Seeing the Future with Imaging Science: Interdisciplinary Research Team Summaries (2011)

The development of treatments for human diseases has followed models and methods that have changed very little during the past 50 years. Capitalizing on the power of imaging technologies offers an opportunity to improve these models and methods and to make the search for improved treatments more efficient and the treatments themselves more efficacious. Traditionally the search for treatments goes through stages: target identification, compound synthesis and screening, evaluation in animal models, phase 1-4 testing, and assessment of outcomes such as efficacy or side effects. The challenge to imaging technology is to find ways with which this search could be improved.

Traditionally, dose-finding studies have relied on a relatively crude trial-and-error approach in which multiple doses are used, and symptom improvement has been the measurable target. The development of methods to image targets that are more primary to the disease process offers the possibility to improve dose-finding methods and outcome assessment. For example, labeled ligands have been developed that bind to neuroreceptors for neurotransmitters that are thought to be overactive or underactive in brain diseases (e.g., dopamine in schizophrenia, Parkinson’s disease; serotonin in

mood disorders) and therefore are conceptualized as more precise treatment targets. These ligands (e.g., [¹¹C]raclopride for D2 receptors, [¹⁸F]setoperone for 5-HT₂receptors) can be used to measure the occupancy of receptors induced by varying doses of medications, and the degree of occupancy can then be correlated with level of symptom improvement or side effects to determine the level of receptor occupancy required for optimal treatment. This approach is now widely used in order to determine the optimal doses of some psychoactive drugs. The number of available ligands that are U.S. Food and Drug Administration (FDA)-approved is limited, however, and so the development of new and better ligands for drugs of many types continues to be a significant challenge.

For many years [¹⁸F]fluorodeoxyglucose-2-deoxy-D-glucose (FDG) has been used to identify the location and size of cancer lesions and to monitor their response to treatment using PET. FDG, while useful, is also relatively crude and nonspecific. The application of imaging technology to monitoring treatment targets can be substantially enhanced if investigators develop new amino acid ligands (tyrosine, methionine, thymidine) that aim at more specific targets, such as hormones (e.g., receptors for estrogen, testosterone) or substrates involved in protein or nucleic acid synthesis.

Most current applications of imaging technology to improving treatment examine activity on a large scale: systems, organs, lesions, etc. A new technology has recently emerged that permits imaging at the intracellular level and the ability to manipulate cellular function. This technology is referred to as “optogenetics” because it uses light-responsive proteins derived from algae (channelrhodopsin and halorhodopsin) that can be used to manipulate neuronal firing by opening or closing ion channels. Channelrhodopsin responds to blue light and produces neuronal firing, while halorhodopsin responds to yellow and silences the cell. Because halorhodopsin also responds to red or near infrared, and because infrared can pass more deeply into tissue, optogenetics offers the possibility of providing nonsurgical control over circuits deeply located in the brain. Optogenetics—a relatively new technology—has been used to study many facets of neuroscience, such as brain reward circuits and mechanisms of memory. It has the potential to supplant deep brain stimulation as a treatment for diseases such

as Parkinson’s disease or depression, and its potential efficacy for restoring or improving vision by activating damaged retinal cells is also being examined.

Although disease symptoms are treated as the classic target in drug development, it is obvious that disease mechanisms are a more appropriate and efficient target. Imaging offers a variety of opportunities to improve target identification.

An obvious example of the potential utility of imaging tools is to apply the many methods available from standard imaging technologies such as structural magnetic resonance (sMR), diffusion tensor imaging (DTI), magnetic resonance spectroscopy (MRS), or positron emission tomography (PET) and to use their various measurements to conduct case-control comparisons and to thereby identify anatomic, biochemical, or physiological indicators and mechanisms of disease onset or progression. For example, this approach has been helpful in studying the mechanisms of both classical Mendelian diseases such as Huntington’s (which displays evidence of tissue pathology prior to clinical onset that is correlated with the number of CAG repeats) and non-Mendelian diseases such as schizophrenia (which also has indications of tissue change prior to onset that have been linked with increasing replicability to a group of candidate genes such as DISC1, NRG1, and BDNF). Although both of these diseases have evaded treatments that prevent or reverse their onset and progression, imaging research can be used to develop methods that point to new treatment targets. MRS studies of schizophrenia have, for example, supported the search for drugs that affect the glutamate system.

Testing of new compounds that are potentially therapeutic has traditionally been done using animal models and measuring behavioral or metabolic changes. Imaging offers the opportunity to improve on this methodology by offering an opportunity to conduct high-throughput studies of a variety of animal models and also to study disease mechanisms. Mouse and rat models are available for a variety of diseases, and finding new and/or improved models is an important challenge. Another important challenge is to find ways to use imaging to test compounds in simpler systems that can be studied more efficiently and inexpensively. For example, the zebrafish has become an important vertebrate animal model for a variety of human diseases because it is relatively easy to modify genetically. There are now many zebrafish models of diseases, ranging from porphyria to hypothyroidism to

age-related cognitive decline. Zebrafish are also susceptible to carcinogens and to infectious agents such as tuberculosis. Therefore, zebrafish offer an attractive option for high-throughput screening of drugs using imaging technologies because such studies can be conducted rapidly and on a large scale. Chemical libraries or potential therapeutic agents could be tested for efficacy or toxicology using zebrafish embryos or larvae and applying digital imaging methods to measure outcome. Investigators are also exploring ways to use more standard imaging technologies (e.g., MR) to screen drugs in mammal animal models.

• How can we be certain that novel imaging methods are yielding valid measures (e.g., the extent to which a specific level of ligand displacement/receptor occupancy reflects optimal treatment levels, the accuracy of tumor volume estimate)? What are their technical or statistical limitations?

• How can existing imaging technologies be applied or modified in novel ways to develop new or better treatments?

• To what extent could new imaging methods such as dual-wavelength laser speckle imaging (measures blood flow, blood volume, and tissue hemoglobin oxygenation) or digital-frequency-ramping optical coherence tomography (images quantitative 3-D vascular network) add new insights to functional imaging?

• What characteristics of an imaging system are most important either for administering therapy or assessing its efficacy?

• What are some ways that high-throughput imaging for drug screening could be enhanced?

Gradinaru V, Mogri M, Thompson KR, Henderson JM, Deisseroth K. Optical deconstruction of Parkinsonian neural circuitry. Science 2009;324:354-9. Accessed online June 15, 2010.

Lieschke GL, Currie PD. Animal models of human disease: zebrafish swim into view. Nat Rev Genet 2007;8:253-367. Preview accessed online June 15, 2010.

Mamo D, Kapur S, Shammi CM, Papatheodorou G, Mann S, Therrien F, Remington G. A PET study of dopamine D2 and serotonin 5-HT2 receptor occupancy in patients with schizophrenia treated with therapeutic doses of ziprasidone. Amer J Psychiatry 2004;161:818-25. Accessed online June 15, 2010.

Because of the popularity of this topic, three groups explored this subject. Please be sure to review the second and third write-ups, which immediately follows this one.

• Robert S. Balaban, National Institutes of Health

• Shelley A. Batts, Stanford University

• Ashley Grant, University of Texas

• Cindy M. Grimm, Washington University in St. Louis

• Joseph J. Jankowski, Case Western Reserve University

• Mark W. Lenox, Texas A&M University

• Anant Madabhushi, Rutgers University

• Amina A. Qutub, Rice University

• Lyudmila A. Sakhanenko, Michigan State University

• Kimani C. Toussaint, Jr., University of Illinois at Urbana-Champaign

• Roma Subramanian, Texas A&M University

Roma Subramanian, NAKFI Science Writing Scholar, Texas A&M University

It is the year 2030. Wanda goes in for her annual mammogram, which is performed using monochromatic X-ray imaging. The image reveals a breast tumor. Wanda’s DNA profile reveals mutations in tumor-suppressor genes BRCA1/2, indicating increased breast-cancer risk. Image-guided biopsy is then performed using multimodal imaging, which provides both functional and anatomical data. The resulting imaging data together with Wanda’s genomic data are fed into a national systems-medicine database to identify possible treatments and their probable outcomes. Minimally invasive image-guided surgery is selected to resect the tumor. Further, an inoperable metastasis detected in the lung is treated by targeted drug administration and real-time image-guided evaluation of the efficiency and efficacy of drug delivery.

After two days of lively brainstorming at the 2010 National Academies Keck Futures Initiative Conference on Imaging Science, an interdisciplinary team of nine researchers, with backgrounds ranging from electrical engineering to infectious disease, began their presentation of the solutions to

the challenge posed to the them with the above narrative, which presents a futuristic vision of medical imaging.

As the team members acknowledged, achieving this vision rests on three foundations: creating improvements in imaging technology, diagnostics, and therapeutics.

IDR team 7A began tackling the challenge by drawing up the following list of novel imaging technologies and their applications in diagnostics and therapeutics.

• Two-photon fluorescence microendoscopy: This minimally invasive imaging technology provides micron-scale resolution images of tissues in regions inaccessible by light microscopy. It combines two-photon fluorescence microscopy, which eliminates light scatter from deep tissues, and microendoscopy, which enables the visualization of deep-seated structures through the use of endoscope probes composed of microlenses.

  A clinical application of this technology is in the surgical technique of cochlear implant electrode insertion. Handheld portable fluorescent microendoscopes have been developed to visualize the middle and inner ear during this procedure, which is currently being done almost blind. Cochlear sensory cells that are destroyed during the process of electrode insertion result in the loss of any residual hearing. By enabling visualization of the cochlear space, the microendoscope will enable more precise location of the electrode and increase the electrode-nerve interface of cochlear implants.

• Multimodality imaging: An ideal imaging system would be able to simultaneously provide anatomical, physiological, biochemical, and molecular information with high sensitivity and specificity. However, currently, no such single imaging system exists. Different imaging modalities provide different types of information. For example, although magnetic resonance imaging provides high-resolution anatomical images (for example, of brain regions and the connections between them), it cannot image human molecular events. With positron emission tomography, on the other hand, biochemical processes such as protein synthesis and amino acid transport can be observed. Through multimodal imaging, the strengths of each of these imaging systems can be combined.

Integrating in a quantitative manner information from various imaging modalities will help better predict patient outcome. The team also discussed the importance of integrating information from not only various imaging modalities but also from animal models and -omics technology (such as genomics, proteomics, and metabolomics) to aid diagnostic, prognostic, and theranostic predictions and to serve as a training tool.

• Monochromatic X-ray imaging: Phase-contrast and fluorescence monochromatic X-ray imaging have the potential to provide nanometer-scale resolution images of deep tissue at a considerably lower radiation dose than conventional X-rays and without the use of ionizing radiation, which is known to cause DNA damage. Medical applications of this technology include obtaining “freeze” organ motion images, for example, of breathing or of the heart beat; imaging tumors, the edges between organs, and internal structures of bones; and enhancing drug dosing and delivery.

  Currently, monochromatic X-rays are generated in a synchrotron, and it is the size and the cost of this machine that are the major barriers to the dissemination of monochromatic X-ray imaging technology. Table-top monochromatic X-ray sources have not yet been developed. However, if this technical challenge could be overcome and a moderately priced, reasonably sized monochromatic X-ray source could be developed, it would revolutionize clinical imaging.

• Photoacoustic imaging: This novel technology is based on the photoacoustic effect, that is, the generation of sound from light (the conversion of non-ionizing laser pulses into ultrasound waves). It combines the resolution provided by ultrasound waves with the high contrast provided by light waves to generate images of deep structures in the body without any health risk. This technology can be used to image blood vessels or tumors or to guide biopsies.

  Other imaging technologies the team touched upon included wavefront engineering (with applications in deep-tissue imaging) and mass spectrometry imaging (for imaging and mapping biomolecules in tissue sections, for example, for identifying biomarkers for early cancer diagnosis).

1. Monitoring drug treatment efficacy: The team discussed the potential of imaging technology for monitoring drug delivery in real time. Applications of image-guided real-time drug delivery include verifying drug delivery on target, monitoring drug release and treatment effects, identifying novel drug targets, determining appropriate drug dosage, and comparing drug treatments.

2. Image-guided surgery: Image-guided therapeutic interventions result in better treatment outcomes by minimizing patient impact and reducing recovery time. For example, the team discussed how minimally invasive image-guided mitral valve repair has the potential to reduce complications associated with open-heart surgery and cardiopulmonary bypass.

For image-guided surgery to improve significantly, surgeons should be able to control the three-dimensional field of view intuitively and manipulate surgical images in real-time. Further, surgical instruments and devices that can be seen and tracked by the selected imaging technique, that have sensors for haptic feedback (to enable remote surgery), and that are integrated with imaging technology so that they have the capability to provide high-magnification, small field-of-view images are required.

As a member of the team explained, there are two parts to the problem of improving imaging technology: (1) acquiring imaging data and (2) processing that data to enhance their information content. Therefore, in addition to enumerating new imaging technologies (or new imaging applications), the team discussed current problems in archiving and processing image data.

Currently, there exists no single clinical image database that can be used to share, search, and retrieve medical imaging records.

One requirement for creating such a database is to store images in an identical format. DICOM—Digital Imaging and Communication in Medicine—is a standard imaging format like “jpg” or “tiff,” created to enable the exchange of digital image information between imaging instru-

ments from various vendors. However, DICOM is not a true standard, and inter-vendor operability continues to be a challenge (that is, there are variations in image format derived from machines manufactured by different vendors because there are variations in the way each vendor conforms to the DICOM standard).

Another issue in creating such databases is tracking the provenance and manipulation of imaging data. For example, information about the subject, machine-specific settings or parameters used to acquire the image, and how the image was processed is often unorganized and stored in files in different machines, making it difficult to reanalyze data, assess the quality or usefulness of the imaging data, or replicate experiments.

Other issues discussed in the context of image database construction were difficulties in enforcing clinical image data disclosure; image annotation; image database encryption, privacy, and security; and the need to create databases that correlate image information with biological mechanisms and that enable cross-referencing of image information provided by different modalities (for example, radiological and histological data).

The team concurred that the “full potential” of imaging lay in being able to share imaging data and discussed some applications of image processing/data mining of large sets of collected images.

• Radiation dose modulation: With regard to prostate cancer, computational image segmentation in conjunction with multiparametric magnetic resonance imaging can help determine the location of the tumor so that high radiation dose can be targeted to the tumor area, thereby minimizing the radiation dose to other areas.

• Personalized medicine: Again, in the context of prostate cancer, multiparametric magnetic resonance imaging along with machine learning or pattern recognition tools can be used to distinguish high- and low-grade prostate cancer patients (that is, determine the stage of the disease). This information can thus be used to triage patients for either surgery or enrolment in an active surveillance (or wait-and-watch) program.

• Statistical atlases for diagnosis and treatment: Statistical 3-D population-based atlases of a particular organ provide statistical information on how the structure and function of that organ vary by age, gender, and disease states in large populations. These atlases are generally constructed by integrating data from multiple subjects from different sources (for example, MRI, PET, histology). They are useful for determining normal variations present in a population. Further, because such atlases provide information

on the spatial distribution of a disease, they can help determine which part of an organ the disease is most likely to occur, which in turn will facilitate disease diagnosis and enable accurate biopsies and targeted treatment.

The team concluded its discussion by recommending the endorsement of the following policies by the National Academy of Sciences.

1. Existing programs that promote the deposition of research data in public repositories should be encouraged. Repository creation, administration, and access should be supported by the National Institutes of Health and enforced by scientific journals. Such repositories will enable the creation of an adaptive systems-medicine database and will be necessary for the statistical analysis of large-scale data. Other advantages of such repositories are better detection of disease subtypes and developing personalized treatment.

2. The development of low radiation, minimally invasive imaging for clinical use should be encouraged, for example, monochromatic X-rays. As discussed earlier, these provide contrast and resolution at lower radiation doses.

3. In keeping with the interdisciplinary nature of the NAKFI conference, grants that encourage and fund scientists in different but complimentary disciplines should be encouraged. Funding mechanisms should be leveraged to encourage interdisciplinary research collaborations among individuals with backgrounds in computation, biology, imaging, and clinical medicine.

4. Existing imaging and information technology should be harnessed to improve global health. For example, thermal imaging and T-rays (terahertz radiation) can be used to screen for infectious diseases at airports to understand mechanisms of disease transmission, and mobile phones can be used to image, share, and review global disease data.

• Stephen A. Boppart, University of Illinois at Urbana-Champaign

• Danny Ziyi Chen, University of Notre Dame

• Teng-Leong Chew, Northwestern University

• Ivan J. Dmochowski, University of Pennsylvania

• David S. Lalush, North Carolina State University

• Philip J. Santangelo, Georgia Institute of Technology/Emory University

• Joseph A. Zasadzinski, University of California, Santa Barbara

• Laura Smith, University of Georgia

Laura Smith, NAKFI Science Writing Scholar, University of Georgia

Medical imaging is an invaluable tool in the diagnosis and treatment of diseases and has the potential to provide even more extensive insight into development and effectiveness for patients with all kinds of acute and long-term diseases. However, the process of imaging is not as simple as taking a picture of the patient’s body, discerning some abnormality, and devising an appropriate treatment plan. Instead, the process is complex and involves varying components that are each essential to the smooth operation of the imaging system and the subsequent health implications for the patient. Understanding each of these elements within the overall process of medical imaging was the challenge presented to a group of interdisciplinary scientists that convened at the National Academies Keck Futures Initiative Conference on Imaging Science.

The conference brought together experts in a wide range of disciplines, from biomedical engineering to chemistry. IDR team 7B was charged with tackling the question of how to use imaging to improve the treatment of diseases. During the two days of the conference, the team developed a multi-stage approach to the challenge by taking into account the complex nature of medical imaging and disease pathology.

The key to finding a solution became apparent within the first few minutes of the initial group meeting: treatment. More specifically, the temporal aspects of treatment were considered to be a major concern and one of the focal points around which the entire model revolved. The care of a patient typically follows a prescribed path, beginning with a diagnosis. A treatment decision is then made and implemented. The efficacy of treatment is assessed after a certain amount of time has passed. In the case of cancers, for instance, treatment assessment may not occur until months after the treatment application. The goal of the team became to establish a

model that would effectively shorten, or possibly eradicate, the amount of time between these three stages.

In addition, the team decided to focus on developing a plan to aid in the acquisition of imaging data that could improve the therapy decision-making process, as well as the prediction of treatment outcomes. This is especially important when treating a disease that may require long-term management or that has the ability to metastasize, such as cancer and neurodegenerative disorders.

The outcome of two days of group deliberation was not a quick-fix or even a conclusion that just one aspect of imaging could be improved to aid in disease treatment. What arose from the meetings was similar to a business plan or model. The team honed in on what it considered the most crucial aspects of imaging to create an “integrated platform” that would improve the overall state of medical imaging and its ability to aid in the treatment process. This platform would require extensive research in certain areas and includes innovative concepts related to computation, data integration, and human observers of imaging data, technical and chemical components of instrumentation, and the patient. The platform would be used in a situation in which an abnormality or problem of some kind has clearly been identified rather than during the screening process.

The team approached the challenge by first discussing the current state of imaging technology and what will need to change to make significant improvement. Positron emission tomography (PET) has become the most efficient method of imaging on the molecular and cellular level, capturing the functional processes within the body, such as metabolic activity. More traditional modalities like magnetic resonance (MR) and computed tomography (CT) image anatomical aspects of the body. To improve the information derived from imaging systems, the team concluded that functional imaging or a hybrid of modalities (PET/CT, for example) will be needed to adequately portray activity possibly related to the patient’s disorder. The importance of functional imaging lies in the fact that chemical or biological changes related to a disease process often show up before morphological or anatomical evidence becomes noticeable. Functional imaging also provides

better contrast in images, which is crucial in discerning abnormalities, like differences between healthy tissue and abnormal tissue.

Functional imaging requires the use of an agent to create a signal in the targeted area of the body, and the team began to think about novel ways agents could be used to not only produce images, but also to aid in the treatment process as well. What it came up with was the option of a multifunctional theranostic agent, or possibly a set of agents. A theranostic combines an imaging agent with a therapeutic agent, thus combining two steps into one. More specifically, this theranostic would contain several imaging probes, but also a therapeutic agent that could be activated once in the targeted area. Such an agent would be able to provide more definitive contrast between the normal and abnormal tissue or structures, perhaps at an earlier stage in pathology than current probes can. In addition, the number of probes included would ensure activation and a signal at any sign of an abnormality. This “cocktail” of agents could also be personalized to the patient’s unique case, making a streamlined process allowing diagnosis and treatment to occur close to one another, reducing the treatment time gap mentioned earlier.

The team also considered the importance of quantization and computation in imaging, where it is sometimes difficult to obtain quantified information. However, after some debate over the importance of computer analysis in understanding medical images, the team arrived at the idea of large-scale data integration as beneficial to the disease treatment process. Although computers cannot perform all the functions a human observer can, humans have limitations that a computer does not. Computer-aided analysis could potentially change the manner in which imaging and data analysis are performed and used to decide on treatment plans for patients.

To speak to the issue of data integration the team suggested a “library” of medical images to aid in the treatment decision-making process. A massive, image-based catalog (similar to the DNA sequence database) would be available for radiologists to compare a patient’s image with other existing images demonstrating disease characteristics, matching it as closely as possible to an image in that database. The scaled, catalogued images would represent the standard of human normality, along with any deviations from that baseline according to diseases and their processes. A radiologist would be able to look at an image and determine how much, in quantitative terms,

the target is out of the norm. Although a person would make the final call, having a streamlined, computerized way of helping to determine the closest match to a patient’s condition would cut out a large chunk of the time spent on human analysis of imaging data.

Although quantization of data is an important aspect of medical imaging, the team made sure to retain the human observer as a major, vital component of the framework. Computer-aided analysis may be helpful in speeding up the decision-making process, but in the end a human will decide the meaning of the image data, affecting how it is addressed through a treatment plan. Because of this, human interpretation of data needs to be optimized through training, improved communication, and a specialized decision-maker interface between the quantified data and the human interpreter. By understanding that computers cannot perform certain tasks that a person can, the team remained grounded and focused on finding answers that could be more easily obtained in the near future after the improvement of already existing technologies and methods.

The model created by the team was composed of several interrelated layers, so it made a point of identifying all areas to be explored further in order for the model to be a viable framework to follow. It is crucial to understand that this platform cannot go forward without recognizing what is lacking in medical imaging for disease treatment and proposing areas that need further research.

The most pressing issue when it comes to advancements in functional imaging is biological target identification. If the targeted object is not an appropriate biomarker, or indicator of the desired biological process, then the data generated from that image is useless for the purpose of designing a treatment plan. This stems from limited understandings of the biological processes involved in certain diseases. Furthering our knowledge of what biological processes signal the onset or presence and progression of diseases will open several doors, allowing researchers to correctly target activities indicative of disease. Furthermore, current technology lacks the sensitivity and specificity to target these biomarkers, limiting what processes can be imaged. In addition to biomarker identification, suitable multifunctional

imaging agents would need to be developed to provide the ingredients for the theranostic cocktail.

Another area of research that must be addressed is that of extensive data synthesis. Currently there are no known efforts to integrate multiple datasets in medical imaging. Data would need to be integrated across time, scales, targets, and modalities, which would be a huge effort requiring extensive funding and organization. Without the compilation of this information, however, the decision-making process and subsequent treatment plans would remain time consuming and the interpretation of data divided. Although computer-aided analysis is an exciting concept for medical imaging, it is also one of the most difficult to put into action effectively. It would require the collaboration of many people, the development of reliable hardware, software, and computational algorithms, and acceptance of which data should be synthesized. Additionally, such integrated data would be meaningless without the human observer to visualize and communicate it to others effectively. This creates the need for a dependable interface with which to interpret and communicate the data, as well as training of those involved.

Although the technical limitations of such a model are numerous and necessary to attend to before implementation, one piece of the overall process should be kept in mind at all times: the patient. As with any medical process, it is important to take into account the needs of the person receiving treatment, as well as the demands such procedures may place on someone suffering from a disease. Human compliance is not only desired, but also necessary for such an integrated platform to aid in the treatment of disease.

The outcome of the team deliberations is not simply the wishful thinking of scientists and researchers invested in the medical imaging field. The ideas generated during the conference are forward-looking and reflect an aspiration for improvement in the overall quality of medical services. The suggested model would provide patients with more effective, personalized treatments tailored to their specific situations. It could also enhance the quality of life for patients by reducing the amount of time spent undergoing, assessing, and changing treatments. The team further predicts patient outcomes to be improved by such personalized treatment regimens. For doctors and researchers, the benefits are widespread; they will be able to bet-

ter understand disease pathogenesis and biological/physiological processes of the body. Technologically, the advancements in imaging modalities and computerized data will no doubt have spin-off effects for other areas. In addition, the medical field will benefit from the development of computational simplification and data integration by improving ways of handling large amounts of data on patients and conditions. Although time, cost and manpower may all be issues to consider in the present, the future of medical imaging holds promise for the treatment of disease and ultimately the quality of health care and patient health.

• Rigoberto C. Advincula, University of Houston

• Stuart S. Berr, University of Virginia

• Frank Chuang, University of California, Davis

• Allan V. Kalueff, Tulane University

• Philip R. LeDuc, Carnegie Mellon University

• John D. MacKenzie, University of California, San Francisco

• Mark J. Schnitzer, Stanford University

• Andrew Tsourkas, University of Pennsylvania

• Alexander Walsh, University of Southern California

• Andrew Z. Wang, University of Northern Carolina

• Nadia Drake, University of California, Santa Cruz

Nadia Drake, NAKFI Science Writing Scholar, University of California, Santa Cruz

In considering how imaging could be used to improve disease treatment, IDR team 7C chose cancer as a disease model. The team’s goals were three-pronged: streamlining diagnostic processes for patients by developing multimodal, multiplexed imaging; improving treatments by identifying imaging markers correlating with good or bad outcomes; and making these proposed technologies inexpensive, portable, and accessible to all patients.

At present it can sometimes take months to diagnose and begin treat

ing cancer. Between the first suspicion of a tumor to initial treatment there may be an arduous diagnostic journey. Patients may require more than one kind of image before the size or spread of a tumor can be determined. Biopsies are usually performed. Then, a treatment plan is prepared, often limited to a standard protocol that generally cannot yet be tailored to the specific patient or to his or her specific tumor. Follow-ups happen every three to six months because technologies aren’t sensitive enough to detect small numbers of regrowing tumor cells.

In an ideal world, one envisioned by the IDR team, the process of diagnosing and monitoring tumors could be significantly improved, and made more efficient, if new, highly sensitive imaging technology could identify and monitor minute changes in tumor status or spread.

What if, instead of months, the process took days? A tumor is suspected—and a small, portable instrument with multiple imaging capabilities detects and characterizes the tumor that same day. Treatment is based on a detailed dataset containing outcomes for specific biomarkers within a specific lesion, courtesy of finely detailed images that help clinicians characterize the tumor. Follow-ups happen regularly and at home, with remote, continuous monitoring that is non-invasive and sensitive enough to provide a clear picture of even small changes in a tumor or detect metastatic tumor cells traveling in the blood.

Instruments enabling sensitive and efficient ways to diagnose and treat disease might be on the horizon. Such diagnostic and therapeutic innovations might help not only with cancer, but also other conditions such as neurological disorders and serious infections.

Team 7C comprised scientists of all stripes. One team member uses video tracking to study how zebrafish respond to a variety of drugs. Another attaches tiny microscopes to mice and studies the neural correlates of behavior. A third tracks tumors using immune cells. Another studies nanoprobes, which might be useful as new diagnostic or therapeutic tools. They all have an interest in imaging. They all spent two days pooling their collective experience and imaginations to propose an answer to the team’s challenge: Find novel ways to use imaging methods to improve the treatment of diseases.

Initial thoughts on using imaging to improve disease treatment were as varied as cell surface markers.

One idea was to do the inverse of conventional post-treatment tumor

imaging and develop an imaging agent that identifies unresponsive cells. Another idea was to develop computers recognizing histological patterns—and biomarkers—in a high-throughput manner. A third suggestion was to develop synthetic cells based on fluorescent cell mapping—and in that way, develop a clearer picture of tumor cells and how they might respond. Yet another suggestion was something similar to the Star Trek tricorder—a hand-held scanner that quickly identifies anomalies.

Common to all these ideas? The marriage of technology and clinical application, and the need for multi-functional detection and imaging systems.

Team members cautioned against developing a new technology doctors can’t use or understand, saying clinical utility should be paramount in any design process. “We need to give some kind of meaning to it, get the clinical side to validate an imaging technique,” a team member said.

Multifunctional imaging systems will improve the speed with which diseases are identified and diagnosed—and also provide more specific information about what a patient is facing. For example, the reagents used in certain types of imaging—positron emission tomography (PET) scans or magnetic resonance imaging (MRI)—are different. The machines are different. The type of data produced by the images is different.

What if imaging modalities, like MRI and PET, could be combined into a single technology routinely providing both anatomical (MRI) and molecular (PET) information? Dual PET-MRI has just been developed and is not widely used. Although seemingly ideal, combining modalities runs into serious stumbling blocks, including finding plausible contrast agents. Such agents are used to increase the visible distinction between different tissues or structures, and can be injected, ingested, or inhaled.

Technologies like computed tomography (CT) use contrast reagents in the form of dyes typically containing iodine or barium. PET scans use radioactive biological analogs to provide images of functioning tissues. MRI enables visualization of sub-surface structures with the help of paramagnetic gadolinium-containing contrast reagents.

“Can we find a contrast agent that will work for both CT and MRI?” a team member asked. Reagents are needed that can be used either sequentially or in parallel, and although no answer readily presented itself, the team considered the possibility of nanoparticles fitting the bill.

What should be considered for multiplexed, multimodal imaging development?

Modalities refer to different imaging systems—those that use different target molecules as sources of data. For example, magnetic resonance records a proton signal. CT scans record X-ray attenuation. PET scans look at electron-positron annihilation events. Ultrasound relies on different tissue densities reflecting sound waves. “Multiple markers might not be a big challenge, but seeing them simultaneously is.”

Integrating methods requires simultaneously detecting all the different contrast agents or markers in play and making sense of the data they’re producing. That also requires that different markers play well with another.

“Energy conversion systems” might easily lend themselves to multimodal integration. These systems involve using things like light and sound as contrast agents—meaning tissues respond to each element differently, allowing a detector to convert the contrasting signals into a picture. Photoacoustic imaging is one example: when light waves are projected into the skin, some are converted into heat and then into ultrasonic waves, which return a highly detailed image, different from the somewhat fuzzy ultrasound images we are used to seeing (ultrasound sends sound waves into the skin). Piezoelectric detection, another form of energy conversion, takes advantage of related electrical and mechanical properties. Traditionally, mechanical stress applied to a material can induce electrical activity, which can be detected and used for imaging.

Multiplexing probes means using the same mode to measure different markers. In theory, a unique probe could be made for different biological markers—either at a subcellular, cellular, or system level. Then, one detector could be used to image all the probes and provide a more integrated picture than a single probe allows.

For example, if a subcutaneous fluorescent imager could “see” many different colors of injected fluorescent markers—each attached to a different type of cell or tissue—then it could weave together a more instructive image than one simply looking at, for example, green-tagged ovarian tumor cells.

Two or three or five multiplexed probes could help researchers extract a lot of information from a single imaging procedure. For instance, if yellow-tagged blood vessels were in the image along with markers of the tumors’s genes, then physicians could simultaneously study tumor vascularization and genetics. If blue-tagged epithelial cells were visible, then physicians

would know whether the mass was epithelial in origin—and likely cancerous. If orange-tagged germ cells were detectable, then clinicians would know whether the mass might be a benign germ cell tumor. Combing many probes could provide information about which biomarkers are associated with cancerous growth and enable correlations between biomarker presence and clinical outcome. It would be like a cocktail of different probe molecules, each with a unique signature.

Team members also considered whether CT scans could identify different nanoparticle densities—or even make use of Mossbauer probes.

In essence, these proposed imaging technologies would facilitate a level of analysis approaching in vivo cytology and provide an instant, accurate picture of what’s going on beneath the skin’s surface.

Then, that information could be included in detailed datasets about how different biomarkers—identified by different probes—correlate with treatment outcomes. Morphology, proteins, cell cycle alterations, apoptosis, immortality, location, and gene expression and sequence could all be monitored. Instead of two-dimensional datasets—like those considering drug dose and response—team members proposed including the above markers and additional variables like age, sex, time, treatment, genetic information, and environmental factors. This way, imaging could enable more tailored therapies, instead of the one-size-fits-all generic treatment blanket currently covering treatment options. Are we going to reach the level of personalized medicine? Maybe one day. But for now, more specificity would be a good start.

Post-treatment monitoring of metastatic potential could use noninvasive imaging systems that track labeled tumor cells, for example, before they begin cancerous regrowth. The group suggested that constant monitoring—on a detailed scale—is necessary for both the patient’s and the clinician’s peace of mind. New multimodal monitoring of blood vessels for metastatic tumor cells would offer a kind of constant vigilance that would be more reassuring and medically beneficial than the staggered, routine three- to six-month follow-ups made necessary by current limitations in finding tiny numbers of migrating cells.

In addition to simply creating technologies allowing multiple imaging modalities and the monitoring of specific biomarkers, team 7C considered the accessibility of such technologies.

Ideally, instruments would be small. Portable, even. And inexpensive.

Like the Star Trek tricorder, for example—a hand-held device used to scan the body and detect aberrations. What if something like that existed? Something not much larger than a cell phone, with an image screen, that could detect tumors on the spot? Instead of suspecting a tumor—and, in some cases, having no idea where it is—the “tricorder” could assist in determining location, cell type, and prognosis.

Team 7C agreed such a device would be ideal—and maybe even possible. But there are substantial technological hoops to jump through. And developing it would take time. At least 25 years. It’s not on the immediate horizon, but within the realm of possibility.

Post-treatment monitoring could be accomplished by equally non-invasive, portable devices: think cell phones, goggles, transcutaneous patches, and fingertip scans—like those used to measure blood oxygenation today. What if organs, tumors, or cells were labeled with a detectable marker, and a smart phone app could turn the phone into a detector? Post-treatment monitoring for tumor regrowth would be possible at home. And results could be text-messaged to a clinician.

Similarly, monitoring the blood for unwelcome travelers (metastatic cells) could be done by taking advantage of the eye’s or skin’s relative transparency. 3-D goggles could simultaneously scan retinal blood vessels and provide an exciting cinematic experience (Star Trek in 3-D?). A transcutaneous patch or fingertip scanner could be worn overnight and provide data in the morning about whether anything unwanted is running around in the blood. And, in theory, these types of remote monitoring instruments might be possible in about a decade.

Every journalist is familiar with the inverted pyramid—and team 7C is, too. In fact, so are most conference attendees. While giving the team’s preliminary report, Andrew Tsourkas pointed to an inverted pyramid—it held broad symptoms of cancerous lesions on top, and drilled down to molecular specifics at its point. The point was to represent new ways of

thinking about disease diagnosis and treatment and blurring the lines generally separating levels of disease characterization. Alas, Andrew was accused of not knowing what a pyramid looks like, as the one the group developed was situated on its head.

But team 7C was ready to defend itself: “Building a pyramid right is easy. Building it upside down is impossible. Do the impossible,” they said. “Think big.”