International Journal of Computational Intelligence and Applications

Granting those heart failure patients who are recipients of an implantable rotary blood pump (iRBP) greater functionality in daily activities is a key long-term strategy currently being pursued by many research groups. A reliable technique for noninvasive detection of the various pumping states, most notably that of ventricular collapse or suction, is an essential component of this strategy. Presented in this study is such a technique, whereby various indicators are derived from the noninvasive pump feedback signals, and a suitable computational methodology developed to classify the pumping states of interest. Clinical telemetry data from ten implant recipients was categorized (with the aid of trans-oesophageal echocardiography) into the normal and suction states. These data are used to develop a pumping state classifier based on an artificial neural network (ANN). Nine indices, derived from the noninvasive impeller speed signal, form the inputs to this ANN classifier. During validation, the resulting ANN classifier achieved a maximum sensitivity of 98.54% (609/618 samples of 5 s in length) and specificity of 99.26% (12,123/12,213 samples) for correct detection of the suction state. The ability to detect the suction state with such a high degree of accuracy provides a critical parameter both for control strategy development, and for clinical care of the implant recipient.

Body surface potential maps were investigated to identify a set of optimal recording sites required to discriminate between several diseases. Specifically, recordings captured from subjects exhibiting myocardial infarction or left ventricular hypertrophy, as well as a control group consisting of healthy subjects, were investigated. Owing to the fact that multi-class problems are inherently difficult to solve we divided the problem into several two-class scenarios. Six data sets were generated from the available 744 records, each viewing the available data differently, to form several two-class problems. A data-driven selection algorithm was applied to each of the generated data sets to produce six classification models, each utilizing as features those recording sites offering most to the discrimination task being investigated. Subsequently, a framework was introduced to facilitate the combination of outputs from each classifier. Essentially, the framework used the outputs from half of the classification models to determine which of the remaining models would be employed to form a final decision. A benchmark, in the form of a multi-group classifier, was introduced to evaluate the perceived benefits of the proposed approach. An improvement of approximately 10% upon the benchmark was observed resulting in an overall accuracy of 79.19%.

An automated method is presented for segmentation of two-dimensional HRCT images of the lung into regions of four lung patterns: normal, emphysema, honeycombing, and ground-glass opacity (GGO). Segmentation was implemented in two stages. At the first stage, pixel-wise classification of the lung area was performed using local textural features extracted by the wavelet transform. At the second stage, classification results were refined by application of knowledge-based rules. Performance of the method was compared on two sets of HRCT images: one included HRCT images with characteristic examples of lung patterns and the other consisted of unselected HRCT images that represented a model of routine operations at a general radiology practice. On the first set of images, sensitivity of the method ranged from 0.92 to 0.99, and specificity ranged from 0.96 to 0.99. On the second set of images, sensitivity and specificity were, respectively, 0.49 and 0.95 for emphysema, 0.87 and 0.55 for normal, 0.34 and 0.99 for honeycombing, and 0.57 and 0.94 for GGO. The two-stage approach allowed for simple and effective application of high-level knowledge about appearance of lung patterns on HRCT images and did not require feature and region of interest size selection for the first stage of pixel-wise lung pattern classification.

Surface electromyogram (SEMG) has numerous applications, but the presence of artifacts and cross talk especially at low level of muscle activity makes the recordings unreliable. Spectral and temporal overlap can make the removal of artifacts and noise, or separation of relevant signals from other bioelectric signals extremely difficult. Identification of hand gestures using low level of SEMG is one application that has a number of applications but the presence of high level of cross talk makes such an application highly unreliable. Individual muscles may be considered as independent at the local level and this makes an argument for separating the signals using independent component analysis (ICA). In the recent past, due to the easy availability of ICA tools, a number of researchers have attempted to use ICA for this application. This paper reports research conducted to evaluate the use of ICA for the separation of muscle activity and removal of the artifacts from SEMG. It discusses some of the conditions that could affect the reliability of the separation and evaluates issues related to the properties of the signals and a number of sources. This paper also identifies the lack of suitable measure of quality of separation for bioelectric signals and it recommends and tests a more robust measure of separation. This paper also proposes semi-blind ICA approach with the combination of prior knowledge of SEMG sources with ICA to identify hand gestures using low level of SEMG recordings. The theoretical analysis and experimental results demonstrate that ICA is suitable for SEMG signals. The results demonstrate the limitations of such applications due to the inability of the system to identify the correct order and magnitude of the signals. This paper determines the suitability of the use of error between estimated and actual mixing matrix as a mean for identifying the quality of separation of the output. This work also demonstrates that semi-blind ICA can accurately identify complex hand gestures from the low-level SEMG recordings.

In this study, we have investigated the electrophysiological differences between alcoholic and control subjects using two different approaches namely complexity and energy (power) analyses. The electroencephalogram data used in this study were recorded from 77 alcoholic and 44 control subjects while the subjects were performing delayed matching to sample object recognition task for three types of stimuli. These were a single stimulus and a second matching or nonmatching stimulus that followed the single stimulus after a delay. The experimental paradigm evokes object recognition, visual short-term memory, and decision-making abilities. The results indicated that all regions (i.e. frontal, central, temporal, parietal, and occipital) in the brain exhibit more complexity and less energy for alcoholic subjects as compared to controls. When different visual stimuli pairs were compared among alcoholic and control subjects, the results from energy analysis showed groupwise differences in occipital and parietal regions. These results provide a strong indication on the impairment in brain's electrophysiological activity for alcoholic subjects due to a history of long-term alcohol abuse.

Trip-related falls are a major problem in the elderly population and research in the area has received much attention recently. The focus has been on devising ways of identifying individuals at risk of sustaining such falls. The main aim of this work is to explore the effectiveness of models based on Support Vector Machines (SVMs) for the automated recognition of gait patterns that exhibit falling behavior. Minimum toe clearance (MTC) during continuous walking on a treadmill was recorded on 10 healthy elderly and 10 elderly with balance problems and with a history of tripping falls. Statistical features obtained from MTC histograms were used as inputs to the SVM model to classify between the healthy and balance-impaired subjects. The leave-one-out technique was utilized for training the SVM model in order to find the optimal model parameters. Tests were conducted with various kernels (linear, Gaussian and polynomial) and with a change in the regularization parameter, C, in an effort to identify the optimum model for this gait data. The receiver operating characteristic (ROC) plots of sensitivity and specificity were further used to evaluate the diagnostic performance of the model. The maximum accuracy was found to be 90% using a Gaussian kernel with σ² = 10 and the maximum ROC area 0.98 (80% sensitivity and 100% specificity), when all statistical features were used by the SVM models to diagnose gait patterns of healthy and balance-impaired individuals. This accuracy was further improved by using a feature selection method in order to reduce the effect of redundant features. It was found that two features (standard deviation and maximum value) were adequate to give an improved accuracy of 95% (90% sensitivity and 100% specificity) using a polynomial kernel of degree 2. These preliminary results are encouraging and could be useful not only for diagnostic applications but also for evaluating improvements in gait function in the clinical/rehabilitation contexts.

Patients with transhumeral amputations and C5/C6 quadriplegia may be able to use voluntary shoulder motion as command signals for powered prostheses and functional electrical stimulation, respectively. Spatiotemporal synergies exist between the shoulder and elbow joints for goal-oriented reaching movements as performed by able-bodied subjects. We are using a multi-layer perceptron neural network to discover and embody these synergies. Such a network could be used as a high-level controller that could predict the desired distal arm joint kinematics from the voluntary movements of the shoulder joint of an able-bodied subject. We evaluated this for a task that involved reaching to 16 targets in a horizontal plane. After reaching reasonable offline prediction accuracy for our neural networks, we then deployed the best network to make real-time predictions of the elbow angles and examined its performance on both inter- and intra-subject trials. Finally, we extended the model to utilize the five degrees-of-freedom at the shoulder to control the five degrees-of-freedom required for a prosthetic arm and hand to reach and grasp variously oriented objects in the extrapersonal workspace. Such a system, although very simple, was readily controllable for a reach and grasp task presented to the subject in a virtual reality environment.

The availability of inexpensive, miniaturized sensors and the development of efficient and accurate filtering algorithms are important elements in order that pedestrian navigation system (PNS) will become an enabling technology for monitoring unconstrained daily living activities of human subjects.

In this paper, we develop and test a computational method that estimates the path traveled by a walker from in-shoe mounted tri-axis inertial and magnetic sensors. The interest for inertial and magnetic sensors stems from their low cost, suitable form factors, small power consumption, which may give rise to self-contained, portable PNS whose hindrance to walking is limited.

An important element in our approach is represented by an extended Kalman filter (EKF), whose aim is to estimate the body part orientation and to perform the in-line calibration of the tri-axis magnetometer. Two validation tests are applied to either acceleration or magnetic vector measurements, in order to adapt the measurement noise covariance matrix against the effects of body motion and external magnetic fields on the sensed gravity and earth's magnetic fields involved in orientation determination. Additionally, some biomechanical facts about how humans normally move are exploited to enhance the positioning performance of the developed method.