Projects
The proposed projects are:
1. Kolmogorov Complexity and Image Similarity
2. Content-Based Image Retrieval
4. Radlex Ontology and Natural Language Processing
5. Computer-Aided Detection of Flat Lesions in CT Colonography
6. Probabilistic Based Semantic Retrieval of Lung Nodules
Project 1 (Furst): “Kolmogorov Complexity and Image Similarity”. We are investigating the application of Kolmogorov Complexity and Normalized Information Distance (NID) as universal similarity measure. Specifically, we are investigating the impact of image linearization, spatial shifts, intensity shifts, noise, and feature images on the Normalized Compression Distance between images.
Project 2 (Raicu): “Content-Based Image Retrieval”. Our objective is to create a content-based image retrieval (CBIR) system for lung nodules. Our objective is to bridge the semantic gap between radiologists’ ratings and image features. We have been developing a conceptual-based similarity modelderived from content-based similarity to improve CBIR.
Project 3 (Raicu): “Lung Nodule Segmentation”. Our project has two main goals. Our first focus is to measure the variability among radiologist outlines in lung nodules in CT scan data. Our second goal is to create a soft segmentation of lung nodules in CT scans using a probabilistic classification that incorporates the variability metric.
Project 4 (Channin): “Radlex Ontology and Natural Language Processing”. The goal is to search through medical articles and identify new radiology terms using natural language processing and the LexEVS vocabulary service. In particular, we are looking for imaging observations and characteristics to add to Radlex, a specialized radiological ontology.
Project 5 (Suzuki): “Computer-Aided Detection of Flat Lesions in CT Colonography”. We are working on a scheme for computer aided detection of non-polypoid colorectal neoplasms from CT colonography, with an emphasis on Type 0-II A lesions. Our main goal is to improve the sensitivity and specificity achieved by previous algorithms.
