With the ever-increasing role of medical images in diagnosis, treatment, and evaluation of treatment effects, extraction of quantitative information from these images and efficient use of the results have become a necessity. In recent years, others and we have developed novel two-dimensional (2D) and three-dimensional (3D) deformable models for a variety of image analysis applications in medicine and industry. We have also developed reliable automated methods for defining the initial shape of the model for segmentation and characterization of hippocampus from magnetic resonance imaging (MRI). These methods need to be extended and feature extraction methods developed to segment and characterize (i.e., determine multi-parametric intensity distribution, texture, shape, surface area, and volume of) brain structures such as hippocampus, amygdala, red nucleus, substantia nigra, globus pallidus, putamen, corpus callosum, and thalamus from MRI. In addition, new databases need to be developed to hold the results with other clinical information (e.g., textual data) in a manner that can be searched, retrieved, and queried conveniently from any computer station. The goal of this project is to develop novel approaches for the above needs. Developments will be done in the context of an important biomedical application and will localize, segment, and characterize hippocampus from MRI. The proposed database will be able to evaluate correlation between a variety of risk factors and post-operative outcomes. The methods will be tested; evaluated, and validated, using simulated images and clinical studies of epileptic patients. Clinical diagnosis based on EEG studies and surgery outcome will be used as """"""""gold standards"""""""" for evaluation and validation of the image analysis methods. The proposed research will be a breakthrough in the development and application of computerized methods for medical image quantitation and object characterization, and will advance image analysis science in the direction of integrating knowledge-based systems, deformable models, texture analysis, and database technology. The proposed approach is applicable to the identification, segmentation, and characterization of other biological structures (e.g., lung, liver, kidneys, cells, neurons). It is also applicable to virtually any image analysis task for which object quantitation and characterization are used.
Showing the most recent 10 out of 20 publications
