The medical community is presently in a state of transition from a situation dominated by the paper medical record to a future situation where all patient data will be available online by an electronic clinical information system. In data-intensive clinical environments, such as intensive care units (ICUs), clinical patient data are already fully managed by such systems in a number of hospitals. However, providing facilities for storing and retrieving patient data to clinicians is not enough; clinical information systems should also offer facilities to assist clinicians in dealing with hard clinical problems. Extending an information system's capabilities by integrating it with a decision-support system may be a solution. In this paper, we describe the development of a probabilistic and decision-theoretic system that aims to assist clinicians in diagnosing and treating patients with pneumonia in the intensive-care unit. Its underlying probabilistic-network model includes tempo...

Intelligent systems are increasingly being deployed in medicine and healthcare, but there is a need for a robust and objective methodology for evaluating such systems. Potentially, receiver operating characteristic (ROC) analysis could form a basis for the objective evaluation of intelligent medical systems. However, it has several weaknesses when applied to the types of data used to evaluate intelligent medical systems. First, small data sets are often used, which are unsatisfactory with existing methods. Second, many existing ROC methods use parametric assumptions which may not always be valid for the test cases selected. Third, system evaluations are often more concerned with particular, clinically meaningful, points on the curve, rather than on global indexes such as the more commonly used area under the curve. A novel, robust and accurate method is proposed, derived from first principles, which calculates the probability density function (pdf) for each point on a ROC curve for any given sample size. Confidence intervals are produced as contours on the pdf. The theoretical work has been validated by Monte Carlo simulations. It has also been applied to two real-world examples of ROC analysis, taken from the literature (classification of mammograms and differential diagnosis of pancreatic diseases), to investigate the confidence surfaces produced for real cases, and to illustrate how analysis of system performance can be enhanced. We illustrate the impact of sample size on system performance from analysis of ROC pdf’s and 95% confidence boundaries. This work establishes an important new method for generating pdf’s, and provides an accurate and robust method of producing confidence intervals for ROC curves for the small sample sizes typical of intelligent medical systems. It is conjectured that, potentially, the method could be extended to determine risks associated with the deployment of intelligent medical systems in clinical practice.

Assessing software costs money and better assessment costs exponentially more money. Given finite budgets, assessment resources are typically skewed towards areas that are believed to be mission critical. This leaves blind spots: portions of the system that may contain defects which may be missed. Therefore, in addition to rigorously assessing mission critical areas, a parallel activity should sample the blind spots. This paper assesses defect detectors based on static code measures as a blind spot sampling method. In contrast to previous results, we find that such defect detectors yield results that are stable across many applications. Further, these detectors are inexpensive to use and can be tuned to the specifics of the current business situations. 1

I hereby declare that I am the sole author of this thesis. This is a true copy of the thesis, including any required final revisions, as accepted by my examiners. I understand that my thesis may be made electronically available to the public. ii The work reported in the thesis consists of two parts. One part is concerned with the development of a feature selection method based on statistical significance test, which can be generally used in any supervised pattern classification. The other part applies this proposed feature selection method to conduct proteomic pattern analysis for prostate cancer detection. For a given classification problem, we need to determine a set of relevant features to generate a classifier. In real-world problems, many features in initial feature set are usually irrelevant to the classification task and redundant with each other, which will increase the computational complexity and reduce the recognition rate. The task of feature selection is to choose a small feature subset in order to achieve better classification performance. As such,

The construction of a neural network-based decision support system for the diagnosis of acute abdominal pain was investigated. Different neural network algorithms were compared to define the optimal algorithms for this diagnostic classification problem. The problem of missing input data values was examined with various replacement techniques. Special attention was paid to the evaluation of confidence for the outputs of the networks. The results of the classification were also examined with different databases from two countries (Finland and Germany). The results were also compared with the results of statistical analysis. In our tests, the two best neural network algorithms, backpropagation (BP) and learning vector quantization (LVQ), classified patient cases with a very high degree of accuracy (90%), which is as high as that achieved in the best studies using other methods. A new method to present the results of the classification with the LVQ algorithm was developed. The use of k-nea...

This paper sets out a novel method for generating robust and accurate probability distributions for all points of a non--parametric ROC curve based on samples of any size. From these probabilty distributions it is straight forward to derive confidence boundaries as required

CADIAG-2 is a medical expert system to assist differential diagnosis in several sub-specialties of internal medicine. A patient’s symptoms, signs, laboratory test results, and various clinical findings constitute the starting point of the computer-assisted differential diagnostic process. Lists of confirmed and excluded diagnoses as well as diagnostic hypotheses are the output. In this paper we logically verify CADIAG-2’s knowledge base which consists of about 20,000 rules, by using a classical theorem prover. We identified ten inaccuracies in the present knowledge base. One of the inaccuracies is presented and discussed here.