C. Okan Sakar

Telediagnosis of Parkinson’s Disease Using Measurements of Dysphonia

Parkinson’s disease (PD) is a neurological illness which impairs motor skills, speech, and other functions such as mood, behavior, thinking, and sensation. It causes vocal impairment for approximately 90% of the patients. As the symptoms of PD occur gradually and mostly targeting the elderly people for whom physical visits to the clinic are inconvenient and costly, telemonitoring of the disease using measurements of dysphonia (vocal features) has a vital role in its early diagnosis. Such dysphonia features extracted from the voice come in variety and most of them are interrelated. The purpose of this study is twofold: (1) to select a minimal subset of features with maximal joint relevance to the PD-score, a binary score indicating whether or not the sample belongs to a person with PD; and (2) to build a predictive model with minimal bias (i.e. to maximize the generalization of the predictions so as to perform well with unseen test examples). For these tasks, we apply the mutual information measure with the permutation test for assessing the relevance and the statistical significance of the relations between the features and the PD-score, rank the features according to the maximum-relevance-minimum-redundancy (mRMR) criterion, use a Support Vector Machine (SVM) for building a classification model and test it with a more suitable cross-validation scheme that we called leave-one-individual-out that fits with the dataset in hand better than the conventional bootstrapping or leave-one-out validation methods.

A feature selection method based on kernel canonical correlation analysis and the minimum Redundancy-Maximum Relevance filter method

Highlights? We propose a feature selection method based on a recently popular mRMR criterion, which we called KCCAmRMR. ? Our method is based on finding the unique information that a candidate variable possesses about the target variable. ? We propose using correlated functions explored by KCCA instead of using the features themselves as inputs to mRMR. In this paper, we propose a feature selection method based on a recently popular minimum Redundancy-Maximum Relevance (mRMR) criterion, which we called Kernel Canonical Correlation Analysis basedmRMR (KCCAmRMR) based on the idea of finding the unique information, i.e. information that is distinct from the set of already selected variables, that a candidate variable possesses about the target variable. In simplest terms, for this purpose, we propose using correlated functions explored by KCCA instead of using the features themselves as inputs to mRMR. We demonstrate the usefulness of our method on both toy and benchmark datasets.

Pulmonary crackle detection using time-frequency and time-scale analysis

Pulmonary crackles are used as indicators for the diagnosis of different pulmonary disorders in auscultation. Crackles are very common adventitious transient sounds. From the characteristics of crackles such as timing and number of occurrences, the type and the severity of the pulmonary diseases may be assessed. In this study, a method is proposed for crackle detection. In this method, various feature sets are extracted using time-frequency and time-scale analysis from pulmonary signals. In order to understand the effect of using different window and wavelet types in time-frequency and time-scale analysis in detecting crackles, different windows and wavelets are tested such as Gaussian, Blackman, Hanning, Hamming, Bartlett, Triangular and Rectangular windows for time-frequency analysis and Morlet, Mexican Hat and Paul wavelets for time-scale analysis. The extracted feature sets, both individually and as an ensemble of networks, are fed into three different machine learning algorithms: Support Vector Machines, k-Nearest Neighbor and Multilayer Perceptron. Moreover, in order to improve the success of the model, prior to the time-frequency/scale analysis, frequency bands containing no-crackle information are removed using dual-tree complex wavelet transform, which is a shift invariant transform with limited redundancy compared to the conventional discrete wavelet transform. The comparative results of individual feature sets and ensemble of sets, which are extracted using different window and wavelet types, for both pre-processed and non-pre-processed data with different machine learning algorithms, are extensively evaluated and compared.

Feature extraction using time-frequency/scale analysis and ensemble of feature sets for crackle detection

Pulmonary crackles are used as indicators for the diagnosis of different pulmonary disorders. Crackles are very common adventitious sounds which have transient characteristic. From the characteristics of crackles such as timing and number of occurrences, the type and the severity of the pulmonary diseases can be obtained. In this study, a novel method is proposed for crackle detection. In this method, various feature sets are extracted using time-frequency and time-scale analysis. The extracted feature sets are fed into support vector machines both individually and as an ensemble of networks. Besides, as a preprocessing stage in order to improve the success of the model, frequency bands containing no-information are removed using dual tree complex wavelet transform, which is a shift invariant transform with limited redundancy and an improved version of discrete wavelet transform. The comparative results of individual feature sets and ensemble of sets with pre-processed and non pre-processed data are proposed.

A method for combining mutual information and canonical correlation analysis

Highlights? We propose a hybrid measure of relevance based on MI and KCCA. ? Our measure PMI weighs more the samples with predictive powers. ? PMI effectively eliminates the samples with no predictive contribution. ? We show that PMI has improved feature detection capability. Feature selection is a critical step in many artificial intelligence and pattern recognition problems. Shannons Mutual Information (MI) is a classical and widely used measure of dependence measure that serves as a good feature selection algorithm. However, as it is a measure of mutual information in average, under-sampled classes (rare events) can be overlooked by this measure, which can cause critical false negatives (missing a relevant feature very predictive of some rare but important classes). Shannons mutual information requires a well sampled database, which is not typical of many fields of modern science (such as biomedical), in which there are limited number of samples to learn from, or at least, not all the classes of the target function (such as certain phenotypes in biomedical) are well-sampled. On the other hand, Kernel Canonical Correlation Analysis (KCCA) is a nonlinear correlation measure effectively used to detect independence but its use for feature selection or ranking is limited due to the fact that its formulation is not intended to measure the amount of information (entropy) of the dependence. In this paper, we propose a hybrid measure of relevance, Predictive Mutual Information (PMI) based on MI, which also accounts for predictability of signals from each other in its calculation as in KCCA. We show that PMI has more improved feature detection capability than MI, especially in catching suspicious coincidences that are rare but potentially important not only for experimental studies but also for building computational models. We demonstrate the usefulness of PMI, and superiority over MI, on both toy and real datasets.

Online naive bayes classification for network intrusion detection

Intrusion detection system (IDS) is an important component to ensure network security. In this paper we build an online Naïve Bayes classifier to discriminate normal and bad (intrusion) connections on KDD 99 dataset for network intrusion detection. The classifier starts with a small number of training examples of normal and bad classes; then, as it classifies the rest of the samples one at a time, it continuously updates the mean and the standard deviations of the features (IDS variables). We present experimental results of parameter updating methods and their parameters for the online Naïve Bayes classifier. The obtained results show that our proposed method performs comparably to the simple incremental update.

Ensemble canonical correlation analysis

Canonical Correlation Analysis (CCA) aims at identifying linear dependencies between two different but related multivariate views of the same underlying semantics. Ignoring its various extensions to more than two views, CCA uses these two views as complex labels to guide the search of maximally correlated projection vectors (covariates). Therefore, CCA can overfit the training data, meaning that different correlated projections can be found when the two-view training dataset is resampled. Although, to avoid such overfitting, ensemble approaches that utilize resampling techniques have been effectively used for improving generalization of many machine learning methods, an ensemble approach has not yet been formulated for CCA. In this paper, we propose an ensemble method for obtaining a final set of covariates by combining multiple sets of covariates extracted from subsamples. In comparison to those obtained by the application of the classical CCA on the whole set of training data, combining covariates with weaker correlations extracted from a number of subsamples of the training data produces stronger correlations that generalize to unseen test examples. Experimental results on emotion recognition, digit recognition, content-based retrieval, and multiple view object recognition have shown that ensemble CCA has better generalization for both the test set correlations of the covariates and the test set accuracy of classification performed on these covariates.

A Hybrid Method for Feature Selection Based on Mutual Information and Canonical Correlation Analysis

Mutual Information (MI) is a classical and widely used dependence measure that generally can serve as a good feature selection algorithm. However, under-sampled classes or rare but certain relations are overlooked by this measure, which can result in missing relevant features that could be very predictive of variables of interest, such as certain phenotypes or disorders in biomedical research, rare but dangerous factors in ecology, intrusions in network systems, etc. On the other hand, Kernel Canonical Correlation Analysis (KCCA) is a nonlinear correlation measure effectively used to detect independence but its use for feature selection or ranking is limited due to the fact that its formulation is not intended to measure the amount of information (entropy) of the dependence. In this paper, we propose Predictive Mutual Information (PMI), a hybrid measure of relevance not only is based on MI but also accounts for predictability of signals from one another as in KCCA. We show that PMI has more improved feature detection capability than MI and KCCA, especially in catching suspicious coincidences that are rare but potentially important not only for subsequent experimental studies but also for building computational predictive models which is demonstrated on two toy datasets and a real intrusion detection system dataset.

Determination of the optimal threshold value that can be discriminated by dysphonia measurements for unified Parkinson's Disease rating scale

Recently, there is an increasing motivation to develop telemonitoring systems that enable cost-effective screening of Parkinsons Disease (PD) patients. These systems are generally based on measuring the motor system disorders seen in PD patients by the help of non-invasive data collection tools. Vocal impairments one of the most commonly seen PD symptoms in the early stages of the disease, and building such telemonitoring systems based on detecting the level of vocal impairments results in reliable motor UPDRS tracking systems. In this paper, we aim to determine the optimal UPDRS threshold value that can be discriminated by the vocal features extracted from the sustained vowel phonations of PD patients. For this purpose, we used an online available PD telemonitoring dataset consisting of speech recordings of 42 PD patients. We converted the UPDRS prediction problem into a binary classification problem for various motor UPDRS threshold values, and fed the features to k-Nearest Neighbor and Support Vector Machines classifiers to discriminate the PD patients whose UPDRS is less than or greater than the specified threshold value. The results indicate that speech disorders are more significantly seen in the patients whose UPDRS exceeds the experimentally determined threshold value (15). Besides, considering that the motor UPDRS ranges from 0 to 108, relatively low UPDRS threshold of 15 validates that vocal impairments can be used as early indicators of the disease.

Effect of different window and wavelet types on the performance of a novel crackle detection algorithm

Pulmonary crackles are used as indicators for the diagnosis of different pulmonary disorders. Crackles are very common adventitious sounds which have transient characteristic. From the characteristics of crackles such as timing and number of occurrences, the type and the severity of the pulmonary diseases can be obtained. In this study, a novel method is proposed for crackle detection, which uses time- frequency and time-scale analysis, and the performance comparison for different window types in time-frequency analysis and also for different wavelet types in time-scale analysis is presented. In the proposed method, various feature sets are extracted using time-frequency and time-scale analysis for different windows and wavelet types. The extracted feature sets are fed into support vector machines both individually and as an ensemble of networks. Besides, as a preprocessing stage in order to improve the success of the model, frequency bands containing no-information are removed using dual tree complex wavelet transform, which is a shift invariant transform with limited redundancy and an improved version of discrete wavelet transform. The comparative results of individual feature sets and ensemble of sets with pre-processed and non pre-processed data for different windows and wavelets are proposed.