Recent advances in the fields of medical physics, applied mathematics, computer science, and biomedical engineering have attracted the attention and interest of investigators from many disciplines. In the area of neoplastic disease, principles of biomedical image science are being applied to the development of diagnostic methods, prostheses, minimally invasive therapies, and radiation oncology. The potential of these advances and new technological strategies to improve patient care is great. Development and application of new image-based methods of tissue identification and characterization and multimodal image display technologies, are needed to improve detection of neoplastic disease, as well as its staging and treatment.
New avenues of scientific inquiry may enable the development of a novel set of imaging technologies that are designed to perform ideally in their respective biological environments. The disciplines that are needed to achieve this include medical physics, applied mathematics, biomedical imaging science and biomedical engineering. These fields overlap and are complementary to one another.
To determine how best to facilitate development in these areas, the National Cancer Institute Diagnostic Imaging Program convened a workshop on July 17-18, 1997 which brought together experts in the fields of biomedical imaging sciences and diagnostic radiology as well as representatives from the National Cancer Institute, other federal agencies and industry. These experts were charged with developing recommendations that will provide a foundation for a research program in those areas having specific application to the mission of the Institute.
The workshop was divided into two strategic parts. During the first part, overview presentations of research needs and opportunities were made. In the second part, seven task forces were defined to address general issues that pertain to all areas of biomedical imaging in oncology research and clinical practice.
This report summarizes the results from one of those groups focused on "Matching clinical and biological needs with emerging imaging technologies". The imaging methods used in biomedical applications that this report discusses include:
- X-ray projection imaging,
- X-ray computed tomography (CT),
- Magnetic resonance imaging (MRI) and magnetic resonance spectroscopy,
- Single photon emission computed tomography (SPECT),
- Positron emission tomography (PET),
- Electrical source imaging (ESI),
- Electrical impedance tomography (EIT),
- Magnetic source imaging (MSI), and
- Medical optical imaging,
This is one of seven task forces assembled to respond to the issues and challenges identified at the NCI Imaging Sciences Working Group meeting on 17-18 July 1997. This task force, "Matching Clinical and Biological Needs With Emerging Technologies", seeks to identify ways to stimulate the development of emerging technologies (optical imaging, MR spectroscopy, intermodality image fusion, ultrasonography, image analysis and display and data management). It is intended to delineate means that coordinate the development of these technologies consonant with the needs of the clinical and biological sciences, and supported by partnerships between industry, academia, clinical practitioners, and government.
This task force has been assembled by invitation to include experts worldwide on biomedical imaging and image processing. The first meeting of this group was held at the Marriott Suites O'Hare on 30 November 1997 near the start of the annual Radiological Society of North America meeting. Twelve experts from more than 25 invitees were able to attend, as noted in the appendix. Three NCI representatives were in attendance as well.
Early detection and diagnosis are critical to increasing one's chances of surviving cancer. Although we cannot, as yet, prevent many types of cancer, early detection through the use of imaging can reduce cancer deaths. However, many lesions detected by imaging are benign resulting in many unnecessary medical procedures including biopsies. Additionally, many cancers are not detected. For example, mammography misses about 15% of cancers. Thus, improved early detection and clinical management depends upon improvements in imaging technologies to detect, diagnose and treat cancer.
A major goal of the NCI Diagnostic Imaging Program should be to coordinate and encourage government-industry-academia collaborative efforts to advance cancer imaging through facilitated identification, evaluation, and transfer of promising and relevant imaging technologies.
To facilitate identification and support of new technologies and concepts, a Problem Statement was developed and distributed, defining the current state-of-the-art in cancer imaging and the technological needs for improved image-guided cancer diagnosis and treatment.
The guiding principles for matching emerging imaging technologies and biomedical needs were described as not that much different from the approach used to solve contemporary problems in quantitative non-destructive testing of inanimate objects. The suggested approach is problem oriented and model based where a set of pertinent specifications are posed in terms of detectability limits and accept/reject criteria. Models are based on the first principles from the physical sciences and engineering that govern imaging instrument performance and limits.
Models are formulated in an abstract setting and are ultimately matched to individuals or groups. The Visible Human Project was described as an example of modeling, where more or less typical examples are augmented by experts in anatomy who delineate and label structures to form an electronic anatomic atlas. The fusion of medical images with non-image knowledge bases was considered advantageous despite the formidable technical and computational problems that are encountered in implementing them.
Potential synergy with other ongoing efforts in biomedical research and development, such as the Human Genome project and especially the NCI Cancer Genome Anatomy Project were identified. This was most appropriately done using knowledge fusion where structural and non-image attributes are applied to explicitly and implicitly map relationships in data bases. Anatomic atlases specific for cancer have not emerged at the scale and extent seen in the Visible Human Project. The development of electronic atlases that accommodate abnormalities such as solid tumors is important.
The issue, "Do we need new imaging modalities?" was discussed and contrasted with the question, "Do we have the tools that we need to effectively use the images we already collect?". Breakthroughs in imaging based on introduction of new modalities would be welcome, but cannot be predicted reliably. The inability of the present funding mechanisms to sponsor the advanced prototype development of promising new imaging modalities and methods was attributed to the use of review criteria that strictly consider biological significance of a project rather than its potential significance, if successful, in a broad spectrum of imaging related problems. In other words, the organ, disease and system approach which lies at the base of NIH and especially NCI does not make imaging system development straightforward since support is gained more by the merits of the system's applications than its intrinsic capabilities or promise. Both MRI and CT scanning, which have obviously revolutionized biomedical imaging in the past 3 decades, were not aggressively supported by the biomedical research establishment in the USA during their early development and were largely developed in the United Kingdom before their adoption in the USA became obvious and necessary.
The relationship of tumor size (especially
The synthesis of specifications for advances in cancer imaging is predicated in better use of a priori knowledge gained by quantification of known abnormalities and setting of accept/reject criteria.
Computer aided diagnosis was emphasized as promising for better utilization of already available imaging data, and tailored approaches have been useful in finding missed lesions for breast and lung cancer on conventional radiographs.
Assembly of an anatomy-based experiential database and training of this system to capture the knowledge available in human practitioners was viewed as a priority. The planning, delivery and results of therapy can be unified in the experiential database to guide further advances and "close the loop" of diagnosis and therapy that is now largely open. The potential benefits of the closed loop approach where experience reinforces and improves performance over time are substantial and extensive. Databases have been developed for imaging, but there is no coherent "standard" approach to their development and integration has emerged. The potential synergy obtained by unifying the knowledge across databases is substantial, and NCI support of this effort would be highly productive.
There is an acute need for better and more robust anatomical reference systems that allow multiple encounters with a single individual across modalities to be unified and applied. The problem is even more complex and difficult when multi-modality population-based differences are to be assessed. The variability of anatomy and its importance in cancer diagnosis and treatment was emphasized, particularly in the prostate and lung where lymphatic drainage patterns that we cannot reliably image now may be inferred from structural data that can be obtained in images.
This group recommends that: 1) image quantification methods for oncology that lead to objective specifications of performance needs are the basis for progress, 2) Translation of cancer imaging needs in research and clinical applications into quantifiable criteria that can guide and measure progress has not been done, but should receive high priority, 3) The relative lack of support to emerging biomedical methods in the past so offshore development of new technologies is the rule rather than the exception should be reversed, 4) Probabilistic model-based image reconstruction and analysis methods that fully incorporate a priori knowledge, including non-image data has promise to move the field forward rapidly, and 5) integration across centers which possess unique data bases to fully exploit their potential synergy must be accomplished.
A summary of the recommendations of the working group/task force and the implementation strategies are presented below.
Imaging of Breast Cancer
Future research in breast imaging will involve evaluating the molecular aspects of breast cancer. To achieve these innovations, collaborations among researchers in radiology, pathology, and molecular biology are necessary. Three areas of active investigation are mammoscintigraphy, digital mammography, and breast magnetic resonance (MR) imaging.Nuclear Imaging of the Breast
- Can provide diagnostic information that is complementary to that given by conventional mammography.
- Requires development of dedicated, small field-of-view imaging devices.
- Breast imaging with use of positron-emitting radioisotopes is based primarily on assessment either of glucose metabolic rate via FDG or of estrogen- or progesterone-receptor density
- Future developments will enable pre- or intraoperative localization, functional characterization of breast lesions, and patient follow-up during and after therapy.
MRI of the Breast
- Image-processing and pattern recognition algorithms will be used to improve lesion detection and characterization
- Full-field digital mammography devices are undergoing preliminary clinical testing.
- The development and exploration of 3D display through tomosynthesis or stereoscopic acquisition should be encouraged.
- Optimal soft-copy displays should consider the following: (a) user-friendly roam and zoom capabilities, (b) on-line image processing and computer-assisted.
- A high priority should be experimentation to determine what spatial resolution is needed to optimize breast lesion detection and characterization.
- Other detectors (eg, that use selenium, cadmium zinc telluride, or lead iodide) might offer improved diagnostic accuracy.
- Multi-center clinical trials of whether digital mammography can improve on the sensitivity and specificity of screen-film mammography
Imaging of Genitourinary CancerOvarian Cancer
- Contrast-enhanced MR imaging of the breast has been shown to be highly effective
- The most important research area for breast MR imaging is contrast material enhancement.
- Current gadolinium chelates suffer from a lack of specificity
- Clinical tools could be developed for non-invasively evaluating lymph node involvement and identifying micrometastases and residual microscopic disease
- Clinical studies should include evaluation of
- suspicious breast lesions
- the accuracy of local extent of breast cancer
- MR screening of high-risk populations.
- Minimally invasive therapy and diagnosis will include
- MR image-guided breast biopsy
- MR image-guided ultrasound ablation and
- MR image-guided tumor excision
- imaging-guided biomarker or tumor-specific delivery of pharmacologic, chemosensitizing, or radiosensitizing agents to tumors
- Early detection of ovarian cancer remains an important priority.
It is the third leading cause of cancer death in women aged 35-74
Ovarian cancer is metastatic at the time of diagnosis in 70% of patients
Current methods of screening are inadequate
Ultrasound is sensitive but not specific
Several genes associated with an increased risk
- Examples of imaging developments that are being pursued:
Fat-suppressed MR imaging with oral and intravenous contrast media,
Estrogen receptor agents tagged with a radioisotope
Monoclonal radiolabeled antibody imaging
- "Cancer markers" (eg, prostate specific antigen, carcinoembryonic antigen) are needed
- Genetically high-risk individuals will need screening
- Imaging-guided therapy of prostate cancer is promising
- Methods of accurately depicting the extent of the tumor must be developed.
- Doppler US with or without contrast agents warrants study
- Endorectal coil MR imaging may be used for local staging
- Phased-array coils are less expensive, more comfortable and may replace endorectal coils
- Major problems include the lack of specificity of some of the anatomic findings and the inability to identify microscopic spread.
- Genetic screening will identify high risk patients for more intensive imaging
- Imaging agents that are specific for renal cancers are needed
- CT and MR imaging can be used to detecting and characterizing small renal masses
- Numerous small lesions that cannot be accurately characterized
- High-resolution targeted CT, PET, MR imaging, or MR spectroscopic imaging and ultrasound may be able to depict more accurately the character of a lesion.
- High-resolution MR imaging and intravesicular US may be used for bladder cancer staging.
- Dynamic, enhanced MR imaging with new contrast agents are needed
- 3D virtual endoscopic techniques will aid urologists in directing biopsies
- Endoluminal and sonographic evaluation of the ureter and renal pelvis will define the extent and invasion of tumors of the urothelium.
Cancer involving Lymph Nodes
- Methods of staging with MR imaging are at a historic high "watermark,"
- Hysterosonography for uterine cancer will need to be evaluated
New Contrast AgentsCT and MR Cancer Imaging
- The "Holy Grail" of lymph node imaging is an agent that enables positive identification of metastatic nodal disease but will not accumulate in normal nodes.
- Current methods of assessing metastatic lymph node disease are inadequate.
- Assessment of volume and perhaps shape may help
- Dynamic enhancement of CT or MR images is needed to detect nodes that enhance similarly to that of the primary tumor.
- MR lymphangiographic contrast agents can be used
Cost containment for CT and MR imaging will be a prevailing concern.CT Developments
Areas of development in CT include:
a. innovative applications of existing technology
Substantially reduced-cost CT could replace traditional radiography or CT expense could be offset by reduced stay and reduced morbidity
b. improvements in data display
3D images will be interactive, readily available in the operating suite, and integrated with surgical instruments in a dynamically changing display
c. improvements in data acquisition
Faster acquisition speed
Better spatial resolution
Improved tube and detector design
d. utilization of contrast media
The search for a nontoxic, organ-targeted contrast agent has been the "Holy Grail"MRI Developments
Blood-pool labeling agents and hepatocyte-specific agents will be developed
A. X-ray projection imaging
- Development of electronic planar array detectors with adequate resolution, size, reliability, and quantum efficiency.
- Development of digital display systems of sufficient resolution and dynamic range.
- Development of real-time 3D fluoroscopy systems for angiographic and interventional use.
- Development of means to detect and use the information in scattered radiation, including mathematical correction schemes. Further understanding of the role of scatter rejection and optimization of the x-ray sources used in digital mammography.
- Development of monochromatic sources, novel x-ray generation techniques, and holographic methods.
- Performance of displays is a limiting factor in the clinical utility of full-breast digital mammography. Current soft copy displays are resolution limited (2K x 2.5K pixels) and provide inadequate contrast and light output. This application requires development and testing of cost-effective digital displays for high resolution (e.g. 50-100 microns per pixel), high contrast (about 12-14 bits), large field of view visualization (4K by 4K, or 2K by 2K with 4K by 4K buffer) combined with practical rate of display and luminescence.
- Key to enhancing interpretation of digital mammograms is determining how to display the most important information in the image in the best (and fastest) possible way for the clinician. This requires development of computer workstations with practical user interfaces for clinical radiologists (e.g. multi-resolution, "region-of-interest" displays and "bright light" display equivalents).
- Development and implementation of novel cost-effective x-ray detectors in prototype digital mammographic systems requires 18 cm x 24 cm image size and 50 micron spatial resolution. Several such detectors are being prototyped and cost effectiveness is a key issue; current mammography equipment sells for around $100,000 and the cost per procedure is $55-$100.
B. X-ray computed tomography
- Increased instantaneous and sustained x-ray power capabilities in both conventional x-ray tubes and electron beam systems.
- New and innovative technologies for the required high-intensity x-ray source, beyond those currently available.
- Two-dimensional detector arrays encompassing a larger solid angle and allowing improved spatial resolution along the scanner axis, including the associated high-throughput data acquisition electronics.
- Decrease in image reconstruction times through high performance low-cost processors and the increased use of Fourier domain-based and iterative reconstruction schemes.
- Mathematical means for utilizing the information contained in scattered photons.
- Effective means for the correction of patient motion-related artifacts.
- Better and easier-to-use three-dimensional data reduction and visualization techniques.
- Monochromatic x-ray sources with high brightness.
- Detectors that provide information on angles of incidence and high energy resolution.
- Algorithms for image reconstruction at reduced X-ray radiation exposure.
- Techniques for 2D and 3D CT fluoroscopy in fan-beam and cone-beam geometry to guide medical interventions.
- Practical and effective algorithms for metal artifact reduction.
C. Magnetic resonance imaging (MRI) and magnetic resonance spectroscopy
Development of cost-effective MRI technologies, including but not limited to innovative contrast-enhancing agents, organ or body region-specific magnets, new or improved pulse imaging sequences, and refinements in RF coil design.
- Magnet Systems
- Development of economical higher-temperature superconducting magnets, using Nb3Sn and other higher-temperature materials.
- Development of designs for economical magnets for special applications, including specific anatomic parts, or for special disciplines such as therapy.
- Development of designs for more economical magnetic field shielding that would allow retrofitting of magnets into existing diagnostic, interventional and surgical rooms.
- Pulsed-field MRI
- Development and validation of strategies for signal readout that minimizes introduction of additional noise and interference.
- Development of means for energy recovery during collapse of the polarizing field.
- RF Coils
- Design of uniform transmitters and receivers that include dielectric and wavelength effects.
- Development of methods for three-dimensional modeling of RF fields.
- Design and evaluation of high-temperature superconducting coils and associated cooled pre-amplifiers for low-field imaging.
- Design of SNR-efficient high-speed image combination and reconstruction techniques for multi-coil arrays.
- Gradient Systems
- Design of very short head gradient coils using current return paths at a greater diameter than that of the primary windings.
- Design of acoustically quiet gradient coils.
- Design of head gradient coils with good subject access for visual and auditory task presentation.
- Optimization of head immobilization devices compatible with head gradient coils.
- Development of methods for electrical decoupling of RF coils and gradient coils in close proximity.
- There is a fundamental trade-off between spatial and temporal resolution (fine detail versus rapid sequence scanning). The development of high performance gradients will allow higher temporal resolution with high spatial resolution.
Dynamic MR Imaging
- Alternative image reconstruction methods to the FFT, including maximum entropy, wavelets, Bayesian, and search-everywhere neural network methods
- Blood Flow
- Development of techniques for rapid measurement of instantaneous velocity in three-dimensional space (six-dimensional problem)
- Modeling of complex flow and its implications for the vascular MR signal
- Development of methods for extracting parameters of physiologic relevance, including shear stress, distensibility, and turbulence
- Diffusion/Perfusion Imaging
- Establishment and experimental validation of a molecular model for anisotropic diffusion in tissues
- Development and validation of methodology for quantitative perfusion measurements
- Development of improved approaches for spatially localized measurement of diffusion coefficients in vivo
- Modeling of heat dissipation in tissue
- In vivo measurement of tissue fiber orientation
- Other Tissue Parameters
- Development of accurate measurement techniques for quantitative relaxation times and their interpretation in terms of clinical diagnosis
- Development of sophisticated segmentation techniques based on multiple parametric acquisitions
- Extractions and meaningful display of strain maps of cardiac function
- Modeling and verification of peripheral nerve stimulation
- Functional MRI
- Development of methods for monitoring data quality during a scanning session and for ensuring that functional activation is being observed.
- Development of new experimental protocols, especially for use with complex stimuli (e.g., visual presentations).
- Modeling and experimental verification of the biophysical contrast-to-noise mechanisms induced by neuronal activation and their dependence on magnetic field strength.
- Establishment of detailed biophysical models to understand the stimulus response permitting separation of vascular from parenchymal changes and arterial from various changes, taking into account parameters such as blood volume and flow.
- Evaluation of noise sources (e.g., stochastic, physiological, instrument instability) and development of strategies for their minimization.
- Multinuclear MRI
Evaluation of polarized noble gases as tracers of pulmonary function and tissue perfusion (e.g., in muscle, brain and other organs).
- MR Microscopy
- Exploration of the theoretical limit of spatial resolution and its dependence on key parameters, including diffusion and detection sensitivity.
- Development of improved means for monitoring of and correcting for the effects of motion, which currently limit resolution of in vivo MR microscopy
- Commercially available MRI systems do not image the microcalcifications (less than 300 microns in size) that indicate the presence of small ductal carcinoma in situ of the breast. MR microscopy may provide a solution to this well recognized limitation.
D. Emission Tomography, including SPECT and PET
- Development of detector technology including -
- Investigation of new room-temperature semiconductor materials, combination of scintillators and solid-state detectors, and readout electronics for use in single photon imaging systems
- Methodological search for new scintillator materials with high light output and fast photo-fluorescent decay time for PET systems
- Development of digital scintillation cameras that use a discrete scintillator matrix with high counting rates and improved imaging capabilities, including improved stability, uniformity, energy resolution, and intrinsic resolution
- New scintillators with enhanced light output and fast decay times
- Depth of interaction determination
- Improved count-rate capabilities
- Time of flight information
- Multiple energy windows for scatter correction and transmission scanning
- Development and implementation of room temperature solid state detectors with enhanced energy resolution.
- Development of miniature gamma cameras for interventional procedures.
- Further development of SPECT/PET hybrid systems.
- Improved count rate capability
- Increased intrinsic efficiency
- Development of improved ECT systems
- Improve transmission scanning in both SPECT and PET
- Special purpose systems
- Small animal systems
- Breast imaging systems
- Development of high efficiency SPECT systems
- Cone beam collimation
- Compton scatter cameras
- Coded apertures
- In research on image reconstruction methods, the most promising areas of investigation include -
- Development of analytical solutions for the SPECT and PET reconstruction problem that include the effects of non-uniform attenuation, spatially varying scatter and spatial resolution
- Development of fast and stable iterative reconstruction methods that incorporate non-uniform attenuation, spatially varying scatter and spatial resolution for accurate SPECT and PET image reconstruction, and
- Development of new three-dimensional reconstruction methods and data acquisition strategies for PET and converging-beam SPECT.
- Wider availability of metabolic and receptor targeted radiopharmaceuticals optimized for breast, prostate and other cancer detection, staging and treatment monitoring; including positron emitting radiopharmaceuticals for PET imaging and single photon emitting radiopharmaceuticals for gamma camera imaging.
- Cost efficient PET systems designed for whole body use as well as devices targeted for specific purposes such as lymph node staging.
- Cost-effective x-ray detectors with high spatial, time, and energy resolution and the capability for measuring the angle and depth of interactions
- Effective and fast means of acquiring low-noise transmission data
- Mathematical techniques to take advantage of improved detector technologies, such as the possible inclusion of time-of-flight information into the reconstruction process
- Fast three-dimensional reconstruction algorithms, in particular for fast dynamic or whole-body studies
- Fast and quantitatively correct iterative reconstruction algorithms for two-and three-dimensional reconstructions
- Inexpensive and powerful reconstruction processors to accommodate parallel iterative reconstruction methods
- Fast real-time sorting electronics for data acquisition, efficient data storage and handling capabilities for the vast amount of projection data in three-dimensional and whole body studies.
- Development of effective wide-band models that include the physical acoustical properties of the tissues in which the beam is propagating, such as absorption and aberration-inducing effects. This is effectively the forward wave propagation problem.
- Further study of the dielectric and ferroelectric properties of materials, with particular emphasis on the development of new ferroelectrics with high dielectric constants, such as the relaxer ferroelectrics; of ferroelectric functions; and of the role of grain and domain boundaries in the piezo-and ferroelectric properties of materials.
- Development of four-dimensional finite-element models of single transducer elements and multi-element arrays.
- Development of super-broadband transducers, array modeling, and effective modeling of multi-mode vibrations of transducer elements.
- Investigation of theories on development of faster beam-forming algorithms that will be flexible enough to allow adaptive corrections for phase and amplitude distortions attributable to both tissue and systems effects. Development of algorithms that ensure that diagnostic information can be obtained from any one patient regardless of intervening tissue components.
- Development of statistical and/or adaptive methods to preserve image detail associated with high-level echoes (such as those that occur when surgical tools interact with tissue) and sharp boundaries around vessels and ducts while improving contrast detail associated with low-contrast or small objects. Development of contrast enhancement routines based on knowledge of underlying tissue types (both healthy and diseased) and their interaction with ultrasound. Investigation of genuine contrast improvement procedures through tissue characterization/parametric imaging (including texture analysis, and tissue elasticity).
- Development of algorithms based on known differentiation between tissue types and between normal and diseased states.
- Improvement of methods for three-dimensional imaging capabilities. Development of better methods for displaying three-dimensional flow data in a format that is easily comprehensible. Development of image processing routines that overcome the specific problems of three-dimensional ultrasound (e.g., acquisition issues, coherent image formation (speckle), inhomogeneous media, non-transmission modes) and thus allow optimal utilization of existing three-dimensional reconstruction techniques. Development of methods for regional segmentation as a preprocessing step to measurements of both area and volume, and application of three-dimensional image data display methods and automatic methods for measuring areas and volumes.
- Development of future imaging systems that use scattered waves more effectively. The present pulse-echo systems use a fractional bandwidth greater than 50%. Because the waves are basically oscillatory and coherent, the images show "speckle," the multiplicative random noise noted in laser images. If a wider bandwidth could be obtained by extending the response to low frequencies, a speckle-free system could result.
- Development of a solution for the inverse problem to include the complex viscoelastic wave equation and its solutions for both transmission and reflection tomography.
- Development of high resolution/high frequency systems to detect smaller masses and provide information on the matrix, margins, and vascularity of lesions. There is a fundamental trade-off between high and low contrast resolution and penetration. Heating of the tissues is also a potential problem with high-frequency, deep imaging of the soft tissues in 3-D. Heating and frame rate problems can be ameliorated by creating several to many image lines for each transmitted pulse.
- Current transducer technology permits variable focusing only within the imaging plane, although arrays with a limited number of elements in the image plane thickness direction are beginning to reach the market. The development of 2D and 3D imaging equipment using 2D arrays could offer new methods for full breast examination using ultrasound and give increased likelihood of success in the most promising area for increasing ultrasound imaging of the breast, i.e. correction for the phase aberrations.
- Current ultrasound technology has a field of view limited to several centimeters at maximum resolution, making full breast examination difficult and time-consuming. This is in part a result of current use of ultrasound for examining already-suspected masses. At the present time, longer arrays are technologically quite feasible at little extra cost and frame rates would be acceptable with the new high-end systems.
- Development and testing of color flow Doppler and/or the use of contrast agents to assess abnormal vascularization in the vicinity of the lesion. For example, use of ultrasound to image the harmonics associated with gas-filled microspheres injected into the bloodstream.
- Development of ultrasound motion analysis and sonoelasticity techniques to improve tissue characterization and differentiation of lesions from normal breast tissue.
- Development of 3-D spatial compounding and image segmentation for delineation of detailed parenchymal patterns that correlate with histological findings, for example to discriminate the lobular, ductal and connective tissue structure of the breast.
- Development of perfusion imaging or imaging of fractional blood volume, mean transit time and wash-in, washout times for distinction of highly active masses, including, possibly, the more aggressive cancers.
F. Electrical Source Imaging
- The inverse computation of cerebral cortex and epicardial potentials requires knowledge of the geometries and conductivities of intervening tissues. It may not be practical, in an experimental or clinical setting, to even measure the position of each body surface electrode. Some convenient strategy of individual-specific anatomic measurement (e.g., by MRI) needs to be developed so that the true geometry and inhomogeneous conductivity can be incorporated into the reconstructions of potentials.
- Resolution must be improved to the point that details of conductive disturbances (e.g., areas of slow conduction in the heart) can be detected and localized. This could be achieved by modifying the mathematical formulation to incorporate higher-resolution temporal and spatial information into the regularization procedure. This would include development of fast algorithms for computing and displaying reconstructed information at close to real time.
- The accuracy of the inverse procedure is dependent on the number and position of the surface electrodes. Methods are needed to characterize the optimal number and positions for expected noise conditions.
- Although potentials can be reconstructed with good accuracy using Tikhonov inversion techniques, a significant improvement is achieved when the time progression is incorporated into the regularization procedure for the cardiac inverse problem. These techniques could be extended to allow similar improvements in studies of the brain and of heart activation.
G. Electromagnetic Tomography - (Electrical Impedance, Microwaves, Conductivity, Terrahertz)
- Develop computable mathematical models for tomography that accommodate multi-path incoherent diffusion in transport of radiation through heterogeneous biological objects.
- Undertake clinical research to determine the role of impedance imaging for applications such as determination of cardiac output, monitoring of pulmonary edema and screening or tissue characterization in breast cancer.
H. Magnetic Source Imaging
- For magnetic source imaging to become an acceptable procedure, the cost of biomagnetic instrumentation must first be reduced. The most expensive component is the screening against external interference, given that the necessary soft magnetic material requires a costly manufacturing process. Signal processing algorithms to suppress such interference are available and can be improved, but they may not be sufficient to completely obviate the need for shielding.
- Point and distributed source reconstructions are both approximations of the true distribution of current in the body, but it is not known which is a better approximation. The two types of reconstructions are so different that it is difficult to directly compare their accuracy. An effective basis of comparison is needed.
- The incorporation of realistic body geometry and conductivity distributions used in the forward models is important. As models become more detailed, they also become more expensive to compute, and additional research in the designs of efficient algorithms are needed.
- How accurate must the forward model be to yield accurate reconstructions from data with some specified noise level; since improved accuracy costs more computer time, it is preferable to use a forward model that is no more accurate than necessary.
I. Medical Optical Imaging
- Development of mathematical tools for the mapping of parameters such as absorption and scattering coefficients that govern the propagation of light in human tissues.
- Investigations to determine the biophysical basis for diffusion of light in tissue.
- Investigations of excited fluorochromes that target specific tissues.
J. Image-guided Minimally Invasive Diagnostic and Therapeutic Interventional Procedure Planning
- Definition of tumor and other surgical target margins or boundaries utilizing various medical imaging techniques(correlating with spatially registered histology to estimate the capabilities of the various imaging modalities in defining the boundaries for various anatomic regions).
- Development of real-time image processing techniques, particularly rapid methods of model creation, three-dimensional rendering, and accurate segmentation of anatomic tissues for various imaging modalities.
- Research in surgical planning and simulation particularly trajectory planning for needle biopsy, its basic surgical application today.
- Improvement, via more complex automated technologies, of current registration or image fusion methods of different medical imaging modalities, especially video-based and laser-scanning techniques with prospectively created models.
Guidance and Localization
- Development of flexible and untethered sensors to provide anatomic fiducial marks or information on the position of needles, catheters, and surgical instruments, for tracking of instruments or for fusing patient and image coordinate systems.
- Development of computational systems and algorithms to enable "instantaneous" reconstruction, reformation, and display of the image data so as to enable real-time following of a physician's actions during a procedure (e.g., advancing a catheter or needle).
Monitoring and Control
- Definition of the temporal resolution required for various image-guided therapeutic procedures, taking into consideration the physical characteristics of the specific imaging modalities and the dynamic properties of the monitored procedures, specifically for multi-slice volumetric monitoring.
- For MRI, development of new pulse sequences specifically for therapy applications rather than diagnostic applications. A particularly important need is the development of highly temperature-sensitive pulse sequences to enable monitoring of "heat surgery."
- Investigations to correlate the factors affecting energy deposition or abstraction (e.g., pulse duration, pulse energy, and power spectrum) with histological and physiological changes in the tissue and resulting image changes, for the purpose of determining mechanisms of thermal damage and the biophysical changes that take place during various thermal surgical procedures such as interstitial laser therapy, cryoablation, and high-intensity focused ultrasound treatment. Such investigations need to be done for various anatomic regions and medical conditions for which such therapy might be appropriate.
- Investigation of the range of medical conditions amenable to treatment with minimally invasive techniques made possible by expanded capabilities for visualization during a procedure via the various medical imaging modalities.
Instruments and Systems
- Although prototypical MRI systems have been created that provide direct and easy access to the patient, more research and development is required to optimize further the geometric configuration of these systems. Similar requirements are appropriate for the other imaging modalities, particularly CT.
- For MRI-guided biopsy and therapy, magnet-compatible needles and other equipment, using materials that do not cause image distortions in a magnetic field, need to be identified and developed. Accessible and easy-to-use guidance systems are required to perform localization or biopsy of lesions detected by MRI alone.
- Development of high performance two-dimensional detector arrays for CT and other x-ray imaging modalities, and of less expensive two-dimensional transducer arrays for ultrasound, along with appropriate means for acquiring, reconstructing, and displaying the data.
- Improved methods of inexpensively shielding the magnetic field to enable inexpensive retrofitting of existing MRI systems into current operating rooms.
- Integration of imaging methods with therapeutic procedures, including feedback systems between data display devices and image information, computer-assisted image-controlled surgical tools, robotic arms, and instruments.
- Creation and development of new instruments and tools to accomplish new tasks enabled by the availability of image-guided therapy, especially specialized surgical tools such as in the case of MRI-guided therapy.
K. Knowledge-based Decision Support Systems for Diagnosis and Therapy
- Longitudinal analysis (change detection) of diagnostic images from temporally spaced exams is likely to be valuable; however, this is technically challenging. Tools and techniques for such longitudinal analysis are required. In particular, spatial registration of these images is confounded by variations in instrumentation, patient positioning, and acquisition parameter settings.
- Development of larger image databases with detailed knowledge of outcomes and truth to facilitate intercenter comparison of various CAD approaches.
- Techniques to correlate multiple views to locate features in pairs or larger sets of images.
- Application and testing of innovative computer algorithms, neural networks and other forms of machine intelligence for computer-aided diagnosis and image processing. For example, mathematical optimization could lead to image processing that is optimized for each breast image. Likewise, the large number of images and volume of data inherent in MRI can benefit from sophisticated techniques for image processing, display, analysis, and computer aided diagnosis. Optimization of image acquisition methodologies and image processing algorithms for exploration of quantitative biologic and physiologic MRI parameters to improve tissue characterization and maximize its impact on cancer diagnosis and management.
- Automated segmentation for 3D image display and processing (e.g. for follow up studies, tumor volume quantification in response to treatment, etc.).
- Image fusion with multiple imaging modalities has the potential to enhance diagnosis and staging. In 3D imaging, multimodality image registration may be important, when MR images may be combined with other 3D displays (e.g. Nuclear Medicine, PET, Ultrasound, etc.) to combine anatomic, physiologic and biologic tumor information in a single 3D image.
- Automated segmentation of images is required for 3D interactive display and segmentation cannot rely on signal intensity alone. Model-based algorithms need to be developed.
- Practical and correct rigid and elastic image registration algorithms need to be developed
- Significant improvements are required in VR technologies, e.g. head-mounted displays, rendering engines, and speed of head/body tracking (latency).
L. Image Processing for Medicine
- Develop software tools for the principal applications of image processing in medicine:
a) Image segmentation
b) Computational Anatomy
c) Registration of multi-modality images
d) Synthesis of parametric images
e) Data visualization
f) Treatment Planning
- Software and image processing methods directed at registering metabolic image sets with anatomic image sets, especially in the breast; taking into account the change in shape with positioning on different imaging devices.
- Extension of the traditional approaches to image segmentation and object classification to include shape information rather than merely image intensity. These techniques, when combined with the ability to accurately register deformable objects, would make a major contribution to interpretation of images from the heart, abdomen, and pelvis.
- Derivation of quantitative methods for analysis of tissue function and correlation of that information directly to anatomy. Application of statistical approaches for identification of subtle changes in time series of three-dimensional images obtained for mapping brain function, the use of prior anatomic information to constrain the reconstruction of low signal-to-noise metabolic data, and the derivation of parametric images that accurately describe the kinetics of biologically relevant anatomic images.
- Radiation therapy and tissue ablation by heating or freezing require precise definition of the anatomic targets and a physical characterization of the processes leading to destruction of abnormal tissues while preserving nearby normal tissues. The need for simulation of the treatment process in three dimensions is a major challenge for high-performance computing and algorithm development.
- Radiation treatment planning of internally distributed radiopharmaceuticals.
- Digital archiving systems are required that can accomplish the following functions:
- An object-oriented image database.
- Client server software.
- 1 to 2 TBytes per year on-line capacity.
- Compression ratio of 50:1.
- Image Database that employs the DICOM 3.0 standard.
M. Mathematics for Biomedical Imaging Science
- Investigation of the trade-offs of stability versus resolution for the inverse problem for the Helmholtz equation, as applied to ultrasound and microwave imaging
- Development of the mathematical theory and algorithms for inverse problems for the transport equation. There are applications not only to light tomography but also to scatter correction in CT and emission CT. Also needed are investigations of diffusion and diffusion wave approximations.
- Extension of anisotropic diffusion in image processing from 2D to 3D and higher dimensions, with inclusion of semi-local or global information. Application-specific tuned filters should be developed, for example in perceptual grouping. Coherence-enhancing anisotropic diffusion and new filter models using other structure descriptors than the (regularized) gradient or the structure tensor, based on wavelets, Gabor filters, or steerable filters are candidates.
- Exploitation of the relation between structures at different scales in the nonlinear context aids in the avoidance of correspondence problems and understanding the deep structure in nonlinear diffusion processes. Using a scale-space stack of these filters to extract semantically important information for a specified task.
- Exploration of novel, more time-critical applications solved by PDE?s using implicit schemes, splitting and multi-grid methods, or grid adaptation strategies.
- Research on guidelines for automatic parameter determination for specific tasks to simplify the application of nonlinear diffusion filters, in combination with other image processing techniques to support data compression, segmentation, tomographic reconstruction and neural networks for learning a priori information.
- Development of numerical methods for parabolic inverse problems for application to light tomography (the diffusion approximation). Methods using adjoint fields are promising.
- Investigation of Gelfand-Levitan theory for multi-dimensional hyperbolic inverse problems. The one-dimensional Gelfand-Levitan theory is the backbone for one-dimensional inverse scattering; extension to three dimensions would solve the mathematical and numerical problems in ultrasound and microwave imaging.
- Development of a general-purpose algorithm for bilinear inverse problems. The inverse problems of medical imaging frequently have a bilinear structure, irrespective of the type of underlying equation (elliptic, parabolic, hyperbolic, or transport). A general-purpose algorithm for discretized problems is conceivable and would advance the field substantially.
- Development of methods of scatter correction through transport models for transmission and emission CT. This promises to be easier to solve than the inverse problem in light tomography, at least for cases in which scatter is not too large.
- Creation of reconstruction algorithms for three-dimensional CT and efficient algorithms for cone-beam and helical scanning using areal detectors.
- Classification of three-dimensional scanning geometries according to the stability of the inversion problem.
- Development of reconstruction algorithms, possibly of Fourier type, for three-dimensional PET, especially by use of the stationary phase principle.
- Development of faster methods for computing maximum likelihood estimates with priors, more efficient iterative methods, and methods that exploit symmetries of the scanning geometries though efficient numerical algorithms such as the FFT.
- Investigation of precondition for nonlinear iterations such as expectation maximization (EM).
- Construction of good priors for SPECT and PET computations.
- Creation of mathematical attenuation corrections for emission CT, i.e., determination of the attenuation map from the emission data (without transmission measurements). Encouraging mathematical results for two source distributions are available, and simulations with templates have been performed.
- Creation of specialized regularization methods, particularly for regularization in time, which have application to magnetic and electrical source imaging.
- Reconstruction of functions under global shape information using templates. A typical application is the reconstruction of attenuation maps for emission CT and magnetic and electrical source imaging.
- Constraint reconstruction, i.e., reconstruction of a function from transmission or emission CT or MRI data with side conditions such as non-negativity. There are applications to limited angle CT and to MRI scans with insufficient sampling.
- Reconstruction of a function from irregular spaced samples of its Fourier transform, which has applications to CT and MRI.
- Removal of artifacts caused by opaque objects, such as hip joints in CT.
- Reconstruction of a function in R3 from integrals over almost-planar surfaces. There are applications to MRI data collected with imperfect magnets.
- Reconstruction of vector fields to determine more than the curl, especially applied to Doppler imaging.
- Use of wavelet techniques in image reconstruction.
- Theory and algorithms for tomotherapy, image fusion and calculation of dose distributions.
N. Evaluation and Measurement Systems
- Develop accepted in vitro models for evaluation and measurement.
- Define relevant outcome/effects standards for human applications of imaging systems.
- Inaugurate translational clinical trial mechanisms and support for biomedical imaging sciences and engineering.
- Develop and foster adoption of clear regulatory guidelines.
O. Training and Career Development
- Develop curricula focusing on "invention" and creativity, aimed at discerning and overcoming roadblocks.
- Train professionals in the area of medical physics, applied mathematics, biomedical imaging science and biomedical engineering to develop basic methods, and link their training with translational clinical research programs
- Develop multidisciplinary training sites; include corporate partners and a mix of both NIH and industrial support for the implementation of such programs.
Throughout the workshop and task force deliberations it was apparent that, for the development of future biomedical imaging methods and applications, active and concurrent participation of many relevant disciplines should work together to facilitate cross-disciplinary research and design through mission- and hypothesis-directed research programs. The workshop/task force participants recommended the following three specific actions:
Programmatic Changes. Encourage new multi-disciplinary research initiatives, educational infrastructure in the areas of biomedical imaging sciences, and creation of opportunities and funding for young scientists.
Central Resources. In order to help assure that different disciplines collaborate effectively, develop information technology that maximizes the efficiency with which ideas can be shared and expanded upon. In this regard, it is considered important that there be adequate support for multidisciplinary training programs focusing on "invention" and creativity.
Integration of Approach. Provide an integrated programmatic approach as well as a coordinated and cooperative pattern of shared funding among NIH institutes, NSF, private foundations, industry, and other sources.
Recent advances in the fields of the biomedical imaging sciences have attracted the attention and interest of investigators from many disciplines. The potential for the emergence of clinically applicable technologies from these fields, particularly in the realm of tissue-based therapies, is very significant. For this reason, a Workshop was held in order to determine the optimal directions that can be taken by the NCI to provide the proper support for research initiatives in these areas. Since the fields of biomedical imaging science, medical physics, visualization, diagnostic radiology and applied mathematics are so interdependent, it was deemed best to incorporate and integrate recommendations in all areas to obtain the best potential interdisciplinary synergy in their evaluation. This report of the recommendations summarizes the deliberations of those present at the workshop, task force meetings, and follow-up exchange of documents and discussions among all of the participants.
The task force participants underwent a facilitated idea structuring process known as the Hoshin facilitation process to address general issues that pertain to biomedical imaging sciences and especially applications in cancer. This process, often used in the corporate world as part of Total Quality Management programs was chosen because it fosters group openness and trust, thereby maximizing idea sharing. The mechanics of the Hoshin process and the method by which its results are interpreted are described in the Appendix.
In addition, background documents were exchanged to stress the importance of corporate technology development, the roles of regulation and liability in the fate of medical technologies and the growing importance of interdisciplinary work in the development of new medical technologies.
NCI Task Force on "Matching Clinical and Biological Needs with Emerging Technologies"
NCI Imaging Sciences Working Group
Michael W. Vannier, M.D.
Department of Radiology
University of Iowa
Iowa City, IA
Laurence P. Clarke, Ph.D.
Department of Radiology
University of South Florida
David Gustafson, Ph.D.
Siemens Ultrasound, Inc.
David J. Hawkes, Ph.D.
Computational Imaging Science Group
UMDS, Guys Hospital Campus
London, United Kingdom
William E. Higgins, Ph.D.
Department of Electrical Engineering
Pennsylvania State University
University Park, PA 16802-2705
Prof. Willi Kalender, Ph.D.
Institute of Medical Physics
University of Erlangen-Nuremberg
Valen E. Johnson, Ph.D.
Institute of Statistics and Decision Sciences
Nico Karssemeijer, Ph.D.
Department of Radiology
University Hospital Nijmegen
Nijmegen, The Netherlands
Chuck Meyer, Ph.D.
Department of Radiology
University of Michigan
Ann Arbor, MI
Kurt R. Smith, D.Sc.
Surgical Navigation Technologies, Inc.
Donald O. Thompson, Ph.D.
Director Emeritus, Center for Nondestructive Evaluation
Aerospace Engineering and Engineering Mechanics
Iowa State University
Prof. Max A Viergever
Image Science Institute
University Hospital Utrecht
Utrecht, The Netherlands
National Cancer Institute:
David Bragg, MD
Daniel Sullivan, MD
Anne Menkens, PhD