Medical Imaging

Radiologists must make life-critical diagnostic decisions from noisy medical images that have complex visual structure and statistics.
Radiologists must make life-critical diagnostic decisions from noisy medical images that have complex visual structure and statistics.

There are times when an experienced physician sees a visible lesion clearly and times when he does not. This is the baffling problem, apparently partly visual and partly psychologic. They constitute the still unexplained human equation in diagnostic procedures.

Henry Garland, M.D., 1959

 

Our research aims at understanding the processes by which radiologists and doctors detect and classify disease from images, and how they use this acquired knowledge to improve the acquisition, processing, and display of medical images to improve the sensitivity and specificity of their diagnostic decisions. The projects range from basic scientific studies, where we try to understand the mechanisms, strategies, and algorithms used by radiologists and doctors to detect and/or classify disease, how AI influences radiological decisions and search, to the development of AI-based metrics of medical image quality and computer algorithms for applying the knowledge we gain of how humans process medical images to assist radiologists and doctors in real-world medical scenarios. 

 

Selected Recent Publications: 

Jonnalagadda, A., Barufaldi, B.B., Maidment, A.D., Weinstein, S.P., Abbey, C.K. and Eckstein, M.P., 2025. Convolutional neural network model observers discount signal-like anatomical structures during search in virtual digital breast tomosynthesis phantoms. Journal of Medical Imaging, 12(5), pp.051809-051809.

Gommers, J.J., Verboom, S.D., Duvivier, K.M., van Rooden, C.J., van Raamt, A.F., Houwers, J.B., Naafs, D.B., Duijm, L.E., Eckstein, M.P., Abbey, C.K. and Broeders, M.J., 2025. Influence of AI decision support on radiologists’ performance and visual search in screening mammography. Radiology, 316(1), p.e243688.

Klein, D.S., Karmakar, S., Jonnalagadda, A., Abbey, C.K. and Eckstein, M.P., 2024. Greater benefits of deep learning-based computer-aided detection systems for finding small signals in 3D volumetric medical images. Journal of Medical Imaging, 11(4), pp.045501-045501.

Lago, M.A., Jonnalagadda, A., Abbey, C.K., Barufaldi, B.B., Bakic, P.R., Maidment, A.D., Leung, W.K., Weinstein, S.P., Englander, B.S. and Eckstein, M.P., 2021. Under-exploration of three-dimensional images leads to search errors for small salient targets. Current Biology, 31(5), pp.1099-1106.

 

Selected "classic" Publications:

Eckstein, M.P., Bartroff, J.L., Abbey, C.K., Whiting, J.S. and Bochud, F.O., 2003. Automated computer evaluation and optimization of image compression of x-ray coronary angiograms for signal known exactly detection tasks. Optics Express, 11(5), pp.460-475.

Eckstein, M.P., Ahumada Jr, A.J. and Watson, A.B., 1997. Visual signal detection in structured backgrounds. II. Effects of contrast gain control, background variations, and white noise. Journal of the Optical Society of America A, 14(9), pp.2406-2419.