Predicting Patient’s Long-Term Clinical Status after Hip Arthroplasty Using Hierarchical Decision Modelling and Data Mining

Construction of a prognostic model is presented for the long-term outcome after femoral neck fracture treatment with implantation of hip endoprosthesis. While the model is induced from the follow-up data, we show that the use of additional expert knowledge is absolutely crucial to obtain good predictive accuracy. A schema is proposed where domain knowledge is encoded as a hierarchical decision model of which only a part is induced from the data while the rest is specified by the expert. Although applied to hip endoprosthesis domain, the proposed schema is general and can be used for the construction of other prognostic models where both follow-up data and human expertise is available.

GenePath: a computer program for genetic pathway discovery from mutant data

The sequencing of the human genome and the genomes of several model organisms is the first step toward the long-term objective of genetic research: the identification of all genes, and the discovery of their functions and mutual interactions. This article presents a methodology and a computer program called GenePath to support the discovery of gene function. GenePath uses mutant data and available genetic knowledge to identify potential genetic pathways. Several pilot applications based on experimental results from Dictyostelium and C. elegans confirmed the usefulness of the proposed schema. Our results suggest that GenePath is a valuable tool that can be used as an intelligent assistant to support genetic reasoning.

Induction of hypotheses concerning hip arthroplasty: a modified methodology for medical research

OBJECTIVES

The objective of this study is to advocate a methodology for medical research that, in contrast to traditional medical methodology, exploits the flexibility of machine learning and retains the kind of statistical tests that are generally accepted in the medical field for the confirmation of hypotheses.

METHODS

First, the medical problem is defined and data for an observed population are collected; then a machine learning tool is used to generate hypotheses regarding the problem; finally, statistical methods are used to determine the validity of the generated hypotheses.

RESULTS

To illustrate this approach, the problem of defining indications for hip arthroplasty after an acute medial femoral neck fracture is investigated as a case study.

CONCLUSIONS

The methodology is similar to the usual style of applying machine learning, but insists on a link to the techniques of statistical tests that are normally used in medicine. It aims at a more flexible and economical use of experimental data than in the usual medical research, which is enabled by techniques of machine learning. At the same time, by reference to traditional statistical tests, it is hoped that this approach will lead to improved acceptance of machine learning in the medical field.

Decisions at hand: a decision support system on handhelds

One of the applications of clinical information systems is decision support. Although the advantages of utilizing such aids have never been theoretically disputed, they have been rarely used in practice. The factor that probably often limits the utility of clinical decision support systems is the need for computing power at the very site of decision making--at the place where the patient is interviewed, in discussion rooms, etc. The paper reports on a possible solution to this problem. A decision-support shell LogReg is presented, which runs on a handheld computer. A general schema for handheld-based decision support is also proposed, where decision models are developed on personal computers/workstations, encoded in XML and then transferred to handhelds, where the models are used within a decision support shell. A use case where LogReg has been applied to clinical outcome prediction in crush injury is presented.

Predicting patient's long-term clinical status after hip arthroplasty using hierarchical decision modelling and data mining

Construction of a prognostic model is presented for the long-term outcome after femoral neck fracture treatment with implantation of hip endoprosthesis. While the model is induced from the follow-up data, we show that the use of additional expert knowledge is absolutely crucial to obtain good predictive accuracy. A schema is proposed where domain knowledge is encoded as a hierarchical decision model of which only a part is induced from the data while the rest is specified by the expert. Although applied to hip endoprosthesis domain, the proposed schema is general and can be used for the construction of other prognostic models where both follow-up data and human expertise is available.

Machine learning for survival analysis: a case study on recurrence of prostate cancer

Machine learning techniques have recently received considerable attention, especially when used for the construction of prediction models from data. Despite their potential advantages over standard statistical methods, like their ability to model non-linear relationships and construct symbolic and interpretable models, their applications to survival analysis are at best rare, primarily because of the difficulty to appropriately handle censored data. In this paper we propose a schema that enables the use of classification methods--including machine learning classifiers--for survival analysis. To appropriately consider the follow-up time and censoring, we propose a technique that, for the patients for which the event did not occur and have short follow-up times, estimates their probability of event and assigns them a distribution of outcome accordingly. Since most machine learning techniques do not deal with outcome distributions, the schema is implemented using weighted examples. To show the utility of the proposed technique, we investigate a particular problem of building prognostic models for prostate cancer recurrence, where the sole prediction of the probability of event (and not its probability dependency on time) is of interest. A case study on preoperative and postoperative prostate cancer recurrence prediction shows that by incorporating this weighting technique the machine learning tools stand beside modern statistical methods and may, by inducing symbolic recurrence models, provide further insight to relationships within the modeled data.

Naive Bayesian-based nomogram for prediction of prostate cancer recurrence

This paper introduces a schema with naive-Bayesian classifier and patient weighting technique to develop a prostate cancer recurrence prediction model from patient data. We propose the graphical presentation of naive-Bayesian classifier with a nomogram, which can be used both for prediction or can provide means to data analysis. The resulting model was experimentally evaluated; the results were favorable both in terms of interpretability and predictive accuracy.