Quantitative Imaging and Computer-Aided Diagnosis
Contact
Summary
This research program (i) develops a framework for quantifying uncertainty in the use of quantitative imaging and methodologies to account for this uncertainty; and (ii) investigates alternate and less burdensome methods for the assessment of devices, utilizing imaging biomarkers and image-extracted features with a special focus on computer-aided diagnosis devices.
Our research resources include an anthropomorphic thoracic phantom with vasculature insert, and synthetic lung nodules with different shapes sizes, and radiodensities.
The output from a CAD system is displayed as an arrow pointing to a lung nodule on a thoracic CT scan.
Medical images contain considerably more information than what clinicians currently use as part of their routine evaluation. To enable clinicians’ use of this information for improved patient care, it is necessary to devise image acquisition/analysis techniques that facilitate the collection and extraction of quantitative information, and pattern recognition/statistical learning techniques that intelligently aggregate extracted information. Examples of application include imaging biomarkers to quantify response to therapy and prognostics as well as computer-aided detection/diagnosis systems used in areas such as screening radiology, malignant/benign tissue discrimination in medical images, and digital pathology.
Our research program impacts the regulatory assessment of a wide variety of systems that either extract quantitative imaging biomarkers or use extracted image features in computer-aided diagnostic devices. Findings from the quantitative imaging component of our project will be helpful for recommendations for appropriate study designs and endpoints for the Center for Devices and Radiological Health (CDRH) and the Center for Drug Evaluation Research (CDER) pre-market submissions, guidance to developers, and to facilitate the qualification of imaging biomarkers. Findings from the computer-aided diagnosis component of our research will be helpful to evaluate device claims that utilize quantitative imaging features, to identify the extent to which synthetic/blended image data may provide benefit in this context, and to reduce the data size/number of clinical trials for these devices.
In addition, we have developed a Public dataset that consists of phantom CT images of an anthropomorphic thoracic phantom, and containing a vasculature insert on which synthetic nodules were inserted or attached. More than 2,500 3D datasets have been acquired or reconstructed with different dose, slice collimation slice overlap, pitch, reconstructed slice thickness, and reconstruction kernels.
Current funding sources
FDA Office of Women’s Health Project
Quantitative Imaging Biomarker Alliance Project
FDA Critical Path Initiative
External collaborators
Quantitative Imaging Biomarker Alliance (QIBA)
Carl E. Ravin Advanced Imaging Laboratories,Duke University
Computational Image Analysis Lab, Columbia University
Imaging Biomarkers and Computer-Aided Diagnosis Laboratory, NIH Clinical Center
Computer-Aided Diagnosis Research Laboratory, University of Michigan Medical Center
Personnel
FDA Staff:
Berkman Sahiner, Ph.D.
Weijie Chen, Ph.D.
Marios Gavrielides, Ph.D.
Kyle Myers, Ph.D
Nicholas Petrick, Ph.D.
Rongping Zeng, Ph.D.
Qin Li, Ph.D.
Aria Pezeshk, Ph.D
Research Fellows:
Ben Berman, Ph.D.
Tomoe Hagio, Ph.D.
Jiheng Wang, Ph.D.
Resource facilities
Anthropomorphic thoracic phantom with vasculature insert
Synthetic lung nodules with different shapes (lobulated, elliptical,spiculated, spherical) sizes, and radiodensities
Coronary vessel phantom that includes known locations, volumes and densities of vascular calcification
Relevant standards & guidances
Guidance for Industry and Food and Drug Administration Staff: Computer-Assisted Detection Devices Applied to Radiology Images and Radiology Device Data – Premarket Notification [510(k)] Submissions
Guidance for Industry and Food and Drug Administration Staff: Clinical Performance Assessment: Considerations for Computer-Assisted Detection Devices Applied to Radiology Images and Radiology Device Data - Premarket Approval (PMA) and Premarket Notification [510(k)] Submissions
Selected peer-review publications
- Gavrielides et al., Observer Variability in the Interpretation of HER2/neu Immunohistochemical Expression With Unaided and Computer-Aided Digital Microscopy
- Zeng et al., Approximations of noise covariance in multi-slice helical CT scans: impact on lung nodule size estimation
- Sahiner et al., Computer-aided detection of clustered microcalcifications in digital breast tomosynthesis: A 3D approach
- Gavrielides et al., Minimum Detectable Change in Lung Nodule Volume in a Phantom CT Study
- Gavrielides et al., Benefit of Overlapping Reconstruction for Improving the Quantitative Assessment of CT Lung Nodule Volume
- Huo et al., Quality assurance and training procedures for computer-aided detection and diagnosis systems in clinical use
- Petrick et al., Evaluation of computer-aided detection and diagnosis systems
- He et al., Computerized characterization of lung nodule subtlety using thoracic CT images
- Petrick et al., Comparison of 1D, 2D, and 3D Nodule Sizing Methods by Radiologists for Spherical and Complex Nodules on Thoracic CT Phantom Images
- Li et al., Volume estimation of low-contrast lesions with CT: a comparison of performances from a phantom study, simulations and theoretical analysis