Modeling Interpretable Fuzzy Rule-Based Classifiers for Medical Decision Support

Decision support systems in Medicine must be easily comprehensible, both for physicians and patients. In this chapter, the authors describe how the fuzzy modeling methodology called HILK (Highly Interpretable Linguistic Knowledge) can be applied for building highly interpretable fuzzy rule-based classifiers (FRBCs) able to provide medical decision support. As a proof of concept, they describe the case study of a real-world scenario concerning the development of an interpretable FRBC that can be used to predict the evolution of the end-stage renal disease (ESRD) in subjects affected by Immunoglobin A Nephropathy (IgAN). The designed classifier provides users with a number of rules which are easy to read and understand. The rules classify the prognosis of ESRD evolution in IgAN-affected subjects by distinguishing three classes (short, medium, long). Experimental results show that the fuzzy classifier is capable of satisfactory accuracy results – in comparison with Multi-Layer Perceptron (MLP) neural networks – and high interpretability of the knowledge base.

Kernel Generative Topographic Mapping of Protein Sequences

The world of pharmacology is becoming increasingly dependent on the advances in the fields of genomics and proteomics. The –omics sciences bring about the challenge of how to deal with the large amounts of complex data they generate from an intelligence data analysis perspective. In this chapter, the authors focus on the analysis of a specific type of proteins, the G protein-couple receptors, which are the target for over 15% of current drugs. They describe a kernel method of the manifold learning family for the analysis of protein amino acid symbolic sequences. This method sheds light on the structure of protein subfamilies, while providing an intuitive visualization of such structure.

Development of Surrogate Models of Orthopedic Screws to Improve Biomechanical Performance

An optimization approach was applied to improve the design of the lag screws used in double screw nails. However, finite element analyses with an optimal algorithm may take a long time to find the best design. Thus, surrogate methods, either artificial neural networks or multiple linear regressions, were used to substitute for the finite element models. The results showed that an artificial neural network method can accurately develop the objective functions of the lag screws for both the bending strength and the pullout strength. A multiple linear regression method can successfully develop the objective function of the lag screws for the pullout strength, but it failed to construct the objective function for the bending strength. The optimal design of the lag screws could be obtained using the artificial neural network method and genetic algorithms.

Pattern Mining for Outbreak Discovery Preparedness

Today, the objective of public health surveillance system is to reduce the impact of outbreaks by enabling appropriate intervention. Commonly used techniques are based on the changes or aberration in health events when compared with normal history to detect an outbreak. The main problem encountered in outbreaks is high rates of false alarm. High false alarm rates can lead to unnecessary interventions, and falsely detected outbreaks will lead to costly investigation. In this chapter, the authors review data mining techniques focusing on frequent and outlier mining to develop generic outbreak detection process model, named as “Frequent-outlier” model. The process model was tested against the real dengue dataset obtained from FSK, UKM, and also tested on the synthetic respiratory dataset obtained from AUTON LAB. The ROC was run to analyze the overall performance of “frequent-outlier” with CUSUM and Moving Average (MA). The results were promising and were evaluated using detection rate, false positive rate, and overall performance. An important outcome of this study is the knowledge rules derived from the notification of the outbreak cases to be used in counter measure assessment for outbreak preparedness.

Intelligent Management of Sepsis in the Intensive Care Unit

Sepsis is a transversal pathology and one of the main causes of death in the Intensive Care Unit (ICU). It has in fact become the tenth most common cause of death in western societies. Its mortality rates can reach up to 60% for Septic Shock, its most acute manifestation. For these reasons, the prediction of the mortality caused by Sepsis is an open and relevant medical research challenge. This problem requires prediction methods that are robust and accurate, but also readily interpretable. This is paramount if they are to be used in the demanding context of real-time decision making at the ICU. In this brief contribution, three different methods are presented. One is based on a variant of the well-known support vector machine (SVM) model and provides and automated ranking of relevance of the mortality predictors while the other two are based on logistic-regression and logistic regression over latent Factors. The reported results show that the methods presented outperform in terms of accuracy alternative techniques currently in use in clinical settings, while simultaneously assessing the relative impact of individual pathology indicators.

Extraction of Medical Pathways from Electronic Patient Records

With the introduction of electronic medical records, a large amount of patients’ medical data has been available. An actual problem in this domain is to perform reverse engineering of the medical treatment process to highlight medical pathways typically adopted for specific health conditions. This chapter addresses the ability of sequential data mining techniques to reconstruct the actual medical pathways followed by patients. Detected medical pathways are in the form of sets of exams frequently done together, sequences of exam sets frequently followed by patients and frequent correlations between exam sets. The analysis shows that the majority of the extracted pathways are consistent with the medical guidelines, but also reveals some unexpected results, which can be useful both to enrich existing guidelines and to improve the public sanitary service.

Visual Data Mining in Physiotherapy Using Self-Organizing Maps

The basis of all clinical science developments is the analysis of the data obtained from a particular problem. In recent decades, however, the capacity of computers to process data has been increasing exponentially, which has created the possibility of applying more powerful methods of data analysis. Among these methods, the multidimensional visual data mining methods are outstanding. These methods show all the variables of one particular problem on the whole allowing to the clinical specialist to extract his own conclusions. In this chapter, a neural approximation to this kind of data mining is shown by means of the valuation analysis of the knee in athletes in the pre- and post-surgery of the anterior cruciate ligament, studying variables of force and measurements at different distances of the knee.

The Use of Prediction Reliability Estimates on Imbalanced Datasets

Data mining techniques are extensively used on medical data, which is typically composed of many normal examples and few interesting ones. When presented with highly imbalanced data, some standard classifiers tend to ignore the minority class which leads to poor performance. Various solutions have been proposed to counter this problem. Random undersampling, random oversampling, and SMOTE (Synthetic Minority Oversampling Technique) are the most well-known approaches. In recent years several approaches to evaluate the reliability of single predictions have been developed. Most recently a simple and efficient approach, based on the classifier’s class probability estimates was shown to outperform the other reliability estimates. The authors propose to use this reliability estimate to improve the SMOTE algorithm. In this study, they demonstrate the positive effects of using the proposed algorithms on artificial datasets. The authors then apply the developed methodology on the problem of predicting the maximal wall shear stress (MWSS) in the human carotid artery bifurcation. The results indicate that it is feasible to improve the classifier’s performance by balancing the data with their versions of the SMOTE algorithm.

Automated Diagnostics of Coronary Artery Disease

The authors present results and the latest advancement in their long-term study on using image processing and data mining methods in medical image analysis in general, and in clinical diagnostics of coronary artery disease in particular. Since the evaluation of modern medical images is often difficult and time-consuming, authors integrate advanced analytical and decision support tools in diagnostic process. Partial diagnostic results, frequently obtained from tests with substantial imperfections, can be thus integrated in ultimate diagnostic conclusion about the probability of disease for a given patient. Authors study various topics, such as improving the predictive power of clinical tests by utilizing pre-test and post-test probabilities, texture representation, multi-resolution feature extraction, feature construction and data mining algorithms that significantly outperform the medical practice. During their long-term study (1995-2011) authors achieved, among other minor results, two really significant milestones. The first was achieved by using machine learning to significantly increase post-test diagnostic probabilities with respect to expert physicians. The second, even more significant result utilizes various advanced data analysis techniques, such as automatic multi-resolution image parameterization combined with feature extraction and machine learning methods to significantly improve on all aspects of diagnostic performance. With the proposed approach clinical results are significantly as well as fully automatically, improved throughout the study. Overall, the most significant result of the work is an improvement in the diagnostic power of the whole diagnostic process. The approach supports, but does not replace, physicians’ diagnostic process, and can assist in decisions on the cost-effectiveness of diagnostic tests.

Building a Lazy Domain Theory for Characterizing Malignant Melanoma

In this chapter, the authors propose an approach for building a model characterizing malignant melanomas. A common way to build a domain model is using an inductive learning method. Such resulting model is a generalization of the known examples. However, in some domains where there is not a clear difference among the classes, the inductive model could be too general. The approach taken in this chapter consists of using lazy learning methods for building what the authors call a lazy domain theory. The main difference between both inductive and lazy theories is that the former is complete whereas the latter is not. This means that the lazy domain theory may not cover all the space of known examples. The authors’ experiments have shown that, despite of this, the lazy domain theory has better performance than the inductive theory.