Super phantoms: advanced models for testing medical imaging technologies

Phantoms are test objects used for initial testing and optimization of medical imaging techniques, but these rarely capture the complex properties of the tissue. Here we introduce super phantoms, that surpass standard phantoms being able to replicate complex anatomic and functional imaging properties of tissues and organs. These super phantoms can be computer models, inanimate physical objects, or ex-vivo organs. Testing on these super phantoms, will enable iterative improvements well before in-vivo studies, fostering innovation. We illustrate super phantom examples, address development challenges, and envision centralized facilities supporting multiple institutions in applying these models for medical advancements.


Introduc�on
Human space flight stands as one of mankind's most remarkable technological achievements.Each step of this ambi�ous endeavor is accompanied by tremendous risks to both humans and machines.However, significant progress has been witnessed from Yuri Gagarin's pioneering spaceflight to the Apollo moon missions, culmina�ng in the present era where crews spend extended periods aboard the Interna�onal Space Sta�on.One crucial element contribu�ng to these achievements is the focus on rigorous tes�ng and op�miza�on on earth using specialized life-sized simulators and training modules.For instance, Neil Armstrong credited the Lunar Landing Training Vehicles [1] for making him feel "comfortably familiar" with the Eagle during the nerve-wracking descent to the lunar surface, where he maneuvered the cra� to land in boulder-ridden territory while low on fuel [2].
We advocate for a comparable emphasis and commitment to rigorous laboratory tes�ng during the development of advanced medical imaging technologies, prior to human tes�ng.Medical imaging research and development are focused on achieving a future where imaging reveals quan�ta�ve imaging biomarkers of anatomy and physiology in three or more dimensions [3].An imaging biomarker is defined as an objec�vely-measured characteris�c extracted from an image used to evaluate normal and pathogenic biological processes.These quan�ta�ve biomarkers will be integral components in the decision-making process of precision medicine [4], with Ar�ficial Intelligence (AI) underpinning every stage of the pipeline, from image reconstruc�on to integra�ve analy�cs and decision-making.However, in the face of this rapid pace of imaging technology advancements, the means and methodologies for tes�ng and valida�ng these technologies in controlled se�ngs before embarking on human studies have not kept pace.
Laboratory tes�ng of new imaging technologies does take place using phantoms.These are inanimate reference objects [5] used to assess the technical specifica�ons of imaging instruments, including spa�al, contrast, and temporal resolu�ons, as also to ascertain bias and precision.In certain cases, advanced phantoms are designed to simulate the anatomy of specific human body parts or organs targeted by the imaging technology.However, these phantoms o�en fail to replicate key aspects of the pathobiology being inves�gated, such as blood flow or perfusion.As a result, imaging phantom studies only provide an ini�al level of technical op�miza�on.Any shortcomings or limita�ons in extrac�ng qualita�ve and preferably quan�ta�ve imaging biomarkers are only discovered once human studies have begun which o�en occurs late in the development process.Human or in general in vivo tes�ng cannot be conducted early during development and improvement stages for various reasons.Firstly, evalua�on of accuracy or devia�on of the imaging method from the gold standard or ground truth, is impossible since the in-vivo ground truth is unknown.Secondly, while precision or variability can be assessed through repeated imaging of human subjects, certain methods involving ionizing radia�on or contrast agents make this impossible.Moreover, repeated imaging o�en fails to provide the necessary insights into error sources required for making essen�al improvements.
There is thus a major gap between tes�ng on straigh�orward imaging phantoms and tes�ng on highly complex human or animal �ssues with their unknown ground truths.In this Perspec�ve, we discuss the need for a new class of phantoms and their tes�ng, which will be a bridge across this chasm.These are digital and physical phantoms that can replicate both the morphology and relevant aspects of physiology in a specific organ under both healthy and diseased condi�ons.These advanced phantoms would facilitate the early evalua�on of an imaging method's ability to extract both anatomic and func�onal imaging biomarkers from acquired data.We call these models 'super phantoms' to signify that they occupy a class of physical and biological realism beyond standard imaging phantoms.We offer several examples of test objects that fit within our defini�on of super phantoms.We close with ways by which the community could arrive at a future paradigm where research into and applica�on of super phantoms, is accepted as an integral component of research and development of imaging technologies.

Super Phantoms
Phantoms are digital or physical test objects used for ini�al tes�ng and op�miza�on of new techniques, as well as for acceptance tes�ng, accredita�on, and quality control [5].They have several atributes depending upon the task at hand, but in general are described as having �ssue-like proper�es.This feature invariably only refers to the interac�on proper�es of the phantom material with the energy beam of the imaging method used, such as x rays or ultrasound waves.They rarely capture the complex pathophysiological proper�es of the organ and their correla�on with image features.
In contrast, we introduce the definition of a "super phantom," as a computer model, an inanimate physical object, or an ex-vivo organ that replicates the most relevant anatomic and functional imaging biomarkers of the structure of interest under carefully controlled laboratory settings.Examples of physiological and func�onal behavior which would produce the relevant imaging biomarkers are vasculature, flow, perfusion, diffusion, oxygen satura�on and mo�on.Examples of structures of interest are organs, body parts, en�re subjects (e.g.humans or small animals) and their various abnormali�es.In the case of digital super phantoms, we can further define them as stochas�c models capable of genera�ng ensembles of phantoms.These ensembles reflect the sta�s�cal proper�es of biological varia�ons present within the cohort to be imaged [6,7].Addi�onally, we can define "blended digital-physical super phantoms" as a combina�on of both digital and physical elements, further expanding the capabili�es of tes�ng and op�miza�on.
We illustrate with Figure 1 the need for super phantoms in two broad scenarios.The first scenario occurs in real-world clinical se�ngs where an established imaging product exhibits inadequate performance in a specific applica�on.To respond to this, improvements in the exis�ng imaging technology or protocols may be recommended.In the second scenario, the clinical need may require a disrup�ve new imaging technology to be developed.In both scenarios, the conceptualiza�on of the solu�on will require early tes�ng with research prototypes on a surrogate of the human organ or body part.During this early phase, digital phantoms can be employed for in silico tes�ng and refinement of the technology and approaches.Following this, digital super phantoms would be used in simula�ons, for valida�on in controlled environments with reliable reference informa�on.Subsequently, experimental studies using physical phantoms, and later super phantoms, should be conducted to evaluate the proposed imaging solu�on.These studies enable inves�ga�on of precision and accuracy, and the es�ma�on of confidence intervals when extrac�ng the required biomarkers.Recommenda�ons for improvements and refinements can be iterated upon un�l an acceptable performance is achieved.Digital super phantoms will serve as valuable tools for conduc�ng virtual imaging trials.These trials allow for the explora�on of a vast range of poten�al solu�ons using objec�ve measures of image quality [6,7].Through this process, ensemble-averaged figures of merit can be computed, facilita�ng robust comparisons and refinement of the technology.
Once the technology solu�on has demonstrated the required level of performance on the super phantoms, the next step involves implemen�ng it in in vivo studies.The studies would be on human subjects but also on small animals in the case of dedicated systems for fundamental studies.A significant advantage of conduc�ng super phantom studies is that these models can represent the ground truth for various proper�es and imaging biomarkers.This not only enables reliable verifica�on, and valida�on of new technologies but also provides data for training Machine Learning algorithms through imaging (both virtual and physical).

Examples of Super Phantoms
It is not our inten�on to provide an exhaus�ve overview of the phantom field nor the budding super phantom field, and will refer the reader to recent extensive reviews [9][10][11][12][13][14].Here, we will briefly discuss three recently published phantoms that align with our defini�on of super phantoms.

1) A digital 4D CT breast imaging super phantom
With advancements in x-ray imaging using digital detectors, there has been a poten�al for func�onal imaging with high spa�otemporal resolu�on, par�cularly in applica�ons such as cardiac and oncological imaging with the use of contrast agents.In the context of breast imaging, there has been a proposal for 4D imaging using dedicated breast CT, which aims to explore the perfusion proper�es of tumors for op�mized treatment [14].
To understand the requirements and specifica�ons of this new approach, Caballo et al. [15] developed a digital breast super phantom.This super phantom not only includes the morphological representa�on of breast �ssues and tumors but also incorporates the wash-in and wash-out kine�cs of iodinated contrastcontaining blood in vasculature and simulated tumors.The morphology of the super phantom is based on segmented pa�ent breast CT scans, which include four �ssue types: skin, adipose �ssue, fibroglandular �ssue, and blood vessels.To create the digital breast super phantom, blood flow characteris�cs from breast MRI data were incorporated into the vessels, background parenchyma, and both benign and malignant tumors [15].Other digital super phantoms for the breast have also been reported [16,17].

2) A physical photoacous�c-ultrasound imaging super phantom
Photoacous�c imaging is an area of ac�ve research for imaging vasculature, hemodynamics, and blood oxygen satura�on, which is par�cularly challenging in quan�ta�ve photoacous�cs [18].Dantuma et al. [19] recently presented a physical breast phantom specifically designed for quan�ta�ve hybrid photoacous�c and ultrasound imaging, which we classify as a super phantom.This phantom accurately represents the main �ssues found in the female breast, including skin, fat, and fibroglandular �ssue, using a custom polymer formula�on with addi�ves to achieve �ssue-specific acous�c and op�cal interac�on proper�es.The �ssue-mimicking materials were me�culously cast layer-by-layer into shapes and sizes that resemble the breast, using 3D-printed molds based on a numerical breast model extracted from MRI images.
To simulate the larger vessels of the breast's vascular anatomy, the super phantom was designed to incorporate wall-less channels.Addi�onally, a flow circuit was developed to perfuse the channels with bovine blood at a controlled oxygen satura�on level.This unique feature of the breast super phantom enables laboratory-based studies in photoacous�cs, allowing for the valida�on and refinement of approaches to recover both qualita�ve and quan�ta�ve features that are sought a�er in in-vivo studies.
3) An ex-vivo 4D flow magne�c resonance imaging super phantom 4D phase-contrast magne�c resonance imaging holds great poten�al for accurate visualiza�on and quan�fica�on of blood flow in three dimensions, along with the assessment of cardiac parameters [20].However, conduc�ng 4D flow MRI studies on live animals presents numerous challenges and limita�ons, including the lack of precise control over the disease process, ethical concerns, high costs, complex housing and maintenance requirements, and the need for anesthesia.
To address these limita�ons, an MRI-compa�ble pla�orm was created centered around an isolated perfused bea�ng pig heart, which meets the criteria of a super phantom.This pla�orm was specifically designed to replicate in-vivo blood flow behavior, thereby genera�ng physiological coronary flow and myocardial perfusion.The isolated heart model ac�vely pumped blood in both ventricles, closely mimicking the natural filling of the coronary arteries and replica�ng physiological heart func�on with precise control over relevant physiological parameters.With this pla�orm, researchers can inves�gate the impact of various surgical procedures on intra-cardiac flow and evaluate the performance of MRI sequences in different pathophysiological se�ngs without the need for live animal experiments.
Figure 2 provides an overview of several advanced phantoms that have been recently published and can be classified as super phantoms according to our defini�on.

Open challenges for super phantoms
It is important to acknowledge cri�cal steps and considera�ons in the emerging field of super phantoms: 1. Enhanced morphological and (patho)physiological representa�on: Super phantoms as we define them, should aim on the one hand to provide an accurate and detailed representa�on of anatomical structures.This represents technical challenges in arriving at accurate variants to the extent necessary in terms of �ssue composi�on and realis�c geometry.On the other hand, super phantoms should also replicate physiological processes such as blood flow, perfusion, and �ssue mo�on, which is technically.Further, there is also much to be learned about replica�ng various pathological condi�ons, such as tumors, vascular abnormali�es, or �ssue pathologies.
2. Mul�-modality integrated imaging: These devices are implementa�ons where the hardware of independent imaging modali�es are physically integrated for complementarity leading to improved diagnos�c and quan�ta�ve accuracy, and beter understanding of disease-associated biological processes [26].Examples of such hybrid imaging embodiments are SPECT (Single Photon Emission Tomography)/CT, PET (Positron Emission Tomography)/CT, and PET/MR imaging, which are available for clinical use and small-animal research [26].Hybrid imaging throws up several challenges in crossmodality image reconstruc�on, image fusion and image interpreta�on, to new requirements for training and educa�on of medical professionals.Further, valida�on of the clinical u�lity and accuracy of hybrid imaging approaches requires �me-consuming and expensive studies.Laboratory studies will help surmount some of the challenges above except that there is a paucity of mul�-modal phantoms and super phantoms.These will be challenging to develop as they should strive to possess the imaging biomarkers for mul�ple imaging modali�es.3. Metrological underpinning: Since super phantoms aim to represent the ground truth for certain quan�ta�ve imaging biomarkers, a metrological framework is essen�al [27].Such a framework is to ensure that super phantom systems and their sub-systems such as Tissue-Mimicking Media (TMM) [28] are accurately characterized.These would be independent methods accepted as reference methods and calibrated to reference materials, and their proper�es labelled with the measurement uncertainty.
Repor�ng of uncertain�es will provide understanding of the reliability and limita�ons of these super phantoms.The measured parameters such as the size, density, flow etc. should be traceable to verifiable physical SI quan��es.Such well characterized super phantoms would provide the ground truth to calibrate the specific imaging systems for which they were designed.4. Technological advancements: There are vast opportuni�es for super phantom development to leverage technological advancements, such as 3D prin�ng [29], novel materials, and advanced manufacturing techniques, to enhance the physical fidelity of phantoms.Incorpora�ng smart materials, sensors, and actuators can enable real-�me feedback and control, further improving realism and func�onality.5. AI-based digital phantoms: State-of-the-art frameworks for synthe�c data genera�on typically leverage prior knowledge on �ssue anatomy and func�onal �ssue parameters for the genera�on of synthe�c phantoms [30].More recently, data-driven approaches are evolving as a promising alterna�ve to the tradi�onal model-based methods due to their ability to close the so-called domain gap between real and synthe�c data [31,32].Such deep learning-based methods have the poten�al of producing large amounts of diverse and highly realis�c digital phantoms [32].A remaining challenge is the combina�on of such digital phantoms with physical phantoms in the spirit of blended digitalphysical super phantoms.Ini�al steps have already been taken in the context of surgical training, in which tradi�onal training (physical) boxes are combined with varying, poten�ally AI-generated, internal anatomy to prac�ce surgical skills.

Digital twins:
A related open challenge is the development of pa�ent digital twins.While personalized 3D prin�ng of individual organs for treatment planning and execu�on is becoming increasingly common, current methods o�en focus on only a specific aspect of pa�ent-individual informa�on, such as the 3D anatomy.Future work should be directed towards more holis�c approaches, combining all the available informa�on on a pa�ent in one digital twin model, which could then serve a mul�tude of purposes, including outcome predic�on, personalized therapy planning and even dry-runs of surgical interven�on on the digital twin of a pa�ent to be operated.7. Valida�on, standardiza�on and collabora�on: Standardized protocols for super phantom fabrica�on and characteriza�on as in 5. above, along with standardized protocols for their use in terms of image acquisi�on and analysis are essen�al for making super phantom reliable tools.This will facilitate reproducibility across different laboratories promo�ng inter-laboratory comparisons and enable the exchange of data with confidence [33,34].EIBALL (European Ins�tute for Biomedical Imaging Research) and QIBA (Quan�ta�ve Imaging Biomarkers Alliance) are examples of ini�a�ves focused on improving quan�ta�ve imaging biomarkers, where standardized ground truth phantoms [35,36] are crucial.Another ini�a�ve specific to the photoacous�c imaging modality is IPASC (Interna�onal Photoacous�c Standardiza�on Consor�um) [37] one of whose tasks is to secure standardiza�on of design, fabrica�on, and use of phantoms to validate and benchmark photoacous�c imaging systems.Similar efforts will be required for super phantoms as well, where collabora�on among researchers is vital with sharing of phantom designs, fabrica�on and valida�on protocols, and imaging data.8. Improve awareness of the field: The realiza�on is growing that we need imaging on super phantoms, as an intermediate phase between imaging on standard phantoms and imaging on humans.However, more efforts should be made to improve awareness of super phantoms, and their roles in valida�ng and assessing the latest genera�on of imaging technologies.This will provide an impulse to atract talented engineers to contribute to the field, and also help to draw interest from other academic domains such as material sciences and addi�ve manufacturing.

Next steps for super phantoms
To address the challenges and considera�ons above, it is necessary to establish super phantoms as an integral part of research and development in imaging technologies.We believe that this is possible with the following concrete steps:

A. Establishment of Super Phantom Core Facili�es
Phantom development is predominantly carried out by PhD students and Post-Docs within individual research groups.This fragmented and decentralized approach o�en leads to compromised instrumenta�on quality and redundancy due to limited individual budgets.To address this, we propose the establishment of a centralized shared facility that can cater to mul�ple ins�tu�ons, equipped with high-quality instrumenta�on and dedicated personnel.
This facility would ideally be located at a university with a strong biomedical engineering program and established collabora�ons with academic hospitals, to house a range of equipment and supplies for the following: 1. Fabrica�on and characteriza�on of Tissue-Mimicking Media (TMM): The facility would house a general chemistry laboratory, a 3D prin�ng sta�on [29], and a machine shop to prepare a variety of TMMs [28] using gels, elastomers, plas�cs, and resins.These TMMs would be doped with addi�ves to replicate different �ssue types.The characteriza�on of TMMs would require instruments and tools to measure their interac�on coefficients [38].For example, an acous�c characteriza�on scanning system [38,39] could be employed to measure acous�c atenua�on, sound speed, acous�c impedance, and non-linearity for ultrasound TMMs.These measurements are crucial for assessing the suitability of the TMMs and informing any necessary itera�ons or op�miza�ons during the development process, and for establishing the ground truth values for the super phantom.2. Super phantom pla�orm development: TMMs would be assembled to construct the body part or organ of interest.This stage would involve incorpora�ng physiological features and behaviors into the super phantom, such as circula�on, perfusion, and mo�on.The facility would facilitate the integra�on of these elements to create the super phantoms.

Valida�on and experimenta�on:
The complete super phantom would be imaged using the specific modality for which it was designed.The facility would require access to imaging equipment, similar to that used in clinical se�ngs, to replicate the real-world opera�onal environment.Close collabora�on with radiologists and clinicians would be essen�al for the analysis and interpreta�on of the acquired images.The primary goals would be to validate the super phantom's ability to meet the quan�ta�ve func�onal requirements and to conduct studies using research variants of imaging equipment to implement improvements in technology or protocols.4. Compu�ng: The facility would need high-performance compu�ng resources for genera�on of digital super phantoms as well as for conduc�ng in silico experiments and trials.Furthermore, numerical procedures for objec�ve and quan�ta�ve assessment of both digital and physical super phantoms would be implemented using these tools.

B. Infrastructure for Collabora�on and Standardiza�on
The facility would serve as a central hub for visi�ng research groups from the imaging and biomedical engineering communi�es, fostering collabora�ve efforts in the field of super phantoms.It would also provide the necessary infrastructure for intramural imaging studies, promo�ng a paradigm of collabora�ve experiments.Customized super phantoms could be disseminated within the community for measurements and inves�ga�ons, including inter-laboratory studies.
Se�ng up and opera�ng such a facility would entail significant costs and necessitate substan�al ini�al investment through mul�-ins�tu�onal support and various programs offered by na�onal and interna�onal funding agencies.To ensure sustainability, the facility's running costs would be covered by project grants, with internal and external users budge�ng for their use of the facility.
To promote knowledge sharing and collabora�on, protocols for fabrica�on and characteriza�on measurements on physical phantoms, along with the resul�ng data, would be made accessible to the community through databases.Digital super phantoms would be open source, enabling imaging technology developers worldwide to readily access them for virtual imaging studies.This imaging data would be invaluable for qualita�ve and quan�ta�ve tes�ng of image reconstruc�on, processing, and analysis algorithms and for training machine learning models.To host the data effec�vely, appropriate metadata would require to be recorded with domain-specific ontology for seman�c search and data integra�on.
Online collabora�on environments would facilitate consensus building on various aspects related to super phantoms, aiming to establish standards.The facility would play a vital role in developing and maintaining standard protocols for super phantoms, serving as a repository of knowledge and exper�se in both physical and digital variants of these phantoms.

Concluding Remarks
We firmly believe that super phantoms, capable of mimicking the physical and func�onal characteris�cs of �ssues and organs in both healthy and diseased states, will play a crucial role in the research and development of new imaging technologies.These phantoms offer the opportunity for comprehensive assessments and itera�ve op�miza�on of imaging modali�es at lower Technology Readiness Levels (TRLs) [40], enabling improvements in their chances of success before human studies.The outcomes of studies using these sophis�cated phantoms may also provide the first evidence of readiness for regulatory boards when assessing applica�ons for subsequent clinical studies and CE (conformité européenne) cer�fica�on [41].
In order to integrate super phantom work into the research and development of imaging technologies, it is essen�al to establish shared facili�es housing the necessary infrastructure for designing, developing, characterizing, and valida�ng super phantoms.Such a super-phantom facility could serve as a hub for researchers from within and outside the hos�ng ins�tu�on, fostering improved rigor in studies and facilita�ng the free flow of ideas for con�nuous enhancements of the super phantoms and the studies conducted with them.
With growing awareness of the nature and significance of super phantoms and the establishment of such facili�es, talented biomedical engineers will be atracted to engage in the research, development, and u�liza�on of these phantoms.The field is replete with numerous fundamental and applied ques�ons and challenges, offering ample opportuni�es for interdisciplinary research.For instance, we s�ll have much to discover regarding the relevant interac�on proper�es of imaging modali�es (op�cal, acous�c, x-ray, etc.) in various pathologies, which is essen�al for developing super phantoms that accurately reflect clinical reality.The level of realism required in super phantoms and how to define and measure this realism also represent important and rich areas of explora�on.Naturally, technical challenges abound, such as accurately mimicking and measuring pathophysiological flow and perfusion in different organs.
The role of industry will be important in the future for dissemina�on and use of super phantoms The acquisi�on and use of these phantoms for assessing and valida�ng expensive advanced imaging systems hold significant commercial appeal.As awareness of super phantoms grows, this nascent sector is expected to steadily expand, encouraging increased industry par�cipa�on and direct investment in the future.
While our Perspec�ve has centered on super phantoms for imaging applica�ons, these tools can play a roles in planning therapeu�c interven�ons such as radia�on therapy [42], thermal abla�ons [43], and surgery [44].The principles and insights discussed in our work can serve as a founda�on for applica�on of super phantoms in these areas.More research and development will be necessary to simulate the biological and physical responses of �ssue subjected to therapeu�c interven�ons.
In summary, we strongly believe that the field of super phantoms will become increasingly important, significantly accelera�ng progress, fostering breakthroughs, and expedi�ng the clinical transla�on of new and effec�ve imaging modali�es and approaches.

Figure 1 :
Figure 1: Mapping from real-world clinical setting to the research lab super phantom setting.The concept of "imaging super-phantoms" pertains to surrogates of tissues and organs that reproduce in vivo imaging features.These super phantoms are available in carefully-controlled laboratory settings to facilitate the assessment and refinement of emerging imaging technologies and methodologies, prior to human studies and clinical application.

Figure 2 :
Figure 2: Examples of test-objects that we classify as super phantoms.The asterisks indicate those presented in text, reference numbers are provided in square brackets, and the imaging modality for which the super phantom is intended for is in round brackets.(Acronyms: 4D -four-dimensional, SaO2 -oxygen satura�on, CT -Computed Tomography, MRI -Magne�c Resonance Imaging, DSA -Digital Subtrac�on Angiography, MPI -Magne�c Par�cle Imaging.)The drawing of the body is inspired by the Vitruvian man created by Leonardo Da Vinci in 1490.The drawing was commissioned by the corresponding author, and he may use this in publica�ons, presenta�ons, web-sites and project proposals.