms, it is imperative to thoroughly comprehend their intended objectives and functions. Phantoms fulfil various tasks along the full translational pipeline of an imaging modality^29,30 including:

Phantoms are often designed to meet one or several of these purposes, which dictates their final design and intrinsic properties.

2.4.

Step 4: Defining Desired Properties of a Phantom Material

The ideal properties of a phantom are specified by its final application. For example, phantoms that are targeted toward testing the signal repeatability over time (precision phantoms) require high temporal and mechanical stability, along with reproducible fabrication. Phantoms that are used for training, testing, or verification purposes (accuracy phantoms) should mimic tissue properties accurately (e.g., specific tissue types, pathologies, and species of interest) either in a static or dynamic manner, to replicate expected signals. It is recommended that the bulk medium from which a phantom is prepared fulfils the following base properties³⁰ (ranked in no preferential order):

A phantom material for PAI should also maintain its structural and material integrity when in contact with an aqueous medium as water-based solutions are often used as acoustic coupling agents during signal acquisition. In addition, the material should be photostable at the visible and NIR wavelength range under safe exposure limits as encountered during imaging, handling, and storage. A suitable bulk material should further allow the embedding of target inclusions without their degradation to enable quantitative assessment of image quality metrics for specific applications. Ideally, the material should allow for the inclusion of targets made out of the same material type as well as of targets made out of different material types (e.g., microspheres, wires, etc.). Further properties should be tailored toward the system type, PA diagnostic procedure, and/or tissue type of interest. For example, for surgical training phantoms, a material type with “self-healing” properties (e.g., self-removal of applied needle tracks or cuts) may be beneficial to increase the life span of the phantom and minimize costs and manufacturing time. Phantoms for macroscopic systems may only require mimicking spatially averaged properties of biological tissue, whereas phantoms for microscopic imaging applications may need to replicate the fine structural details and heterogeneous composition of tissues to accurately support high-resolution imaging. This also necessitates adapting respective manufacturing methods (step 6).

Importantly, an ideal material should be accessible to everyone in the scientific community. If it cannot be procured in a “ready to manufacture” state at more than one standard scientific material supplier, its ingredients (including chemical abstract service numbers) and detailed manufacturing process should be openly published. Reproducible fabrication should be evidenced by a multi-center study (see step 7) to ensure that the material has broad accessibility, achieving desired properties within acceptable uncertainty limitations. An overview of tissue-mimicking materials (TMMs) proposed for PAI phantoms can be found in Table S2 in the Supplementary Material.

2.5.

Step 5: Defining Means to Characterize the Desired Properties

Phantoms require detailed characterization of their intrinsic properties with specialist equipment that is regularly calibrated. Metrology institutes often host such facilities and several already participate in biophotonic standardization initiatives. Equipment hosted in research and industrial laboratories should ideally be cross-referenced to such reference institutes to determine the accuracy with which local characterization can be undertaken for a given material. Ideally, traceability to gold-standard metrology, such as those supported by the National Institute of Standards and Technology (NIST), should be achieved. Guidance is given here for characterizing the optical (${{\mu }_s}^{\prime }$, ${\mu }_a$) and acoustic properties ($v_f,{\alpha }_f$) of a material. Advice on techniques to measure the density,³⁶ non-linearity parameter,³⁷ the acoustic back-scattering coefficient,³⁸ echo reduction,³⁹ and Grüneisen parameter^19,40 can be found elsewhere as referenced.

2.6.

Step 6: Designing a Phantom

After defining the properties of the base phantom material, a suitable design needs to be found that allows the task of interest to be performed (Fig. 3). A phantom design is dependent on the user requirements, which—as outlined in step 3—can be manifold,³⁰ ranging from user training, instrument calibration or optimization, to prototype testing. The design parameters that arise from these requirements can be: (1) qualitative (relating to the overall architecture), e.g., concerning the shape, form, size, dimensionality, or the presence of certain particular design features, such as flow circuits; or (2) quantitative (relating specific TMM properties), specifying optical $({{\mu }_s}^{\prime },{\mu }_a,g,n)$, acoustic ($v_f,{\alpha }_f$), and/or thermal ($\mathrm{\Gamma }$) values (Table 2).

Fig. 3

PAI phantom designs. (a) Overview of the general phantom design process. (b) Representative example of a basic image sensitivity PAI phantom: schematic (top) and PA image (bottom) of a PVCP phantom containing PTFE tubes filled with different concentrations of India ink. (c) Representative example of a more complex, anthropomorphic phantom: photograph (top) and PA image (bottom) of a breast-shaped PVCP phantom. Panel (b) adapted with permission from Ref. 62, Optica. Panel (c) adapted with permission from Ref. 63, Optica.

Having identified suitable design parameters, the phantom can be developed using appropriate manufacturing methods.^64,65 These can include, but are not limited to, (1) casting and molding techniques, (2) additive manufacturing techniques (e.g., 3D-printing, layer-by-layer assembly, or lithographic techniques), and (3) subtractive and fiber formation techniques (e.g., machining and milling; electrospinning). Following phantom fabrication, the design parameters need to be verified using suitable characterization methods (see step 5). Finally, validation needs to be performed by imaging the fabricated phantom to test whether it meets its functional requirements. For simple phantoms designed for quantification of basic imaging parameters, guidance on consensus test methods best practices have been recently summarized.⁶⁶ PAI standards for phantom designs (e.g., relating to phantom geometry, target inclusion sizes and patterns, and measurement/analysis protocols) do not yet exist but will be addressed by IPASC in future efforts to enable objective, quantitative assessment of image quality across different devices. Such standards may not only be beneficial for basic performance assessment phantoms^{22,62,67–69} but should ideally also extend to biomimetic, anthropomorphic phantoms⁶³ to provide more clinically relevant image quality assessment approaches.

2.7.

Step 7: Ensuring Reproducibility

The last—and arguably most important step—is ensuring reproducibility of the phantom manufacture. Only if a material can be manufactured reproducibly, does it have the potential to be reliably used at different centers allowing inter-device comparison and calibration. If a phantom material cannot be obtained from a commercial supplier that can certify its properties, it is recommended to conduct a multi-center study to verify the compliance of a material candidate with the aforementioned requirements. This ensures that the candidate material can be reproduced and characterized reliably using a provided protocol. For such study, three batches of the material should be repeatedly ($n\ge 3$) manufactured by $\ge 3$ independent institutions to enable statistical analysis. It is recommended that relevant material parameters of the resulting material batches should be verified by at least two institutions to avoid bias toward a certain measurement instrument. Variations of the phantom properties should fall within specified acceptance ranges. For acoustic properties, uncertainty ranges for the speed of sound and acoustic attenuation values of TMMs have been specified before⁷⁰ (acoustic attenuation coefficient $\alpha \pm 7\%$, speed of sound $v\pm 1\%$ within a $\pm 1.5\mathrm{MHz}$ bandwidth). The measurement of optical properties with high accuracy and precision is more challenging, and uncertainty values will highly depend on the measurement technique used^30,42,71 and the final properties of the phantom. For this reason, users are advised to adapt the optical uncertainty limits as appropriate to their final application.

Using ingredients available from standard scientific suppliers can help to minimize batch-to-batch variations and maximize availability. The temporal variation/stability of the material properties should be determined under prescribed characterization and material storage conditions. Properties should ideally remain stable for a time-period of at least 6 months. Accurate and precise assessment of material properties requires detailed knowledge of the characterization systems and, in particular, an assessment of the uncertainty associated with each characterization measurement.

The criteria employed for stability should be explicitly stated for each test, but it will typically involve any drift in the parameter being less than 2× the accepted uncertainty figure.