Invited Tutorial II
Presenter: Dr. Peter Hammer, Rutgers University
Time: Saturday, February 4, 2006 (5:45 - 6:45 pm)
Title: Logical Analysis of Data and Applications to Medical Diagnosis and Prognosis
Abstract:
The Logical Analysis of Data (LAD), a combinatorics, optimization, and Boolean algebra-based methodology for extracting information from data, was first proposed in 1986, with a first paper on the topic published in 1988 (Annals of Operations Research 16, 1988), and a first report on its implementation, along with some applications, published in 2000 (IEEE Transactions on Knowledge and Data Engineering 12, 2000). This methodology has been applied since then for classification (including diagnosis and prognosis), for the analysis of the importance and the role of variables, for discovering new classes, for development of decision support systems, etc. One of the most important areas in which LAD has been recently applied is biomedical informatics.
The survey will present an outline of the basic concepts, techniques and algorithmic issues (e.g. support set selection, discretization, pattern enumeration, model formation, construction of discriminants) of LAD. We shall also present results of recent applications of LAD to some problems in biomedical informatics, including ovarian cancer diagnosis using proteomic datasets (Proteomics, March 2004), risk stratification among cardiac patients (Circulation 106, 2002), and the distinction between various types of diffuse large B-cell lymphomas (Artificial Intelligence in Medicine 34, 2005). In addition, we will briefly sketch some yet unpublished results concerning biomaterial design optimization and the analysis of genomic data for the prediction of metastases development in breast cancer.