PathMiner: A Web Based Tool for Computer
Assisted Diagnostics in Pathology.
Lin Yang(1,3), Oncel Tuzel(2), Wenjin Chen(3), Peter Meer(1),
Gratian Salaru(4), Lauri A. Goodell(4), Adam Bagg(5)
and David J. Foran(3)
(1)Department of Electrical and Computer Engineering
(2)Department of Computer Science
Rutgers University, Piscataway, NJ 08854, USA
(3)BioImaging Laboratory, Department of Pathology and Laboratory Medicine
UMDNJ-Robert Wood Johnson Medical School
Piscataway, NJ 08855, USA
(4) Cancer Institue of New Jersey, New Brunswick, NJ, 08903, USA
(5) Department of Pathology and Laboratory Medicine
University of Pennsylvania, Philadelphia, PA, 19104, USA
Large-scale, multi-site collaboration has become indispensable for a
wide range of research and clinical
activities which rely on the capacity of individuals to dynamically
acquire, share and assess images and correlated
data. In this paper, we introduce a web-based system, Pathminer, for
interactive telemedicine, intelligent archiving
and automated decision support in pathology. The PathMiner system
supports network-based submission of queries
and can automatically locate and retrieve digitized pathology
specimens. It correlated molecular studies of cases from
"ground-truth" databases which exhibit spectral and spatial profiles
consistent with the query image. The statistically
most probable diagnosis or structural classification is provided to
the individual who is seeking decision support.
To test the system under real-case scenarios a network-based
laboratory has been established at strategic sites at
UMDNJ - Robert Wood Johnson Medical School, Robert Wood Johnson
University Hospital, the University of
Pennsylvania School of Medicine, Hospital of the University of
Pennsylvania, The Cancer Institute of New Jersey,
and Rutgers University. The average five class classification accuracy
of the system is 93.18% based on ten fold
cross validation on a dataset containing 3691 images. We also show
prospective performance of PathMiner in real
application where the images exhibited large variances in staining
characters compared with the training data. The
average five class classification accuracy in this open set
experiments is 87.22%.
IEEE Transactions on Information Technology in BioMedicine,
13 , 291-299, 2009.
