RESEARCH INTEREST
Dr. Craig Abbey investigates image analysis as applied to diagnostic and quantitative tasks using medical or biomedical images. These studies are used to evaluate imaging technologies and to optimize scanner design. He is currently funded to evaluate in vivo quantitative estimates of progression and proliferation in a murine model of breast cancer using a dedicated positron emission tomography (PET) scanner. The goal of the study is to quantify the accuracy and precision of in vivo measurements compared to traditional histological endpoints that preclude further biologic and behavioral analysis, and to quantify the advantage of longitudinal studies for evaluating therapeutic effect in preclinical trials.
Another component of his research is diagnostic image quality: how imaging systems and image processing produce an impact on diagnostic performance. The motivation for this work is the recognition that improvements to imaging systems-whether they are improvements to the imaging devices or to the processing of image data-ought to result in more accurate diagnosis in the clinic (or else equally good diagnosis at reduced cost or radiation dosage). This kind of research requires knowledge of how images are acquired and subsequently processed, as well as an understanding of how observers extract diagnostic information from images. Professor Abbey's research has primarily focused on this latter component of the diagnostic process. The majority of his work has been on building models that predict the diagnostic performance of human observers and using these models to optimize image-processing methods. For diagnostic tasks, the performance of human observers (radiologists or other clinicians) is typically evaluated through psychophysical studies in which each observer must read and diagnose a set of test images for which the true state of the patient has been obtained. These studies are generally costly and time consuming, and therefore impractical for optimizing the imaging process. A model that could predict the outcome of the psychophysical studies of diagnostic tasks would allow the exploration of a much broader range of free parameters in the imaging.
Along with his collaborators, Professor Abbey developed and applied models of diagnostic performance to a number of different imaging problems, including image reconstruction in nuclear medicine and multi-frame averaging in coronary angiography. More recently, he has been investigating new ways to understand human observers by directly measuring how an observer weights an image while performing a diagnostic task. The resulting spatial maps of observer weights have great potential for building more accurate models of the human observer.
PUBLICATIONS
C. K. Abbey and M.P. Eckstein, "Optimal estimates of human-observer templates in two-alternative forced-choice experiments," IEEE Transactions on Medical Imaging 21(5):429-440. 2002.
C. K. Abbey and H. H. Barrett, 2001, "Human and model-observer performance in ramp-spectrum noise: Effects of regularization and object variability." Journal of the Optical Society of America A. 18(3): 473-488.
MAJOR RESEARCH INTERESTS
Diagnostic and quantitative image analysis and system optimization, modeling observer performance, image reconstruction of emission data (PET and SPECT), molecular imaging of breast cancer proliferation and progression in murine systems.
|
