Brain Informatics and Health

Neuronal morphology modeling is one of the key steps for reverse engineering the brain at the micro level. It creates a realistic digital version of the neuron obtained by microscopy reconstruction in a visualized way so that the structure of the whole neuron (including soma, dendrite, axon, spin, etc.) is visible in different angles in a three dimensional space. Whether the modeled neuronal morphology matches the original neuron in vivo is closely related to the details captured by the manually sampled morphological points. Many data in public neuronal morphology data repositories (such as the NeuroMorpho project) focus more on the morphology of dendrites and axons, while there are only a few points to represent the neuron soma. The lack of enough details for neuron soma makes the modeling on the soma morphology a challenging task. In this paper, we provide a general method to neuronal morphology modeling (including the soma and its connections to surrounding dendrites, and axons, with a focus on how different components are connected) and handle the challenging task when there are not many detailed sample points for soma.

Although statistical graph comprehension has been investigated in cognitive psychology, it has not been reported in cognitive neuroscience. The study designed an experimental condition, i.e., a statistical graph (SG), and two control conditions, i.e., a text (ST) and a statistical graph with text (SGT), where the ST is a verbal description of the information from the SG, and the SGT is a mixed graph + textual description. We used fMRI to analyze the brain activity of 36 normal subjects while they passively view the statistical information presented in any of SG, ST, and SGT. The results indicate that statistical graph comprehension requires the involvement of both ventral and dorsal streams, with more dependence on the ventral stream than the dorsal.

The high dimensionality of functional magnetic resonance imaging (fMRI) data presents major challenges to fMRI pattern classification. Directly applying standard classifiers often results in overfitting or singularity, which limits the generalizability of the results. In this paper, we propose a ”Doubly Regularized LOgistic Regression Algorithm” (DR LORA) which penalizes the voxels of the brain that are of no importance for the classification using the Alternating Direction Method of Multipliers (ADMM) and therefore alleviate this overfitting problem. Our algorithm was compared to other classification based algorithms such as Naive Bayes, Random forest and support vector machine. The results show clear performances for our algorithm.

Visual decision-making involving pairs of individuals tasked with determining the location of an object is a cognitive process combining independent systems together. Although it has been observed that combined systems can improve each of the individual systems, it remains a challenging problem to determine why and how this will occur. In this paper, we use Combinatorial Fusion Analysis (CFA) as a methodology through which we can effectively combine the decisions of two independent visual cognition systems. An experiment with 20 trials is performed in which participants are tasked with determining an object location, and stating the uncertainty factor for their decision. Our results demonstrate that the combination of two visual cognition systems using CFA can match or improve the performance of each individual system only if the pair of systems perform relatively well and are cognitively diverse.

Regarding to aversive stimuli, previous studies on emotion response and formation are plentiful, whereas concentrations on the emotional recovery are comparatively insufficient. The present study focused on the discomfort induced by looking at aversive pictures, and the emotional self-regulation during the following recovery period. A functional magnetic resonance imaging (fMRI) experiment with prolonged paradigm was recruited to investigate how brain-state shifted across three stages: picture viewing, earlier resting period, and latter resting period. Comparing with neutral pictures, aversive pictures activated the caudate nucleus centric subcortical areas, which also kept firing during the resting period. Meanwhile an activation pattern gradually appeared in fronto-parietal regions that were found negatively correlated to subcortical areas. Our findings suggest that the emotional recovery from discomfort is also a procedure accompanied by the strategy shift from passively suppressing emotional response to actively controlling the attention.

The paper presents a novel multimodal-based methodology for consciousness study of individuals with unresponsive wakefulness syndrome. Two interfaces were employed in the experiments: eye gaze tracking system – CyberEye developed at the Multimedia Systems Department, and EEG device with electrode placement in the international 10-20 standard. It was a pilot study for checking if it is possible to determine objective methods based on multimodal techniques which could replace or support current expensive and difficult to access neuroimaging techniques, like fMRI, PET, utilizing in evaluation of consciousness state. The multimodal-based methodology consists of several phases of research involving subjects. Hearing examination based on objective methods (OAE, ABR), consciousness test based on analysis of visual activity, examination of visual neural pathway with Steady State Visually Evoked Potentials and EEG-based comprehension test were proposed. The results obtained within conducted experiments and presented in this paper suggest that proposed objective-subjective methodology could potentially be introduced into clinical facilities after further validation.

Computer-based cognitive assessment programs for children have recently become increasingly popular. This assessment tool has many advantages over traditional assessment approaches including the option of offering an immediate feedback, the ability to systematize delivery of the test items and to modify the difficulty level and the ability to quantify progress.

Purpose:

the purpose of the study is to establish a reference baseline for the cognitive skills among Egyptian school-aged children.

Method

: This study is a cross-sectional prospective design. A sample of 223 healthy children of both sexes, of age ranged from 6-12 years, from urban areas’ elementary schools in Upper Egypt were recruited.

Results:

Rehacom program tool produced a separate progress report for the individual progress of every child.

Conclusions

: Based on the study‘s results the executive function ability was the first to initiated followed by the logical reasoning and finally the topological memory and vigilance

There are various kinds of methods on activated regions detection, including model-driven method and data-driven method, univariate method and multivariate method, frequency domain analysis and time-domain analysis etc. We investigated the problems of principal component analysis applied to activated regions detection,an autocovariance based principal component analysis method was proposed. Firstly,the time series were converted to the autocovariance series, and then the principal component analysis was employed. Meanwhile, the tactic of principal component selection was discussed. The validity of the proposed method was illustrated by experiments on a synthetic dataset and a real dataset. It was shown that the error rate of the new approach was lower compared with the principal component analysis itself.

Recently, the Multivariate Pattern Analysis (MVPA) studies for fMRI not only focus on cognitive state prediction, but also explore the interpretations of brain activity using model predictors (selected voxels). A model is considered to be good for interpreting brain activity if the selected voxels are all relevant to the specific cognitive state. Classical MVPA methods select voxels based on their prediction power; the selected ones are those that provide the best prediction performance. This precision based voxel selection method can guarantee the prediction performance, but it cannot ensure that all the selected ones are relevant. The interpretation of brain activity is therefore not ideal. This paper addresses this issue by introducing the concept of stability to the MVPA studies. If only the stability is emphasized in the selection process, the probability of selecting irrelevant voxels is highly reduced with the sacrifice of the prediction precision. We, therefore, propose a method to combine the stability assessment with the prediction precision assessment. In this paper, the proposed voxel selection method is integrated into a linear sparse predictor, Random Subspace Sparse Bayesian Learning (RS-SBL). The experiment results of simulation datasets demonstrate that our method can simultaneously reduce false positive and false negative rates while maintaining the prediction performance.

This paper describes improvement of the Hilbert-Huang transform (HHT) for detection of ERP components in the EEG signal. Time-frequency domain methods, such as the wavelet transform or matching pursuit, are commonly for this task. We used a modified Hilbert-Huang transform that allows the processing of quasi-stationary signals such as EEG. The essential part of the HHT is an Empirical Mode Decomposition (EMD) that decomposes signal into intrinsic mode functions (IMFs). We designed additional stopping criteria for better selection of IMFs in the EMD. These IMFs positively affect later computed instantaneous attributes and increase classification success. We tested the influence of additional stopping criteria on classification reliability using the real EEG data acquired in our laboratory. Our results demonstrated that we were able to detect the P3 component by using the HHT with additional stopping criteria more successfully than by using the original implementation of modified HHT, continuous wavelet transform and matching pursuit.

Spike sorting is often required for analyzing neural recordings to isolate the activity of single neurons. In this paper, a new feature extractor based on Wavelet and kernel PCA for spike sorting was proposed. Electrophysiology recordings were made in Sprague-Dawley (SD) rats to provide neural signals. Here, an adaptive threshold based on the duty-cycle keeping method was used to detect spike and a new spike alignment technique was used to decrease sampling skew error. After spikes were detected and alimented, to extract spike features, their wavelet transform was calculated, the first 10 coefficients with the largest deviation from normality provided a compressed representation of the spike features that serves as the input to KPCA algorithm. Once the features have been extracted, k-means clustering was utilised to separate the features and differentiate the spikes. Test results with simulated data files and data obtained from SD rats in vivo showed an excellent classification result, indicating the good performance of the described algorithm approach.

This paper presents a method to remove ocular artifacts from electroencephalograms (EEGs) which can be used in biomedical analysis in portable environment. An important problem in EEG analysis is how to remove the ocular artifacts which wreak havoc among analyzing EEG signals. In this paper, we propose a combination of Wavelet Transform with effective threshold and adaptive filter which can extract the reference signal according to ocular artifacts distributing in low frequency domain mostly, and adaptive filter based on Least Mean Square (LMS) algorithm is used to remove ocular artifacts from recorded EEG signals. The results show that this method can remove ocular artifacts and superior to a comparison method on retaining uncontaminated EEG signal. This method is applicable to the portable environment, especially when only one channel EEG are recorded.

Colorectal cancer (CRC) is a relatively common cause of death around the globe. Predictive models for the development of CRC could be highly valuable and could facilitate an early diagnosis and increased survival rates. Currently available predictive models are improving, but do not fully utilize the wealth of data available about patients in routine care nor do they take advantage of the developments in the area of data mining. In this paper, a first attempt to generate a predictive model using the CHAID decision tree learner based on anonymously extracted Electronic Medical Records is reported, showing an area under the curve (AUC) of .839 for the adult population and .702 for the age group between 55 and 75.

Electronic health records (EHRs) are becoming an increasingly important source of patient information. Unfortunately, EHR data do not always directly and reliably map to medical concepts that clinical researchers need or use. Some recent studies have focused on EHR-derived phenotyping, which aims at mapping the EHR data to specific medical concepts; however, most of these approaches require labor intensive supervision from experienced clinical professionals.

In this paper, we use Limestone, a nonnegative tensor factorization method to derive phenotype candidates from claims data with virtually no human supervision. Limestone represents the interactions between diagnoses and procedures among patients naturally using tensors (a generalization of matrices). The resulting tensor factors are reported as phenotype candidates that automatically reveal patient clusters on specific diagnoses and procedures. To the best of our knowledge, this is the first study that successfully extracts useful phenotypes by applying sparse nonnegative tensor factorization to a large, public-domain EHR dataset covering a broad range of diseases. Our experiments demonstrate the interpretability and the promise of high-throughput phenotypes generated from tensor factorization.

The paper aims at two tasks of electronic medical record (EMR) processing: EMR retrieval and medical term extraction. The linguistic phenomena in EMRs in different departments are analyzed in depth including record size, vocabulary, entropy of medical languages, grammaticality, and so on. We explore various techniques of information retrieval for EMR retrieval, including five retrieval models with six pre-processing strategies on different parts of EMRs. The learning to rank algorithm is also adopted to improve the retrieval performance. Finally, our retrieval model is applied to extract medical terms from EMRs. Both coarse-grained relevance evaluation on department level and fine-grained relevance evaluation on treatment level are conducted.

Recently there has been a growing attention on the use of web and social data to improve traditional prediction models in politics, finance, marketing and health, but even though a correlation between observed phenomena and related social data has been demonstrated in many cases, yet the effectiveness of the latter for long-term or even mid-term predictions has not been shown. In epidemiological surveillance, the problem is compounded by the fact that infectious diseases models (such as susceptible-infected-recovered-susceptible, SIRS) are very sensitive to current conditions, such that small changes can produce remarkable differences in future outcomes. Unfortunately, current or nearly-current conditions keep changing as data are collected and updated by the epidemiological surveillance organizations. In this paper we show that the time series of Twitter messages reporting a combination of symptoms that match the influenza-like-illness (ILI) case definition represent a more stable and reliable information on “current conditions”, to the point that they can replace, rather than simply integrate, official epidemiological data. We estimate the effectiveness of these data at predicting current and past flu seasons (17 seasons overall), in combination with official historical data on past seasons, obtaining an average correlation of 0.85 over a period of 17 weeks covering the flu season.

Hospital charges are determined by numerous factors. Even the cost for the same procedure can vary greatly depending on a patient’s conditions, complications, and types of facilities. With the advent of Obamacare, estimating hospital charges has become an increasingly important problem in healthcare informatics. We propose a hierarchical ensemble of

α

-Trees to delicately deal with this challenging problem. In the proposed approach, multiple

α

-Trees are built to capture the different aspects of hospital charges, and then these multiple classifiers are uniquely combined for each hospital. Hospitals are characterized by unique weight vectors that explain the subtle differences in hospital specialties and patient groups. Experimental results based on the 2006 Texas inpatient discharge data show that our approach effectively captures the variability of hospital charges across different hospitals, and also provides a useful characterization of different hospitals in the process.

Despite of the spread of Electronic Health Records (EHRs) in Spanish hospitals and Spanish occupying the second place in the ranking of number of speakers, to the best of our knowledge there are no natural language processing tools for medical texts written in Spanish.

This paper presents an approach based on OpenNLP to process natural language texts written in Spanish for information extraction. The main goal is to integrate our development with cTAKES. As cTAKES has been specifically trained for the clinical domain, in this paper we will train the main modules from a general purpose annotated Spanish corpus and an in-house corpus developed with medical documents, testing both on a set of medical documents. Best performance of individual components when tested with medical documents: Sentence boundary detector accuracy = 0.872; Part-of-speech tagger accuracy = 0.946; chunker = 0.909.

The documents selection related brain information based on the data-brain ontology not only has an important significance in the promotion of data-brain ontology, but also lays the foundation for knowledge integration. However, traditional research of documents selection focuses on the concept, and cannot meet the requirement of the systematic Brain Informatics study. This paper analyzes the characteristics of source knowledge firstly with concepts, attributes and relations. Then, we calculate the weight of documents by using the improved method of VSM. Finally, the experiments using real documents associated with brain science are given and calculating the weight of each document achieves a better effect of ranking selection.

Finding and learning related research is a necessary work in Brain Informatics studies. However, the keyword-based search on brain and mental big data center often brings a large amount of unnecessary results. It is very difficult to find needed research from those results for researchers. This paper proposes a Brain Informatics research recommendation system based on the Data-Brain and BI provenances. By choosing interest aspects from the Data-Brain and applying the unification of search and reasoning based on Data-Brain interests, the more accurate search can be realized to find really related literatures for supporting systematic Brain Informatics studies.

Lots of heterogeneous complex data are collected for diagnosis purposes. Such data should be shared between all caregivers and, often at least partly automatically processed, due to its complexity, for its full potential to be harnessed. This paper is a feasibility study that assesses the potential of Hadoop as a medical data storage and processing platform using EEGs as example of medical data.

In the continuing goal to merge the fields of computational neuroscience with medical based neurodiagnostic clinical research this paper presents advancements on machine learning Big Electroencephalogram (EEG) Data. The authors’ clinical decision-support systems (CDSS) presented in previous work was able to distinguish, within minutes, pathological oscillations hidden in terabytes of complex signal analysis. This paper presents training and learning elements that compliment and advance this previous work. This paper shows how perceptrons, that predate modern-day neural network constructs, remain relevant in many modern classification applications where a clear linear separation is present in the data. Furthermore, the perceptrons also compliment the domain adaptation covariant shifts later used when the system is used in the neuroICU (Intensive Care Unit). Accordingly, we present supervised learning for the neuroICU using single-layer perceptron classifiers.

Efficient encoding of Roman rules is based on the neural bases of mathematical cognitive abilities. The present imaging studies have shown that information granule representing a form of Roman rules is associated with arithmetical domain-sensitive parietal cortex, indicating a switch from the data to the information granule retrieval of memory rules. So far, however, little is known about the developing neural substrate for the establishment of rules from data to information granule. The aim of the present fMRI study is to investigate whether and how mathematical intelligence might be enhanced from data to information granule of Roman arithmetic rules in the parietal cortex. Concerning the same rules, the paired t-test analysis indicated that different activation in the bilateral parietal lobule associated with different retrieval levels. In conclusion, the present study yielded some evidence that a successful model for knowledge-building of rules is accompanied by modifications of brain activation patterns.

Semantic networks inspired by semantic information processing by the brain frequently do not improve the results of text classification. This counterintuitive fact is explained here by the multiple inheritance problem, which corrupts real-world knowledge representation attempts. After a review of early work on the use of semantic networks in text classification, our own heuristic solution to the problem is presented. Significance testing is used to contrast results obtained with pruned and entire semantic networks applied to medical text classification problems. The algorithm has been motivated by the process of spreading neural activation in the brain. The semantic network activation is propagated throughout the network until no more changes to the text representation are detected. Solving the multiple inheritance problem for the purpose of text classification is similar to embedding inhibition in the spreading activation process – a crucial mechanism for a healthy brain.

PubMed is a search engine used to access the MEDLINE database, which comprises the massive amounts of biomedical literature. This an make more difficult for accessing to find the relevant medical literature. Therefore, this problem has been challenging in this work. We present a solution to retrieve the most relevant biomedical literature relating to Cholangiocarcinoma in clinical trials from PubMed. The proposed methodology is called ontology-based text classification (On-TC). We provide an ontology used as a semantic tool. It is called

C

an

c

er

T

echnical

T

erm

N

et (CCT-Net). This ontology is intergrated to the methodology to support automatic semantic interpretation during text processing, especially in the case of synonyms or term variations.

Traditional medicines can play a major role in global health care, due to its indigenous nature, easy access, and cost effectiveness. However, knowledge of this intellectual property is in danger of being lost. It is either undocumented or if documented, it is inaccessible and local in context. World Health Organization signifies the necessity to preserve and maintain this knowledge. Unani medicines, a subfield of traditional medicines, have been continuously practiced in Asia for about 2500 years, and it is facing the same situation of knowledge lost. To preserve knowledge of Unani medicines, initial kind of an effort has been done but a formal semantic structure, that is machine readable and reusable, is required to preserve this knowledge efficiently and effectively. This research focuses on conceptual structure of Unani medicines by presenting domain ontology which includes core principles and philosophy of Unani medicines, diseases, symptoms, diagnosis, drugs, and treatment. Knowledge about fundamentals is captured from expert interviews and books and then this knowledge is converted into ontologies using Protégé. Although it is not exhaustive domain ontology, however it may serve as a starting point for any knowledge based application of Unani medicines. In this research a semantic queries based case study along with a prototype expert system is also proposed.

We evaluated the efficiency of robotic therapy for stroke survivors by using a computational approach in motor theory with a stroke rehabilitation model. In computational neuroscience, hand movement can be represented by population coding of neuronal preferred directions (PDs) in the motor cortex. We modeled the recovery processes of arm movement in conventional and robotic therapies as reoptimization of PDs in different learning rules, and compared the efficiencies after stroke. Conventional therapy did not induce complete recovery of stroke lesions, and the neuronal state depended on the training direction. However, robotic therapy reoptimized the PDs uniformly regardless of the training direction. These observations suggest that robotic therapy may be effective for recovery and not have a negative effect on motor performance depending the training direction. Furthermore, this study provides computational evidence to promote robotic therapy for stroke rehabilitation.

In this paper, we propose a new health monitoring system (HMS) based on a new classification method consisting of the real time support vector machines (RTSVM). The new HMS denoted by RTSVM-MS deals with problems of monitoring systems in intensive care unit (ICU). The main aim of this new system is to considerably reduce the rate of false alarms and keep a high and stable level of sensitivity. Besides, it overcomes the main issue of the existing HMS by proposing a classification model that considers the variation of the patient states over time. In addition, the thresholds set has to be modified when patients are getting better. However, thresholds are stable and do not translate the states of patients over time since, all existing systems in ICU do not take into account of the patients’ states evolution. Our proposal has the ability to generate an initial model that classifies states of patients to normal and abnormal (critical) using the LASVM. Then, it updates its model by considering the evolution in the states of patients using RTSVM. As a result, the new system gives what the medical staff wants as information and alarms relative to monitored patient.

Dementia is the loss of cognitive brain functioning including thinking, remembering, and reasoning to such an extent that it interferes with a person’s daily life and activities. Individuals with high levels of intelligence, educational, and occupational attainment may sustain greater brain damage before demonstrating functional deficit. This is mainly contributed to their cognitive reserve, the mind’s resistance to damage of the brain. This paper proposes brain aerobics, a set of stimulating mental exercises that aims at building the cognitive reserve to protect elderly people from developing dementia’s symptoms. Brain aerobics exercises are developed for mobile devices for ease of use. Experimental results demonstrate the benefits of the proposed brain aerobics in increasing IQ levels by 7.2% for elderly people who have dementia high risk factors.

In neurosurgical treatment of the Parkinson Disease (

PD

) the target is a small (9 x 7 x 4 mm) deep within brain placed structure called

Subthalamic

Nucleus

(

STN

). The goal of the Deep Brain Stimulation (

DBS

) surgery is the permanent precise placement of the stimulating electrode within target nucleus. As this structure poorly discriminates in CT or MRI it is usually stereotactically located using microelectrode recording. Several microelectrodes are parallelly inserted into the brain and in measured steps they are advanced towards expected location of the nucleus. At each step, from 20 mm above the target, the neuronal activity is recorded. Because

STN

has a distinct physiology, the signals recorded within it also present specific features. By extracting certain features from recordings provided by the microelectrodes, it is possible to construct a classifier that provides useful discrimination. This discrimination divides the recordings into two classes, i.e. those registered within the

STN

and those registered outside of it. Using the decision tree based classifiers, the best results have been obtained using the Random Forest method. In this paper we compared the results obtained from the Random Forest to those provided by the classification based upon rules extracted by the rough set approach.

The most common method used by neurologist to evalute the Parkinson’s Disease patients are different rating scales. They give overall picture of the PD patient, but are not objective and different experts can make different observations. In this article the results of correlation between UPDRS evaluation and different quality measures calculated based on MOCAP data recorded during walking for group of PD patients with implanted DBS stimulator have beed presented. This is a continuation of our previous research related to analysis of gait in Parkinson’s Disease Patients.

Deep brain stimulation (DBS) is a well established method used as treatment in patients with advanced Parkinson’s disease (PD). Our main purpose is to increase precision of DBS method by determining which parts of cortex are stimulated in different set-ups. In this paper we have analyzed MRIs that are performed as a standard procedure before and after the DBS surgery. We have used 3D Slicer for registration of MRIs with anatomical brain atlas. In addition, we have generated trajectories of neural tracts (tractography) connecting STN with cortex using data colected by DTI (Diffusion Tensor Imaging). In the following step we have used Rougt Set Theory to compare MRI data with neurological findings acquired by neurologists. We have tested prediction of DBS electrode contact’s position and stimulating parameters in individual patients on improvements of particular neurological symptoms. Our results may give a basis to set optimal parameters of stimulation and electrode’s position in order to obtain the most effective PD treatment.

One of the most common variants of mood disorders is depression. According to various authors, the incidence of depression in the population is 3-10 %, in Parkinson’s disease (PD) – 40-50 % of patients. Most researchers are considering depression in PD as endogenous and finds it to be an important and independent component of the disease manifestations. We examined 49 patients with PD complicated by depression. All patients underwent MRI followed by postprocessing using FreeSurfer (http://surfer.nmr.mgh.harvard.edu). When depression occurs it affects lingual area, parahippocampal areas on both sides and straight gyrus. Regression analysis showed a predominant involvement of the frontal and temporal brain lobes. Prognostically three most important areas involved in the formation of depression were revealed - right and left parahippocampal area and the average occipital-temporal sulcus. The risk of depression manifestation, against the background of left parahippocampal cortex thinning at rates below 2,597 mm, increases 46.8 times.

The adoption of hospital EHR technology is significantly growing and expected to grow. Digitalized information is the basis for health analytics. In particular, patient medical records contain valuable clinical information written in narrative form that can only be extracted after it has been previously preprocessed with Natural Language Processing techniques. An important challenge in clinical narrative text is that concepts commonly appear negated. Though worldwide there are nearly 500 million Spanish speakers, there seems to be no algorithm for negation detection in medical texts written in that language.

Thus this paper presents an approach to adapt the NegEx algorithm to be applied to detect negation regarding clinical conditions in Spanish written medical documents. Our algorithm has been trained with 500 texts where 422 different sentences and 267 unique clinical conditions were identified. It has been tested for negated terms showing an accuracy obtained is of 83,37%. As in the detection of definite affirmed conditions, the results show an accuracy of 84,78%.

Predicting the evolution of individuals is a rather new mining task with applications in medicine. Medical researchers are interested in the progress of a disease and in the evolution of individuals subjected to treatment. We investigate the evolution of patients on the basis of medical tests before and during treatment after brain trauma: we want to understand how similar patients

can

become to healthy participants. We face two challenges. First, we have less information on healthy participants than on the patients. Second, the values of the medical tests for patients, even after treatment started, remain well-separated from those of healthy people; this is typical for neurodegenerative diseases, but also for further brain impairments. Our approach encompasses methods for modelling patient evolution and for predicting the health improvement of different patient subpopulations, dealing with the above challenges. We test our approach on a cohort of patients treated after brain trauma and a corresponding cohort of controls.

Chronic diseases such as diabetes and hypertension have been recognized in the last decade among the principal causes of death in the world. Mitigating and controlling the elicited risks necessitate a continuous monitoring to produce accurate recommendations for both patients and physicians. For patient, it will help in adjusting his/her lifestyles, medications, and sport activities. However, for physicians, it helps in taking guided therapy decision. In this paper, we propose an adaptive Expert System (ES) that relies, not only on a set of rules validated by experts, but also linked to an intelligent continuous monitoring scheme that copes with semi-continuous data streams by implementing smart sensing and pre-processing of data. In addition, we implemented an iterative data analytic technique that learns from the past ES experience to continuously improve clinical decision-making and automatically generates validated advices. These advices are visualized via an application interface. We experimented the proposed system using different scenarios of monitoring blood sugar and blood pressure parameters of a population of patients with chronic diseases. The results we have obtained showed that our ES combined with the intelligent monitoring and analytic techniques provide a high accuracy of collected data and evident-based advices.

The paper presents an innovative method for the diagnosis of the arterio-venous fistula based on recorded acoustic signals. A fistula is an artificial connection between an artery and a vein made to obtain a suitably large blood flow for haemodialysis. If the fistula does not work properly, thrombosis or other health- or life-threatening conditions may develop. Based on the analysis of sound generated by blood flowing through the fistula, the occurrence of pathological conditions may be diagnosed. An artificial neural network implemented using an FANN (Fast Artificial Neural Network) library has been used to evaluate the fistula condition.

This paper studies the performance of recently introduced asymptotic statistics for connectivity inference in the frequency domain, namely via information partial directed coherence (

i

PDC) and information directed transfer function (

i

DTF) and compares them to the behaviour of a classic time domain multivariate Granger causality test (GCT) by using Monte Carlo simulations of three widely used toy-models under varying the simulated data record lengths. In general, the false-positive rates for non-existing connections and the false-negative rates for existing connections are found to decrease with longer record lengths.

We employ toy models to re-examine the notion of causality and its implications in unravelling networks in neuroscience. We conclude that even though multivariate representations of neural dynamic data is indispensable, current popular terminologies for addressing connectivity are insufficiently precise and may even be misleading for fully describing the breadth of information multivariate models now provide. This imposes the need to consider a brand new link centered paradigm of network description where the directed nature of the links plays a central role.

Resting-state Functional Magnetic Resonance Imaging (FMRI) analysis has consistently shown the presence of specific spatial activation patterns. Independent component analysis (ICA) has been the analysis algorithm of choice even though its underlying assumptions preclude deeper connectivity analysis. By combining novel concepts of group sparsity with contiguity-constrained clusterization, we developed a new class of Local dimension-reduced Dynamical Spatio-Temporal Models (LDSTM) for estimating whole-brain dynamical models whereby the causal relationships between well localized spatial components can be identified. Experimental results of LDSTM on group resting-state FMRI data reveal physiologically plausible spatio-temporal brain connectivity patterns among participants.

Localization of sources of brain electrical activity from electroencephalographic and magnetoencephalographic recordings is an ill-posed inverse problem. Therefore, the best one can hope for is to derive a source localization method which is guaranteed to find sources belonging to the set of possible solutions to this problem. Recently, a few methods with this property have been proposed as a non-trivial generalizations of the classical neural activity index based on the linearly constrained minimum-variance (LCMV) spatial filtering technique. In this paper we propose a family of reduced-rank activity indices achieving maximum value when evaluated at true source locations for uncorrelated dipole sources and any nonzero rank constraint. This fact shows in particular that this key property is not confined to a selected few activity indices. We present a series of numerical simulations evaluating localization performance of the proposed activity indices. We also give an overview of areas of future research which should be considered as an extension of the results of this paper. In particular, we discuss how new families of activity indices can be derived based on the proposed technique.

How information processes in the human brain relate to action formation is an interesting research question and with the latest development of brain imaging and recording techniques more and more interesting insights have been uncovered. In this paper a cognitive model is scrutinized which is based on cognitive, affective, and behavioural science evidences for situation awareness. Situation awareness has been recognized as an important phenomenon in almost all domains where safety is of highest importance and complex decision making is inevitable. This paper discusses analysis, modelling and simulation of three scenarios in the aviation domain where poor situation awareness plays a main role, and which have been explained by Endsley according to her three level situation awareness model. The computational model presented in this paper is driven by the interplay between bottom-up and top-down processes in action formation together with processes and states such as: perception, attention, intention, desires, feeling, action preparation, ownership, and communication. This type of cognitively and neurologically inspired computational models provide new directions for the artificial intelligence community to develop systems that are more aligning with realistic human mental processes and for designers of interfaces of complex systems.

Neuroimaging studies of rare disorders, such as dyslexia, require long term, multi-centre data collection in order to create representative disease specific cohorts. However, multi-site data have inherent heterogeneity caused by site specific acquisition protocols, scanner setup, etc. The aim of this study was the analysis of the influence of the two confounding factors: site location and field strength on feature selection procedure. We propose two methods: site-dependent whitening and site-dependent extension and compare with naive approach using classification accuracy as a quality measure of selected features subset. The proposed methods outperform the naive approach, and significantly improves the classification performance of developmental dyslexia.

Human brain imaging techniques, such as Magnetic Resonance Imaging (MRI) or Diffusion Tensor Imaging (DTI), have been established as scientific and diagnostic tools and their adoption is growing in popularity. Statistical methods, machine learning and data mining algorithms have successfully been adopted to extract predictive and descriptive models from neuroimage data. However, the knowledge discovery process typically requires also the adoption of pre-processing, post-processing and visualisation techniques in complex data workflows. Currently, a main problem for the integrated preprocessing and mining of MRI data is the lack of comprehensive platforms able to avoid the manual invocation of preprocessing and mining tools, that yields to an error-prone and inefficient process. In this work we present K-Surfer, a novel plug-in of the Konstanz Information Miner (KNIME) workbench, that automatizes the preprocessing of brain images and leverages the mining capabilities of KNIME in an integrated way. K-Surfer supports the importing, filtering, merging and pre-processing of neuroimage data from FreeSurfer, a tool for human brain MRI feature extraction and interpretation. K-Surfer automatizes the steps for importing FreeSurfer data, reducing time costs, eliminating human errors and enabling the design of complex analytics workflow for neuroimage data by leveraging the rich functionalities available in the KNIME workbench.

A kind of metacognitive and cognitive activity patterns in the brain have been found in the task state, showing the gradient distribution from abstract to concrete processing in the frontal and parietal cortex especially. In our early study, it is observed that this kind of gradient organization is intrinsic and prepared in the resting state. Learning experience is a process from metacognitive to cognitive processing, which might change the spontaneous activity patterns in the resting brain. This study is to explore how the learning experience, including both long-term practice and short-term task experience, influences the intrinsic gradient organization in the human brain. Focused on the task-evoked metacognitive and cognitive pattern regions, by comparing four resting state data, before and after task performing in Day 1 before practice named pre-pre and pre-post respectively, and before and after task in Day 7 after 5-day’s practice, named post-pre and post-post respectively, we investigated the change of gradient organization in the human brain with the approach of functional connectivity (FC) analysis. The result showed that the gradient organization is quite stable across the four resting states, which is similar with our previous finding. Task performance enhanced the correlation between cognitive and mixed functional network, especially after long-time practice, suggesting the key role of cognitive network in the task execution. Moreover, after long practice, the internal connectivity within the metacognitive network and the connection between mixed and cognitive functional network were both weakened, which suggested a functional modulation and separation when task performance became more and more skilled and automatic.

Biomedical experts are confronted with ”Big data”, driven by the trend towards precision medicine. Despite the fact that humans are excellent at pattern recognition in dimensions of ≤ 3, most biomedical data is in dimensions much higher than 3, making manual analysis often impossible. Experts in daily routine are decreasingly capable of dealing with such data. Efficient, useable and useful computational methods, algorithms and tools to interactively gain insight into such data are a commandment of the time. A synergistic combination of methodologies of two areas may be of great help here: Human–Computer Interaction (HCI) and Knowledge Discovery/Data Mining (KDD), with the goal of supporting human intelligence with machine learning. Mapping higher dimensional data into lower dimensions is a major task in HCI, and a concerted effort including recent advances from graph-theory and algebraic topology may contribute to finding solutions. Moreover, much biomedical data is sparse, noisy and time-dependent, hence entropy is also amongst promising topics. This tutorial gives an overview of the HCI-KDD approach and focuses on 3 topics: graphs, topology and entropy. The goal of this intro tutorial is to motivate and stimulate further research.

High-throughput genomic profiling technology provides us detailed information of biological systems. However, it also increases the dimensionality in data, which makes it harder to identify key features and their relations to other features hidden in feature spaces. In this paper we propose a new idea based on the structure learning for the Gaussian Markov random field, which provides us an efficient way to represent a feature space as a collection of small graphs, where nodes represent features and edges represent conditional dependency between features. In our approach a collection of small graphs is created for each subgroup of a cohort, where our interest lies in finding characteristic patterns in each subgroup graph compared to the other subgroup graphs. A simple but effective method is proposed using polarized adjacency matrices to find topological differences in collections of graphs.

Cancers follow a clonal Darwinian evolution, with fitter subclones replacing more quiescent cells, ultimately giving rise to macroscopic disease. High-throughput genomics provides the opportunity to investigate these processes and determine specific genetic alterations driving disease progression. Genomic sampling of a patient’s cancer provides a molecular history, represented by a phylogenetic tree. Cohorts of patients represent a forest of related phylogenetic structures. To extract clinically relevant information, one must represent and statistically compare these collections of trees. We propose a framework based on an application of the work by Billera, Holmes and Vogtmann on phylogenetic tree spaces to the case of unrooted trees of intra-individual cancer tissue samples. We observe that these tree spaces are globally nonpositively curved, allowing for statistical inference on populations of patient histories. A projective tree space is introduced, permitting visualizations of evolutionary patterns. Published data from four types of human malignancies are explored within our framework.

Pathogenic bacteria present a large disease burden on human health. Control of these pathogens is hampered by rampant lateral gene transfer, whereby pathogenic strains may acquire genes conferring resistance to common antibiotics. Here we introduce tools from topological data analysis to characterize the frequency and scale of lateral gene transfer in bacteria, focusing on a set of pathogens of significant public health relevance. As a case study, we examine the spread of antibiotic resistance in

Staphylococcus aureus

. Finally, we consider the possible role of the human microbiome as a reservoir for antibiotic resistance genes.

Hot topics in knowledge discovery and interactive data mining from natural images include the application of topological methods and machine learning algorithms. For any such approach one needs at first a relevant and robust digital content representation from the image data. However, traditional pixel-based image analysis techniques do not effectively extract, hence represent the content. A very promising approach is to extract graphs from images, which is not an easy task. In this paper we present a novel approach for knowledge discovery by extracting graph structures from natural image data. For this purpose, we created a framework built upon modern Web technologies, utilizing HTML canvas and pure Javascript inside a Web-browser, which is a very promising engineering approach. Following on a short description of some popular image classification and segmentation methodologies, we outline a specific data processing pipeline suitable for carrying out future scientific research. A demonstration of our implementation, compared to the results of a traditional watershed transformation performed in Matlab showed very promising results in both quality and runtime, despite some open problems. Finally, we provide a short discussion of a few open problems and outline some of our future research routes.

This conceptual paper discusses a graph-based approach for on-line terrain coverage, which has many important research aspects and a wide range of application possibilities, e.g in multi-agents. Such approaches can be used in different application domains, e.g. in medical image analysis. In this paper we discuss how the graphs are being generated and analyzed. In particular, the analysis is important for improving the estimation of the parameter set for the used heuristic in the field of route planning. Moreover, we describe some methods from quantitative graph theory and outline a few potential research routes.

Heart rate variability (HRV) is the variation of the time interval between consecutive heartbeats and depends on the extrinsic regulation of the heart rate. It can be quantified using nonlinear methods such as entropy measures, which determine the irregularity of the time intervals.

In this work, approximate entropy (ApEn), sample entropy (SampEn), fuzzy entropy (FuzzyEn) and fuzzy measure entropy (FuzzyMEn) were used to assess the effects of three different cardiac arrhythmia suppressing drugs on the HRV after a myocardial infarction.

The results show that the ability of all four entropy measures to distinguish between pre- and post-treatment HRV data is highly significant (

p

 < 0.01). Furthermore, approximate entropy and sample entropy are able to differentiate significantly (

p

 < 0.05) between the tested arrhythmia suppressing agents.

We propose a graph-based analysis framework to study the dynamics of visual gaze from web users. Our goal is to extract the main characteristics of the information foraging process from an attention-centric perspective. Our approach consists of modeling web objects, such as images and paragraphs, as nodes. The visual transitions are represented as edges. With the resulting graphs, several standard metrics were computed. We performed an initial empirical study with 23 subjects. The visual activity was captured using an eye tracking device. The results suggest that a graph based analysis can capture in a reliable way the dynamics of user behavior and the identification of salient objects within a web site.