FDA Home Page | CDRH Home Page | Search | CDRH A-Z Index | Contact CDRH

Program Area: Medical Imaging and Diagnostics

Scope: A wide variety of new digital imaging and display devices is under development by academia and industry, with a broad range of performance characteristics. The Center requires augmented support for the evaluation of such devices. To this end, OSEL scientists are developing evaluation methodologies for diagnostic medical imaging systems such as mammography and fluoroscopy, computed tomography, nuclear medicine, diagnostic ultrasound, and magnetic resonance imaging, as well as novel soft-copy display devices for viewing medical images. This program is located within the Division of Imaging and Applied Mathematics (DIAM).

Background: The Medical Imaging Program at CDRH was initiated in the early 1970s by its predecessor, the Bureau of Radiological Health (BRH). The goal was to go beyond the traditional BRH laboratory approach of simply measuring the level of radiation emitted by an electronic or diagnostic modality, to measurement of the level of imaging performance as well. Laboratory measurement methods were developed for assessing the performance of contemporary and new technologies in the fields of radiography, mammography, computed tomography, diagnostic ultrasound, radioisotope imaging, magnetic resonance imaging, with current emphasis on digital detectors and displays. The program led to contributions to consensus measurement methodology and international standards that are used here today in the approval process for new technologies, in particular, digital radiography and mammography, and diagnostic ultrasound. In-house research and collaboration with academic investigators have also led to laboratory and clinical systems that optimize the ratio of imaging performance to radiation exposure in mammography.

In the late 1980s it was realized that many of the multivariate statistical methods developed for image evaluation were applicable to the assessment of conventional and neural-network systems for computer-aided diagnosis (CAD) in medicine. These include the fundamental paradigm of the receiver operating characteristic or ROC plot of true-positive fraction (or sensitivity) versus the false-positive fraction (or one minus the specificity). The ROC paradigm provides the unifying framework for the evaluation of all diagnostic devices. Starting in the mid-1990s and up to the present, OST’s imaging group has made fundamental contributions to the multivariate ROC statistical approach to assessment of imaging and CAD systems. A multivariate approach is needed for several reasons, including the great reader variability that has been demonstrated in recent years. Current work includes the development of software for clinical trial design and analysis. Contemporary in-house programs address measurement methodology for the assessment of many new technologies for digital image capture and display; novel ultrasonic methods for in-vivo tissue characterization and bone densitometry; statistical analysis of neural-network decision-assist tools; and the emerging field of DNA micro-arrays.

Program Description: DIAM scientists are engaged in developing appropriate methods for evaluating medical imaging system performance and dose. Investigations take the form of theoretical analysis, numerical simulation of the entire imaging chain, and experimental validation. In some instances, improved/optimized system designs are validated through actual system construction and clinical evaluation. Measurement and analysis procedures are also being developed to evaluate the performance of new soft-copy display devices that can have dramatically different light-emitting structures and associated performance characteristics whose impact on the image interpretation process is currently unknown. OSEL scientists provide reliable, quantitative laboratory measurements of imaging system characteristics to the imaging research community. OSEL scientists are also elucidating the fundamental mechanisms underlying the interaction between the image-forming radiation and the anatomy being imaged.

These investigations inform the Center's regulatory decision-making on new digital image devices. The expertise developed through this program is being applied to the review of PMAs for ultrasound bone sonometers and new digital radiographic imaging systems. This program contributes to the development of premarket guidance documents including "Information for Manufacturers Seeking Marketing Clearance of Digital Mammography System" and "Bone Sonometer PMA Applications: Final Guidance for Industry and FDA." OSEL scientists are applying their expertise to the development of a CDRH web site on CT, the development of amendments to the diagnostic x-ray equipment performance standard, and the development of an advisory pertaining to pediatric CT exposures. Last year OSEL staff, along with OCER staff, assisted in the joint planning of a consensus development conference on CT with NIH, and made presentations at that conference.

Improved knowledge of the fundamental imaging mechanisms will lead to an understanding of the sources of variability in imaging data. Having that, inter-machine and inter-institute measurements can be corrected, leading to absolute, quantitative measures which can then be codified through a measurement standardization process. The x-ray spectral measurements program provides a source of otherwise unavailable data to the entire mammography research community for use in developing new equipment performance standards as digital mammography develops, special procedures and test equipment for MQSA, and will be used to inform decisions on marketing clearance for new products and in compliance actions.

The clinical assessment of medical imaging systems is complicated by the great variability observed in readers in radiology. This variability leads to the necessity of a multivariate approach that includes the range of patient case difficulty, the range of reader skills, and correlations among the patients, readers, and imaging technologies under comparison. Thus a primary goal of this program is to develop statistical methods for analyzing the performance of imaging systems within the context of reader variability. At the same time, new and increasingly sophisticated computer techniques for medical diagnosis are being developed by academia and industry to aid/augment the human reader in the interpretation of high-dimensional image data sets. A second major goal of this program is to develop study designs, objective measurements, and analytical methods for the laboratory and clinical assessment of imaging and other diagnostic systems, systems for computer-aided diagnosis (CAD) used in medical imaging, and stand-alone image-based computerized diagnostic modalities such as high-dimensional DNA micro-arrays (DNA chips).

The methodological tools under development for analyzing the performance of imaging systems within the context of reader variability are referred to broadly as the multiple-reader, multiple- case (MRMC) receiver operating characteristic (ROC) paradigm. The paradigm is a multivariate analysis of the map of reader true-positive rates versus false-positive rates as a function of the variables listed above. A key question we are investigating is that of analyzing not only reader or system average performance but also the multivariate uncertainties that result from the finite sample of patients and readers. The approach to the assessment of systems for computer-aided diagnosis and high-dimensional DNA arrays is an extension and application of the multivariate approach to ROC analysis. In the case of CAD and DNA micro-arrays, the key question is that of analyzing the multivariate uncertainties that result from the finite sample of patients used to train the system, the finite sample of patients used to test the system, and their interaction. The general subject of uncertainty analysis also addresses the classical problem of the "generalizability" of performance of a CAD or micro-array algorithm. We make use of advanced statistical tools for diagnostic decision making under uncertainty, including classical Bayes' discriminants, neural-network architectures, and fuzzy logic, in our studies of CAD algorithms and their performance.

In the last few years CDRH has been receiving an increasing number of premarket submissions for digital imaging modalities and modalities used for CAD, not only in imaging but also for clinical laboratory diagnostic tests. Statistical and analytical methods developed in the OSEL imaging group have been directly used to assist with both the design and the data analysis for several of these submissions, both in imaging and CAD. OSEL has played a significant role in the statistical evaluation of device submissions such as those for automated Pap smear readers, lung cancer, and breast cancer detection devices. A current emphasis is on the development of a draft guidance document, in collaboration with scientists in OSB and ODE, to provide industry and academia with "best" and "acceptable" practices for the laboratory and clinical assessment of diagnostic devices. We are also pursuing the potential for coordinating MRMC ROC clinical study designs in such a way as to optimize the expenditure of resources over the total product life cycle of an imaging technology—from university research, through pilot clinical trials and pivotal FDA studies, to confirmatory ACRIN (American College of Radiology Imaging Network) trials, through downstream cost/benefit studies of interest to public policy makers and insurers working at that higher level.

Relevance to FDA /CDRH Mission and the Public Health Impact: The expertise developed through this program is being applied to the review of PMAs for ultrasound bone sonometers and new digital radiographic imaging systems, the development of amendments to the diagnostic x-ray performance standard, the development of an advisory pertaining to pediatric CT exposures, and last year, to participation in the planning and presentation of a consensus development conference on CT with NIH. The x-ray spectral measurements program provides a source of otherwise unavailable data to the entire mammography research community. Finally, investigation of computer-assisted diagnosis devices provide the Center with the scientific basis to effectively regulate this fast growing field.

Five-Year Objectives: Future efforts in the Medical Imaging and Diagnostics program will focus on contemporary and emerging issues of regulatory interest. In the field of digital imaging, these issues include characterizing imaging performance without the underlying assumptions present for analog imaging systems, validating and refining consensus measures, optimizing overall system performance, and assessing tissue parameters from the digital data. Special emphasis with respect to imaging performance and patient exposure will be put on emerging techniques for volume imaging of the human anatomy including cone beam computed tomography using flat panel detector arrays. In the field of ultrasound, topics of interest include tissue characterization, bone densitometry, contrast agents, ultrasonic imaging system performance characterization, temperature mapping, elastography, and the use of ultrasonic measurements in pattern recognition systems. While DIAM continues to work in the area of ultrasound, the largest activity for this year involves high-intensity focused ultrasound, and is described in the program description for Fluid Dynamics and Ultrasonics .

The Medical Imaging and Diagnostics program has made fundamental contributions to the field of statistical analysis of diagnostic imaging and systems for computer-aided diagnosis. We would like to exploit this work and validate its range of utility through extensive computer simulations. In the process, we would be seeking the most efficient or statistically powerful approaches to the evaluation of medical imaging and computer-assist decision modalities. An ultimate goal is the development of a multiple-reader (e.g., multiple radiologists, multiple pathologists) multiple-case (MRMC) version of our current software for ROC analysis in the absence of ground truth (i.e., without a gold standard). Development of such a system would address one of the most difficult yet most common assessment problems in the field of diagnostic medicine. Many statistical problems in the new field of bioinformatics remain unsolved. We would like to extend our contemporary successful research into new realms of bioinformatics that are opening up due to the accumulation of data from multiple testing and patient demographics. This will require continual upgrading of our computational facilities.

Project Descriptions:

Image Acquisition: A wide variety of new digital imaging devices, with a broad range of performance characteristics, is under development by academia and industry. The Center requires new and improved guidance for the evaluation of such devices. To this end, OSEL scientists are developing evaluation methodologies for diagnostic medical imaging systems such as mammography and fluoroscopy, computed tomography, nuclear medicine, diagnostic ultrasound, and magnetic resonance imaging. There are several research areas of interest within this broad field: simulation of the imaging chain, development of consensus measurement methods for imaging performance, and tracking of efficiency data for current systems. Many of these areas involve experimental laboratory investigations using techniques such as x-ray spectroscopy and computer evaluation of images, from digital modalities or of digitized x-ray film images, of appropriate test objects.

Image Display: Displays are currently considered to be the weakest link of the imaging chain for many applications (e.g., mammography). The purpose of this project is to provide the Center with the scientific expertise to address emerging regulatory issues in the area of diagnostic image display technology by developing measurement and analysis procedures to evaluate the performance of image display devices for digital diagnostic imaging systems. Approaches include the development of laboratory measurement methodologies, theoretical and numerical studies, and participation in standards development. Initial efforts have been directed toward the study of CRT-based video monitors driven by computer-controlled digital frame buffers. Investigations of the performance characteristics of emerging flat-panel display technologies using advanced liquid crystal modes are also underway. In addition, a small effort is being expended to explore the effect of color on diagnostic performance.

Computer-aided Diagnosis: Recent studies indicate that helical CT may be an effective screening tool for lung cancer while full-field digital mammography has advanced into a clinical tool. The primary goal of this project is to develop additional in-house CAD expertise and draft FDA guidance for both the submission and review of CAD applications.

Multivariate Statistical Assessment: The purpose of this project is to develop study designs, objective measurements, and analytical methods for the laboratory and clinical assessment of medical imaging systems, systems for computer-aided diagnosis used in medical imaging, and stand-alone image-based computerized diagnostic modalities such as high dimensional DNA micro-arrays (DNA chips). The approach to the laboratory assessment of systems involves development of models, laboratory measurements, and consensus development. The approach to the clinical assessment of systems is to develop statistical methods for analyzing the performance of imaging systems within the context of the great variability observed in readers in radiology. The approach to the assessment of systems for computer-aided diagnosis and high-dimensional DNA arrays is an extension and application of the multivariate approach to ROC analysis.

Updated January 27, 2005

CDRH Home Page | CDRH A-Z Index | Contact CDRH | Accessibility | Disclaimer
FDA Home Page | Search FDA Site | FDA A-Z Index | Contact FDA | HHS Home Page

Center for Devices and Radiological Health / CDRH