International Journal on Artificial Intelligence Tools

The proper management of patient data and their accessibility are still remaining issues that prevent the full deployment and usage of pervasive healthcare applications. This paper presents an integrated health monitoring system based on mobile pervasive technologies. The system utilizes Cloud Computing for providing robust and scalable resources for sensor data acquisition, management and communication with external applications like health information systems. A prototype has been developed using both mobile and wearable sensors for demonstrating the usability of the proposed platform. Initial results regarding the performance of the system, the efficiency in data management and user acceptability have been quite promising.

A method of adapting the boundaries when extracting the spectral features from heart rate variability (HRV) for sleep and wake classification is described. HRV series can be derived from electrocardiogram (ECG) signals obtained from single-night polysomnography (PSG) recordings. Conventionally, the HRV spectral features are extracted from the spectrum of an HRV series with fixed boundaries specifying bands of very low frequency (VLF), low frequency (LF), and high frequency (HF). However, because they are fixed, they may fail to accurately reflect certain aspects of autonomic nervous activity which in turn may limit their discriminative power, e.g. in sleep and wake classification. This is in part related to the fact that the sympathetic tone (partially reflected in the LF band) and the respiratory activity (modulated in the HF band) vary over time. In order to minimize the impact of these variations, we adapt the HRV spectral boundaries using time-frequency analysis. Experiments were conducted on a data set acquired from two groups with 15 healthy and 15 insomnia subjects each. Results show that adapting the HRV spectral features significantly increased their discriminative power when classifying sleep and wake. Additionally, this method also provided a significant improvement of the overall classification performance when used in combination with other HRV non-spectral features. Furthermore, compared with the use of actigraphy, the classification performed better when combining it with the HRV features.

Nowadays medical software is tightly coupled with medical devices that perform patient state monitoring and lately even some basic treatment procedures. Medical guidelines (GLs) can be seen as specification of a medical system which requires their computer-interpretable representation of medical GLs. Until now most of the medical GLs are often represented in a textual format and therefore often suffer from such structural problems as incompleteness, inconsistencies, ambiguity and redundancy, which makes the translation process to the machine-interpretable language more complicated. Computer-based interpretation of GLs can improve the quality of protocols as well as the quality of medical service. Several GLs formal representation methods have been presented recently. Only some of them enable automatic formal verification by introducing an additional translation path to the existing model checking environments. However, if a verified property fails it is difficult to trace back the result needed to change the model. Moreover, these formalisms provide the notion of time mostly in terms of actions order. In this paper we preset the application of a well-know formal behaviour representation approach of embedded systems design domain to medical GLs interpretation. We use Timed Automata extended with Tasks (TAT) and TIMES toolbox to represent medical GLs as a system behaviour in a computer interpretable form. We discuss the verification issues with the help of the anticancer drug imatinib case study.

In this paper, we propose sparse coding-based approaches for segmentation of tumor regions from magnetic resonance (MR) images. Sparse coding with data-adapted dictionaries has been successfully employed in several image recovery and vision problems. The proposed approaches obtain sparse codes for each pixel in brain MR images considering their intensity values and location information. Since it is trivial to obtain pixel-wise sparse codes, and combining multiple features in the sparse coding setup is not straight-forward, we propose to perform sparse coding in a high-dimensional feature space where non-linear similarities can be effectively modeled. We use the training data from expert-segmented images to obtain kernel dictionaries with the kernel K-lines clustering procedure. For a test image, sparse codes are computed with these kernel dictionaries, and they are used to identify the tumor regions. This approach is completely automated, and does not require user intervention to initialize the tumor regions in a test image. Furthermore, a low complexity segmentation approach based on kernel sparse codes, which allows the user to initialize the tumor region, is also presented. Results obtained with both the proposed approaches are validated against manual segmentation by an expert radiologist, and it is shown that proposed methods lead to accurate tumor identification.

Vascular tissue characterization is of great importance concerning the possibility of an Acute Cardiac Syndrome (ACS). Gray-scale intravascular ultrasound (IVUS) is a powerful tomographic modality providing a thorough visualization of coronary arteries. Among the existing methods, virtual histology (VH) is the most popular and clinically available technique for plaque component analysis, it suffers however from a poor longitudinal resolution. In order to surmount this demerit, a new image-based methodology for plaque assessment is suggested here that differentiates tissue components into four classes: calcium, necrotic core, fibrous and fibro-lipid. A rich set of five textural feature families are extracted from IVUS images, computed at different scales. The main contribution of this paper is that tissue classification is accomplished using the principles of multiple classifiers combination approach. At the first stage, an ensemble of base SVM classifiers is constructed from each feature family, separately. The fuzzy outputs of the individual classifiers are then aggregated to provide the final fused results. We investigate four efficient decision fusion schemes of the literature and the SVM fuser. Extensive experimentation is carried out to highlight the merits of the suggested schemes against single SVM classifiers that use reduced feature subsets obtained after feature selection or the entire feature space. The analysis demonstrates that the decision fusion techniques offer improved classification accuracies, compared to single SVM classifiers and existing methods in IVUS imaging. In addition, the method provides accurate assessments of plaque composition in IVUS images.

Autoimmune diseases are strictly connected with the presence of autoantibodies in patient serum. Detection of Antinucleolar Antibodies (ANAs) in patient serum is performed using a laboratory technique named Indirect Immunofluorescence (IIF) followed by manual evaluation on the acquired slides from specialized personnel. In this procedure, several limitations appear and several automatic techniques have been proposed for the task of ANA detection. In this work we present a method achieving state-of-the-art performance on a publicly available dataset. More precisely, two powerful and rotation invariant descriptors are incorporated into a two stage classification scheme where the feature vectors are represented and fused in the dissimilarity space. Then, in a second level dissimilarity vectors are classified using a linear SVM classifier. Evaluation on the HEp-2 cell contest dataset yields a 70.16% performance on cell-level classification. Furthermore we provide results in Image Level Classification where a 78.57% classification rate was achieved.

Most proteins express their functions by binding with other proteins or molecular compounds (ligands). Since the characteristics of the local portion involved in binding (binding site) often determine the function of the protein, clarifying the location of the binding site of the protein helps analyze the function of proteins. Binding sites that bind to similar ligands often have common surface structures (surface motifs). Extracting the surface motifs among several proteins with similar functions improves binding site prediction.

We propose a method that predicts binding sites by extracting the surface motifs that are frequently observed in only a specific set of proteins that bind to the same ligand (group). Since most binding sites have concave structures (pockets), the pockets are compared and common structures are searched for to extract the surface motifs by applying similar graph mining to the pocket data, which are represented as graphs. Common binding sites across several groups can be predicted in such a way to integrate more than one group. We also proposed a method of protein classification, in which the surface motifs extracted using the above method are evaluated on the assumption that a protein belongs to each one of the groups.

Developing effective artificial intelligence tools to find motifs in DNA, RNA and proteins poses a challenging yet important problem in life science research. In this paper, we present a computational approach for finding RNA tertiary motifs in genomic sequences. Specifically, we predict genomic coordinate locations for coaxial helical stackings in 3-way RNA junctions. These predictions are provided by our tertiary motif search package, named CSminer, which utilizes two versatile methodologies: random forests and covariance models. A coaxial helical stacking tertiary motif occurs in a 3-way RNA junction where two separate helical elements form a pseudocontiguous helix and provide thermodynamic stability to the RNA molecule as a whole. Our CSminer tool first uses a genome-wide search method based on covariance models to find a genomic region that may potentially contain a coaxial helical stacking tertiary motif. CSminer then uses a random forests classifier to predict whether the genomic region indeed contains the tertiary motif. Experimental results demonstrate the effectiveness of our approach.

Recommender systems have proven to be an effective method to deal with the problem of information overload in finding interesting products. It is still a challenge to increase the accuracy and diversity of recommendation algorithms to fulfill users' preferences. To provide a better solution, in this paper, we propose a novel recommendation algorithm based on heterogeneous diffusion process on a user-object bipartite network. This algorithm generates personalized recommendation results on the basis of the physical dynamic feature of resources diffusion which is influenced by objects' degrees and users' interest degrees. Detailed numerical analysis on two benchmark datasets shows that the presented algorithm is of high accuracy, and also generates more diversity.