 |
Anant Madabhushi
Associate Professor
Director
Lab for Computational Imaging & Bioinformatics - LCIB
Rutgers
University
Dept. of Biomedical Engineering
599 Taylor Road
Piscataway. N. J. 08854
(732) 445-4500, Ext 6213
FAX - 3753
anantm@rci.rutgers.edu |
Computer aided diagnosis, prostate cancer, breat cancer, image analysis, machine learning, prognosis, digital pathology, MRI
The complexity of biological systems and the vast amount of information now available at the level of genes, proteins, metabolites, tissues, and organs, requires developing quantitative computational methods to define relationships between structure and function at multiple biological scales. For instance, while Magnetic resonance imaging (MRI) and MR spectroscopy can probe a variety of physiological (e.g. blood vessel permeability) and metabolic characteristics of cancer, little is known about the changes in gene expression that underlie the spectral and imaging features observed in cancer. The ability to map genomic and proteomic expressions to /in-vivo/ imaging opens the potential for new insights into disease characterization and a better understanding of disease mechanisms.
As Director of the Laboratory for Computational Imaging and Bioinformatics at Rutgers University, Dr. Madabhushi's research goals have centered on developing novel computerized image and spectral analysis, and multi-modal registration tools to facilitate synergistic and correlative analysis of disease signatures across multiple scales and functionalities -- from gene and protein expression to spectroscopy to histopathology and to MRI, for early computerized diagnosis, prognosis, and theragnosis of prostate, breast, and ovarian cancer/.
We seek to develop a quantifiable framework to link and integrate the cross-modality phenotype mappings for cancer across spatial scales, which will provide the ability to map profiles of proteins and enzymes to their corresponding imaging and/or histological image parameters on a pixel-by-pixel basis. The framework will facilitate identification of specific subsets of cross-modal features that relate to the emergence of a specific phenotype. We have also been developing machine learning schemes for combining measurements at multiple scales and modalities into an integrated prediction score that can provide a superior outcome prediction compared to a single predictor.
|
