The role of modeling methods in medical diagnosis

Abstract

Modeling methods in medical diagnosis are concerned with medical information processing as it pertains to utilizing biological modeling methods to facilitate patient care. Major considerations in this particular area are (1) the classification problem related to the establishment of disease entities-the taxonomy problem, and (2) the diagnosis of diseases. Available are properties, criteria, signs, symptoms, and manifestations of diseases that have been cumulated and categorized by clinicians and researchers. The problem is to optimally utilize the information content of a sign or set of signs in the practice of patient care as pertaining to the medical diagnosis problem. Some mathematical approaches implemented to facilitate such analyses include cluster analysis, discriminant analysis, Bayesian methods, computer approaches, game theory, information theory, stochastic representations, stepwise procedures, decision analysis, and pattern recognition techniques. Each of these has been studied in depth by numerous researchers advocating computer applications in medicine. Here we discuss the scope and limitations of utilizing modeling methods as a viable approach to interpreting vast amounts of biological data collected on a single patient during an encounter. We consider the following: (1) limitations associated with modeling methodologies; (2) levels of responsibilities, ranging over logging, summarizing, reporting, monitoring, and therapy selection; and (3) operational strategies and considerations as they affect hardware logistics, the actual algorithm utilized, and implementation of these sophisticated analysis systems.

Mathematical modelling of sleep fragmentation diagnosis

“…So, it follows from these results that the proposed mathematical model of sleep fragmentation diagnosis of each clinician we called Mathematic Diagnosis' Index (MDI) based on the three main sleep characteristics (MAR, SSS, ISA) each having its own threshold and weight values for each clinician is a posteriori. According to Lezotte and Scheinok [7] the determination of these thresholds and weights values (input parameters) could have been a severe obstacle in any modelling attempt as evidenced by the differences observed between the theoretical values given by each clinician (see Tab. 2 & 3) and the practical values deduced from statistical computations (see Tab.…”

“…10 that our mathematical model is a posteriori validated. Let's notice that if many mathematical models in medical diagnosis have been developed for a long time (see for example [4][5][6][7][8][9][10][11][12][13]), it doesn't seem, to our knowledge, that there exists any mathematical model of sleep fragmentation diagnosis.…”

“…In their seminal works, Lusted and Ledley [4,5,6] proposed many models from symbolic logic, probability, and value theory as a mathematical basis for logical analysis and in the use of machine aids to diagnosis. In the beginning of the eighties, Lezotte and Scheinok [7] discussed "The Role of Modelling Methods in Medical Diagnosis" using mathematical approaches which include cluster analysis, discriminant analysis, Bayesian methods, computer approaches, game theory, information theory, stochastic representations, stepwise procedures, decision analysis, and pattern recognition techniques. They pointed out some limitations of modelling methods in health care due to the complexity of the proposed model and also due to the sensitivity of the methodology to extract the informational content of the input parameters.…”