Herbert F. Jelinek

Fractals in the Neurosciences, Part I General Principles and Basic Neurosciences

The natural complexity of the brain, its hierarchical structure, and the sophisticated topological architecture of the neurons organized in micronetworks and macronetworks are all factors contributing to the limits of the application of Euclidean geometry and linear dynamics to the neurosciences. The introduction of fractal geometry for the quantitative analysis and description of the geometric complexity of natural systems has been a major paradigm shift in the last decades. Nowadays, modern neurosciences admit the prevalence of fractal properties such as self-similarity in the brain at various levels of observation, from the microscale to the macroscale, in molecular, anatomic, functional, and pathological perspectives. Fractal geometry is a mathematical model that offers a universal language for the quantitative description of neurons and glial cells as well as the brain as a whole, with its complex three-dimensional structure, in all its physiopathological spectrums. For a holistic view of fractal geometry of the brain, we review here the basic concepts of fractal analysis and its main applications to the basic neurosciences.

Assessing the Need for Referral in Automatic Diabetic Retinopathy Detection

Emerging technologies in health care aim at reducing unnecessary visits to medical specialists, minimizing overall cost of treatment and optimizing the number of patients seen by each doctor. This paper explores image recognition for the screening of diabetic retinopathy, a complication of diabetes that can lead to blindness if not discovered in its initial stages. Many previous reports on DR imaging focus on the segmentation of the retinal image, on quality assessment, and on the analysis of presence of DR-related lesions. Although this study has advanced the detection of individual DR lesions from retinal images, the simple presence of any lesion is not enough to decide on the need for referral of a patient. Deciding if a patient should be referred to a doctor is an essential requirement for the deployment of an automated screening tool for rural and remote communities. We introduce an algorithm to make that decision based on the fusion of results by metaclassification. The input of the metaclassifier is the output of several lesion detectors, creating a powerful high-level feature representation for the retinal images. We explore alternatives for the bag-of-visual-words (BoVW)-based lesion detectors, which critically depends on the choices of coding and pooling the low-level local descriptors. The final classification approach achieved an area under the curve of 93.4% using SOFT-MAX BoVW (soft-assignment coding/max pooling), without the need of normalizing the high-level feature vector of scores.

ECG Biometric with Abnormal Cardiac Conditions in Remote Monitoring System

This paper presents a person identification mechanism using electrocardiogram (ECG) signals with abnormal cardiac conditions in network environments. A total of 164 subjects were used in this paper using three different databases containing various irregular heart states from MIT-BIH arrhythmia database (MITDB), MIT-BIH supraventricular arrhythmia database (SVDB), and Charles Sturt diabetes complication screening initiative (DiSciRi) database. We proposed a simple yet effective biometric sample extraction technique for ECG samples with abnormal cardiac conditions to improve the person identification process. These sample points were then applied to four classifiers to verify the robustness of identification. Varying numbers of enrollment and recognition QRS complexes were used to validate the stability of the proposed method. Our experimentation results show that the biometric technique outperforms existing methods lacking the ability to efficiently extract features for biometric matching. This is evident by obtaining high accuracy results of 96.7% for MITDB, 96.4% for SVDB, and 99.3% for DiSciRi. Moreover, high sensitivity, specificity, positive predictive value, and Youden Indexs values further verifies the reliability of the proposed method. This technique also suggests the possibility of improving the classification performance using ECG recordings with low sampling frequency and increased number of ECG samples.

Association of cardiovascular risk using non-linear heart rate variability measures with the framingham risk score in a rural population

Cardiovascular risk can be calculated using the Framingham cardiovascular disease (CVD) risk score and provides a risk stratification from mild to very high CVD risk percentage over 10 years. This equation represents a complex interaction between age, gender, cholesterol status, blood pressure, diabetes status, and smoking. Heart rate variability (HRV) is a measure of how the autonomic nervous system (ANS) modulates the heart rate. HRV measures are sensitive to age, gender, disease status such as diabetes and hypertension and processes leading to atherosclerosis. We investigated whether HRV measures are a suitable, simple, noninvasive alternative to differentiate between the four main Framingham associated CVD risk categories. In this study we applied the tone-entropy (T-E) algorithm and complex correlation measure (CCM) for analysis of HRV obtained from 20 min. ECG recordings and correlated the HRV score with the stratification results using the Framingham risk equation. Both entropy and CCM had significant analysis of variance (ANOVA) results [F(172, 3) = 9.51; <0.0001]. Bonferroni post hoc analysis indicated a significant difference between mild, high and very high cardiac risk groups applying tone-entropy (p < 0.01). CCM detected a difference in temporal dynamics of the RR intervals between the mild and very high CVD risk groups (p < 0.01). Our results indicate a good agreement between the T-E and CCM algorithm and the Framingham CVD risk score, suggesting that this algorithm may be of use for initial screening of cardiovascular risk as it is noninvasive, economical and easy to use in clinical practice.

Predicting cardiac autonomic neuropathy category for diabetic data with missing values

Cardiovascular autonomic neuropathy (CAN) is a serious and well known complication of diabetes. Previous articles circumvented the problem of missing values in CAN data by deleting all records and fields with missing values and applying classifiers trained on different sets of features that were complete. Most of them also added alternative features to compensate for the deleted ones. Here we introduce and investigate a new method for classifying CAN data with missing values. In contrast to all previous papers, our new method does not delete attributes with missing values, does not use classifiers, and does not add features. Instead it is based on regression and meta-regression combined with the Ewing formula for identifying the classes of CAN. This is the first article using the Ewing formula and regression to classify CAN. We carried out extensive experiments to determine the best combination of regression and meta-regression techniques for classifying CAN data with missing values. The best outcomes have been obtained by the additive regression meta-learner based on M5Rules and combined with the Ewing formula. It has achieved the best accuracy of 99.78% for two classes of CAN, and 98.98% for three classes of CAN. These outcomes are substantially better than previous results obtained in the literature by deleting all missing attributes and applying traditional classifiers to different sets of features without regression. Another advantage of our method is that it does not require practitioners to perform more tests collecting additional alternative features.

Poincaré plot analysis of heart rate variability in the diabetic patients in the UAE

Major complications such as cardiac death and cardiac autonomic neuropathy are caused by diabetic autonomic neuropathy. Heart Rate Variability (HRV) analysis has shown to detect variations in the autonomic balance of heart rate and is useful for early detection of autonomic dysfunction. This study presents the outcome of HRV analysis of short ECG recordings taken from nondiabetic and type 2 diabetes patients, applying Poincaré plot indices represented by short term variation (SD1), long term variation (SD2) and complex correlation (CCM) measure which measures the temporal dynamics, for early detection of cardiac autonomic neuropathy. SD1 and the ratio SD1/SD2 were found to be significantly lower in type 2 diabetes patients than the control group. The highest discriminatory power was observed with CCM, indicating the advantage of using a dynamic measure for HRV rather than the static Poincaré plot indices. SD1 and CCM could be markers for CVD risk in type 2 diabetic patients.

Influence of stroke location on heart rate variability in robot-assistive neurorehabilitation

Active mental engagement and a positive emotional state are prerequisites for optimal outcomes of rehabilitation programs for stroke patients. Our program at the ETH, Zurich utilizes a closed loop response in automated robot-assist gait training coupled with virtual reality provided tasks. Heart rate variability has been shown to be sensitive to cognitive as well as emotional states as well as pathophysiological-environmental challenges. We investigated whether adaptation to a task differs between stroke patients with either cortical or subcortical lesions. Seven non-stroke control participants were compared to responses of nine stroke patients with either a diagnosis of cortical or subcortical stroke using heart rate variability. The robot-assist virtual reality training session consisted of a familiarization period, a baseline walking period, an under-challenged, appropriate challenged and over-challenged condition. Time and frequency domain as well as nonlinear features were assessed. Our results indicated that only entropy was sensitive to identifying adaptation to a different level of difficulty. Thus a significant difference was seen between the three stroke groups and control for adaption from baseline to the under-challenged condition (p=0.026), and also from the under-challenged condition to an appropriate level of challenge (p=0.027). We propose that the entropy feature provides a robust index of cognitive and emotional level associated with task difficulty experienced by post-stroke patients that allows real time closed loop regulation of robot-assist gait rehabilitation.

Neurons of the Human Dentate Nucleus: Box-Count Method in the Quantitative Analysis of Cell Morphology

The morphology of neurons from the human dentate nucleus was analyzed estimating the size and shape of the dendritic field, shape of the neuron, space-filling property and the degree of dendrite aberrations. Among them, the last three morphological properties were investigated using the most popular technique of fractal analysis: the box-count method. The box dimensions of binary images and dendritic field area were statistically investigated in order to test whether the binary box dimension can quantify the size of the neuron. The same analysis was carried out using the box dimension of outline images and image circularity. The parameters, presented in this study have proved to be a useful means for quantifying the morphology of dentate neurons as they provide a robust means of differentiating between neuron subtypes in the dentate nucleus. The findings of the present study are in accordance with previous qualitative data.

Richardson's Method of Segment Counting versus Box-Counting

Fractal analysis has become a popular method in all branches of scientific investigation including biology and medicine. This paper presents solution for many unresolved questions about the methodology of fractal theory, precisely in connection between fractal geometry and fractal analysis. While some concepts in fractal theory are determined descriptively and/or qualitatively, this paper provides their exact mathematical definition or explanation. Also, we present results in applying two basic length-related methods on two dimensional neuronal images and discuss their applicability.

Machine Learning Methods for Automated Detection of Severe DiabeticNeuropathy

Objective: The present study aimed at investigating machine learning methods for automated detection of severe diabetic neuropathy. Severe diabetic neuropathy represents a significant neurological problem in diabetes as it requires urgent intervention to reduce the risk of sudden cardiac death. Automated detection provides a tool that can be applied to clinical data and for identifying comorbidities that can trigger diagnosis and treatment. Methods: We applied multi scale Allen factor to determine heart rate variability, a marker for diabetic neuropathy from ECG recordings as features to be used for the machine learning methods and automated detection. The major innovation of this work is the introduction of a new Graph-Based Machine Learning System (GBMLS). This method is intended to enhance the effectiveness of the diagnosis of severe diabetic neuropathy. We applied it to the multi scale Allen factor (MAF) features as a collection of attributes determined from the recorded ECG bio signals. These attributes can be collected as a result of routine ECG investigation of patients regardless of the presenting medical problems. Results: Our experiments compared the sensitivity and specificity of the automated detection produced by GBMLS with analogous outcomes achieved by various other machine learning approaches. To this end we used a comprehensive collection of important classifiers and clusterers available in the open source machine learning software package Scikit-learn. The experiments have demonstrated that the best outcomes were obtained by GBMLS in combination with MAF, which improved sensitivity to 0.89 and specificity to 0.98 and outperformed several other classifiers and clusterers including Random Forest with sensitivity of 0.83 and specificity of 0.92. Conclusion: The novel GBMLS machine learning technique applied to MAF attributes has outperformed other machine learning methods and achieved excellent sensitivity and specificity. These results are significant and sufficiently effective to be recommended for practical application of this technique.