Search

Over the past few years, Deep learning (DL) has revolutionized the field of data analysis. Not only are the algorithmic paradigms changed, but also the performance in various classification and prediction tasks has been significantly improved with respect to the state-of-the-art, especially in the area of computer vision. The progress made in computer vision has produced a spillover in many other domains, such as biomedical engineering. Some recent works are directed towards surface electromyography (sEMG) based hand gesture recognition, often addressed as an image classification problem and solved using tools such as Convolutional Neural Networks (CNN). This paper extends our previous work on the application of the Hilbert space-filling curve for the generation of image representations from multi-electrode sEMG signals, by investigating how the Hilbert curve compares to the Peano- and Z-order space-filling curves. The proposed space-filling mapping methods are evaluated on a variety of network architectures and in some cases yield a classification improvement of at least 3%, when used to structure the inputs before feeding them into the original network architectures.

NeuroEvolution (NE) refers to a family of methods for optimizing Artificial Neural Networks (ANNs) using Evolutionary Computation (EC) algorithms. NeuroEvolution of Augmenting Topologies (NEAT) is considered one of the most influential algorithms in the field. Eighteen years after its invention, a plethora of methods have been proposed that extend NEAT in different aspects. In this article, we present a systematic literature review (SLR) to list and categorize the methods succeeding NEAT. Our review protocol identified 232 papers by merging the findings of two major electronic databases. Applying criteria that determine the paper's relevance and assess its quality, resulted in 61 methods that are presented in this article. Our review article proposes a new categorization scheme of NEAT's successors into three clusters. NEAT-based methods are categorized based on 1) whether they consider issues specific to the search space or the fitness landscape, 2) whether they combine principles from NE and another domain, or 3) the particular properties of the evolved ANNs. The clustering supports researchers 1) understanding the current state of the art that will enable them, 2) exploring new research directions or 3) benchmarking their proposed method to the state of the art, if they are interested in comparing, and 4) positioning themselves in the domain or 5) selecting a method that is most appropriate for their problem.

Deep learning has transformed the field of data analysis by dramatically improving the state of the art in various classification and prediction tasks, especially in the area of computer vision. In biomedical engineering, a lot of new work is directed toward surface electromyography (sEMG)-based gesture recognition, often addressed as an image classification problem using convolutional neural networks (CNNs). In this paper, we utilize the Hilbert space-filling curve for the generation of image representations of sEMG signals, which allows the application of typical image processing pipelines such as CNNs on sequence data. The proposed method is evaluated on different state-of-the-art network architectures and yields a significant classification improvement over the approach without the Hilbert curve. Additionally, we develop a new network architecture (MSHilbNet) that takes advantage of multiple scales of an initial Hilbert curve representation and achieves equal performance with fewer convolutional layers.

Introduction: Low grip work and high feelings of self-perceived fatigue could be an early characteristic of decline in reserve capacity, which comes to full expression as physical frailty in a later stage. When grip work and self-perceived fatigue can be identified as characteristics differentiating between robustness and pre-frailty it might allow to identify pre-frailty earlier. Therefore, this study aimed to investigate whether the combination of grip work and self-perceived fatigue is related to pre-frailty in well-functioning older adults aged 80 and over. Methods: Four-hundred and five community-dwelling older adults aged 80 and over (214 robust and 191 pre-frail) were assessed for muscle endurance (grip Work corrected for body weight (GW_bw)), self-perceived fatigue (MFI-20) and frailty state (Fried Frailty Index, FFI). A Capacity to Perceived Vitality ratio (CPV) was calculated by dividing GW_bw by the MFI-20 scores. ANCOVA analysis (corrected for age and gender) was used to compare robust and pre-frail older adults, and binary logistic regressions were applied to analyze the relationship between CPV and pre-frailty status. Results: Pre-frail older adults who scored negative on the exhaustion item of the FFI still showed significantly lower GW (p < 0.001), CPV ratios (p < 0.001) and higher self-perceived fatigue (p < 0.05) compared to the robust ones. The likelihood for pre-frailty related significantly to higher age, being men and lower CPV ratios. In women, every unit increase in CPV ratio decreased the likelihood for pre-frailty by 78% (OR 0.22; 95% CI: 0.11–0.44), for men this effect was less strong (34%, OR 0.66; 95% CI: 0.47–0.93). Conclusions: Pre-frail community-dwelling persons aged 80 years and over without clinical signs of exhaustion on the FFI still experience significantly higher fatigue levels (lower Grip Work, higher self-perceived fatigue and lower CPV levels) compared to robust ones. CPV ratio could therefore be a good tool to identify subclinical fatigue in the context of physical (pre-)frailty.

OBJECTIVE The traditional freehand technique for external ventricular drain (EVD) placement is most frequently used, but remains the primary risk factor for inaccurate drain placement. As this procedure could benefit from image guidance, the authors set forth to demonstrate the impact of augmented-reality (AR) assistance on the accuracy and learning curve of EVD placement compared with the freehand technique. METHODS Sixteen medical students performed a total of 128 EVD placements on a custom-made phantom head, both before and after receiving a standardized training session. They were guided by either the freehand technique or by AR, which provided an anatomical overlay and tailored guidance for EVD placement through inside-out infrared tracking. The outcome was quantified by the metric accuracy of EVD placement as well as by its clinical quality. RESULTS The mean target error was significantly impacted by either AR (p = 0.003) or training (p = 0.02) in a direct comparison with the untrained freehand performance. Both untrained (11.9 ± 4.5 mm) and trained (12.2 ± 4.7 mm) AR performances were significantly better than the untrained freehand performance (19.9 ± 4.2 mm), which improved after training (13.5 ± 4.7 mm). The quality of EVD placement as assessed by the modified Kakarla scale (mKS) was significantly impacted by AR guidance (p = 0.005) but not by training (p = 0.07). Both untrained and trained AR performances (59.4% mKS grade 1 for both) were significantly better than the untrained freehand performance (25.0% mKS grade 1). Spatial aptitude testing revealed a correlation between perceptual ability and untrained AR-guided performance (r = 0.63). CONCLUSIONS Compared with the freehand technique, AR guidance for EVD placement yielded a higher outcome accuracy and quality for procedure novices. With AR, untrained individuals performed as well as trained individuals, which indicates that AR guidance not only improved performance but also positively impacted the learning curve. Future efforts will focus on the translation and evaluation of AR for EVD placement in the clinical setting.

The latest advancements in the field of deep learning and biomedical engineering have allowed for the development of myoelectric interfaces based on deep neural networks. A longstanding problem of these interfaces is that the models cannot easily be applied to new users due to the high variability and stochastic nature of the electromyography signals. Further training a new model for every new subject requires the collection of large volumes of data. Therefore, this work proposes a transfer learning (TL) scheme which allows reusing the knowledge of a pre-existing model for a new user. Firstly, a convolutional neural network (CNN) is trained on an initial dataset using the data of multiple subjects. Then, the weights of this model are fine-tuned for a new target subject. The approach is evaluated on the Ninapro datasets DB2 and DB7. The experimentation included three different CNN models and eight preprocessing alternatives. The results showed that the success of the TL method depends on how the data are preprocessed. Specifically, the biggest accuracy improvement (+5.14%) is achieved when only the first 20% of the signal duration is used.

Cerebral palsy (CP) is the most common cause of permanent neurological disabilities in children. Many children require long-term daily physiotherapy (PT), and videogaming is a promising tool to increase motivation in rehabilitation. The short- and medium-term effects of an intervention with rehabilitation specific videogames were evaluated on individually defined therapy goals, gross motor function, and motivation. Thirty-two children with bilateral spastic CP, Gross Motor Function Classification level III-IV, and 6-15 years were randomized into an intervention group (regular PT and gaming) or a control group (regular PT), followed by a crossover. The effects of both training periods (each 12 weeks) were compared using the Goal Attainment Scale (GAS), Trunk Control Measurement Scale (TCMS), Pediatric Balance Scale (PBS), Gross Motor Function Measure-88 (GMFM-88), and Dimensions of Mastery Motivation Questionnaire (DMQ). After 3 months follow-up, children were retested using the GMFM, TCMS, and PBS. The GAS change scores were significantly higher after the intervention compared to the control period (8.5 and 2.4

We propose a neuroevolutionary speciation mechanism that is applied on NeuroEvolution of Augmenting Topologies (NEAT) that solely evolves neural networks’ topology and weights and its extension HA-NEAT that also evolves activation functions. The new speciation mechanism is defined based on the behavior of the individuals rather than their topological similarity. Focusing on classification tasks we build artificial datasets of high complexity. Performance is described by (i) median classification accuracy, (ii) computational efficiency (number of generations) and (iii) network complexity (number of nodes and connections). The performance metrics are compared using Kruskal-Wallis hypothesis tests with Bonferroni correction. It is found that the proposed behavioral speciation mechanism outperforms the original speciation solving problems that were not solvable before or improving the accuracy and reducing the network complexity.

Convolutional neural networks are known to require large amounts of data to achieve optimal performance. In addition, data is commonly computationally augmented using a variety of geometric and intensity transformations to further extent the set of training samples. In medical imaging, annotated data is often scarce or costly to obtain, and there is considerable interest in methods to reduce the amount of data needed. In this work, we investigate the relative benefit of increasing the amount of original data, with respect to computationally augmenting the amount of training samples, for the case of false positive reduction of lung nodules candidates. To this end, we have implemented a previously published topology for classification, shown to achieve state of the art results on the publicly available Luna16 dataset. Numerous models were trained using different amounts of unique training samples and different degrees of data augmentation involving rotations and translations, and the performance was compared. Results indicate that in general, better performance is achieved when increasing the amount of data, or augmenting the data more extensively, as expected. Surprisingly however, we observed that after reaching a certain amount of unique training samples, data augmentation leads to significantly better performance compared to adding the same number of new samples to the training dataset. We hypothesize that the augmentation has aided in learning more general {rotation and translation invariant-features, leading to improved performance on unseen data. Future experiments include more detailed characterization of this behavior, and relating this to the topology and amount of parameters to be trained.

Computer aided diagnosis systems are used to assist radiologists in their decision making. The sensitivity of these systems is hindered by the complexity of the structures inside the lungs. Several systems and methods have been proposed to detect and classify lung nodules, but all of them have their strengths and weaknesses. One way to overcome the weaknesses is to combine multiple systems. Systems based on handcrafted features capture a limited set of characteristics from the image, while deep learning based classifiers can deal with a wider range of structures. In this work, several ways to combine a handcrafted feature based classifier with four convolutional neural network are explored. The systems were combined merging the probabilities assigned to the detections in several ways. Support-vector machine, multilayer perceptron and random forest classifiers were used to combine the selected classifiers. The LUNA16 Challenge was used to evaluate the performance of the resulting hybrid systems. In all cases, the hybrid systems outperformed the individual systems. Although the average of sensitivities are similar for most of the combinations, the best hybrid system achieves a gain of 35 extra nodules at 4 FP per scan.

Major hurdles for Microsoft's HoloLens as a tool in medicine have been accessing tracking data, as well as a relatively high-localisation error of the displayed information; cumulatively resulting in its limited use and minimal quantification. The following work investigates the augmentation of HoloLens with the proprietary image processing SDK Vuforia, allowing integration of data from its front-facing RGB camera to provide more spatially stable holograms for neuronavigational use. Continuous camera tracking was able to maintain hologram registration with a mean perceived drift of 1.41 mm, as well as a mean sub 2-mm surface point localisation accuracy of 53%, all while allowing the researcher to walk about a test area. This represents a 68% improvement for the later and a 34% improvement for the former compared with a typical HoloLens deployment used as a control. Both represent a significant improvement on hologram stability given the current state-of-the-art, and to the best of the authors knowledge are the first example of quantified measurements when augmenting hologram stability using data from the RGB sensor.

Objective : The aim of this study was to determine if serious games especially customized for physical rehabilitation could be used in daily clinics with patients that are not familiar with informatics and/or new technologies. Materials and methods : 21 patients participated in this study. Experiments were performed in the physical rehabilitation of a hospital in Morocco. Patients were asked to perform physical rehabilitation exercises with dedicated serious games after their conventional therapy session. A questionnaire was used to evaluate patients’ habits to rehabilitation exercises and satisfaction and expectation about the use of serious games for physical rehabilitation. This study aimed to determine if such clinical approach would be culturally acceptable within a North African population in order to plan further clinical distribution if the hypothesis appears to be positive. Results : The same problem of low adherence to at-home exercises was found in Morocco compared to Europe and the United-States of America: 60 (30) % of the exercises performed by the patients in Morocco, 48 (28) % in Europe and the Stats. Results of this study shown that serious games are well accepted by the patient (100%) and that it can help during the rehabilitation (90%). Most of the patient prefers the exercises with the games compared to conventional rehabilitation exercises. Conclusion: Even in a context of low education (participants in this study attended school until the age of 10 y.o.), and low access and knowledge about new technology and informatics, rehabilitation exercises within serious games seems to be an interesting option to motivate patient during the rehabilitation with the physiotherapist but also eventually placed at-home to facilitate rehabilitation.

The aim of this study was to determine if the results of activities performed using specially developed serious games for physical rehabilitation could be used as an indicator of the natural maturation and decline of motor control in healthy participants. Eighty-one participants (19 children (5–15 years old), 40 adults (18–65 years old) and 22 aged subjects (60–88 years old) participated in this study. Motions performed were recorded using the Kinect sensor. Three different exercises embedded in the games were used to assess upper limb, trunk and lower limb control. The trial duration and accuracy, measures of gross motor function and fine motor control, respectively, were computed for each participant. ANOVA tests shows statistically significant differences between the three groups for duration (53 ± 15, 27 ± 10 and 119 ± 30 s for children, adults and elderly subjects respectively) and accuracy (87 ± 5, 89 ± 10 and 70 ± 8% for children, adults and elderly subjects respectively). The slopes of the curves that approximated the evolution of the performance over various ages are coherent with previous studies about motor control development and physiological decline. The proposed solution, i.e. serious games rehabilitation exercises coupled to motion analysis, seems to be an interesting tool to assess global motor function. Further studies are needed to study the influence of pathologies on the studied parameters.

This study aimed to investigate: 1) the effect of optic flow speed manipulation on active participation during robot-assisted treadmill walking (RATW), 2) the influence of the type of virtual environment, and 3) the level of motion sickness and enjoyment. Twenty-eight healthy older adults were randomized in two groups: “stimulus rich” Park group (50% male, 61± 6 year) and “stimulus poor” Hallway group (43% male, 62± 5 year). Subjects walked in the Lokomat with immersive virtual reality (VR) with a matched, slow and fast optic flow speed, each lasting 7 minutes. Active participation was measured by continuously assessing the human-machine interaction torques at the hip and knee joints and muscle activity of the Vastus Medialis and Biceps Femoris. Motion sickness and enjoyment were assessed with the Simulator Sickness Questionnaire (SSQ) and Physical Activity Enjoyment Scale (PACES) respectively. In both groups optic flow speed manipulation in both directions led to a decrease in bilateral hip interaction torques towards flexion at the end of the stance phase compared to matched speed. In the Hallway group, walking with slow optic flow elicited 32% more muscle activity of the Vastus Medialis. There were no significant differences between both groups for the SSQ and PACES. Optic flow speed manipulation appears to have only a small effect on the active participation of healthy people during RATW. The type of virtual environment did not affect their activity, motion sickness or enjoyment. However, the addition of immersive VR during RATW was well tolerated and enjoyable. Further research with patients is necessary.

Frailty is accompanied by limitations in activities of daily living (ADL). These are associated with reduced quality of life, institutionalization and higher health care costs. Long-term monitoring ADL could allow creating effective interventions and thus reduce the occurrence of adverse health outcomes. The main objective of this study was to evaluate if ADL task performance can be assessed by activity measurements based on IMUs, and whether these measures can differentiate individual’s frailty. ADL data was obtained from seventeen elderly who performed two ADL tasks - tea making task (TEA) and gardening task (GARDEN). Acceleration data of the dominant hand was collected using an activity sensor. Participants were split up in two groups, FRAIL (n = 6; Fried score ≥ 2) and CONTROL (n = 11; Fried score ≤ 1) retrospectively. Collected data were used to determine trial duration (TD), relative activity (RA), peak standard deviation (STD), peaks per second (PPS), peaks ratio (RATIO), weighted sum of acceleration per second (SUM), signal to noise ratio (S2N) and mean peak acceleration (MPA). STD, RATIO, SUM and MPA showed good reliability over both tasks. Four of the calculated parameters (RA, PPS, RATIO, SUM) revealed significant results differentiating between FRAIL and CONTROL (effect sizes 1.30–1.77). Multiple linear regression showed that only STD correlated with the Fried score. In summary, the results demonstrate that ADL task performance can be assessed by IMU-based activity measures and further allows drawing conclusions on the frailty status of elderly, although the predictability of the exact Fried score was limited.

In the Dutch capitalistic consumer society, things are not made to stand the test of time, but to be replaced by other things within the foreseeable future. They are made to be thrown away, and quite often as quickly as possible, because it is this characteristic that guarantees a new purchase. In this contribution the authors will focus on a white T-shirt. The main questions are: What is the current practice in the Netherlands regarding the disposal, separation, reusing, and recycling of this Dutch white trash and how does the legal framework regulate the matter of RMG waste?.

Deep learning (DL) has transformed the field of data analysis by dramatically improving the state of the art in various classification and prediction tasks. Especially in the area of computer vision and speech processing, DL has recently demonstrated better performance and generalisation properties, compared to classical machine learning approaches, which are based on the extraction of hand-crafted model-based features followed by classification. Hand gestures and speech constitute two of the most important modalities in human-to-human communication and man-machine interaction. In biomedical engineering, a lot of new work is directed towards electromyography-based gesture recognition. In this paper, we present a brief overview of DL methods for electromyography-based hand gesture recognition and then we select from literature a simple model based on Convolutional Neural Networks that we consider as the baseline model. The proposed modifications to the baseline model yield a 3% classification improvement. In the current paper, we concentrate on the explanatory analysis of this performance improvement. An ablation study identifies which modifications are the most important ones, and label smoothing is investigated to verify if the results can be improved by reducing a priori bias. The analysis helps in understanding the limitations of the model and exploring new ways to improve the performance.

Summary Background To date, the accuracy of bio-impedance (BIA) to assess body composition & sarcopenia in persons aged 80 and over remains unclear. Objective We aimed to evaluate the agreement between dual energy X-ray absorptiometry (DXA) and BIA equations to determine lean mass, as well as their suitability to identify sarcopenia. Design 174 community dwelling well-functioning persons (83 women, 91 men) aged 80 and over were included. Appendicular lean mass (ALM) was predicted using BIA-based equations available in literature, and compared to DXA outcomes. Through cross-validation and stepwise multiple linear regression, a new ALM-formula was generated suitable for this population. Results Literature-based BIA equations systematically overestimated ALM. The new prediction formula that we propose for the 80+ is: ALM = 0,827+(0,19*Impedance Index)+(2,101*Sex)+(0,079*Weight); R2 = 0,888; SEE = 1,450 kg. Sarcopenia classification based on our new BIA equation for ALM showed better agreement with DXA (k ≥ 0,454) compared to literature-based BIA equations (k Conclusions Despite the high correlation between both methods, literature-based BIA equations consistently overestimate ALM compared to DXA in persons aged 80 and over. We proposed a new equation for ALM, reaching higher agreement with DXA and thus improving the accuracy of BIA for this specific age group.

Deep learning (DL) has transformed the field of data analysis by dramatically improving the state of the art in various classification and prediction tasks, especially in the area of computer vision. In biomedical engineering, a lot of new work is directed towards surface electromyography (sEMG) based gesture recognition, often addressed as an image classification problem using Convolutional Neural Networks (CNN). In this paper, we utilize the Hilbert space-filling curve for the generation of image representations of sEMG signals that are then classified by CNN. The proposed method is evaluated on different network architectures and yields a classification improvement of more than 3%.

In recent years, the successful application of Deep Learning methods to classification problems has had a huge impact in many domains. In biomedical engineering, the problem of gesture recognition based on electromyography is often addressed as an image classification problem using Convolutional Neural Networks. In this paper, we approach electromyography-based hand gesture recognition as a sequence classification problem using Temporal Convolutional Networks. The proposed network yields an improvement in gesture recognition of almost 5% to the state of the art reported in the literature, whereas the analysis helps in understanding the limitations of the model and exploring new ways to improve its performance.

Objective Although dual-task paradigm is commonly used in rehabilitation there is, currently, a lack of information about quantitative balance assessment performed in dual-task conditions. The aim of this study is to validate the use of the Nintendo Wii Balance Board (WBB) to assess balance modifications induced by dual-task activity in healthy subjects. Material and Methods Seventy-five healthy participants (37 ± 14 years old, 173± 8 cm, 73± 16 kg, 35 women) participated in this study. Balance was recorded in control situation (baseline) and under two different cognitive tasks (verbal fluency and calculation). Each trial lasted for 60 seconds and were performed with eyes open and closed. Nine parameters were extracted based on center of pressure displacement based on a previously-validated method. ANOVA tests were used to compare the different conditions followed by Bonferroni’s corrections. Protocol was repeated one week after to assess the reproducibility. Results Statistically significant differences were found comparing eyes open and closed situation. Concerning the influence of dual-tasks statistically significant decrease of the balance was observed for both eyes open and closed conditions (increase of the total sway and the speed of displacement). Concerning the reproducibility all parameters shows good to excellent ICC values for the different conditions. Conclusion In this study, we demonstrated the validity of the WBB to detect the effect of the dual-task on balance in healthy subjects. Further studies are needed to determine if such a kind of evaluation can be used in clinics with subjects at risk of falling.

Objective The aim of this work was to determine if scoring mechanisms embedded in cognitive mobile games (CMG) designed for cognitive training are sensitive enough to detect changes in cognitive function induced by various acute respiratory loads in healthy subjects. Material and methods Thirty healthy subjects participated in this study (25 ± 4 years old, 13 women). A set of three brief CMG was used to assess cognitive function (Rush Back, Must Sort, True Color) in control situation and at four different inspiratory loads (through a mouthpiece without resistance and with 10, 50 and 70% of the maximal inspiratory pressure). Results Statistically significant decreases in CMG scores were observed in Rush Back (p = 0.032) and True Color (p = 0.002) when the subjects breathed through the mouthpiece without resistance compared to the control condition. A statistical difference was observed for Must Sort (p = 0.003) between baseline and 10% of Inspiratory Threshold Load (ITL). Significant differences for the three games were observed between the baseline and 50% of ITL and between baseline and 70% of ITL. Conclusion CMG designed for cognitive training are sensitive in detecting transitional changes in cognitive function induced by low, medium and high acute respiratory loads in healthy subjects. This offers interesting new possibilities for the assessment and long-term follow-up of patients suffering from chronic respiratory disease, since this type of assessment could easily be completed independently by patients in their own homes, and could be combined with rehabilitation exercises as an evaluative measure.

The aim of this paper was to investigate the effect of commercial video games (VGs) in physical rehabilitation of motor functions. Several databases were screened (Medline, SAGE Journals Online, and ScienceDirect) using combinations of the following free-text terms: commercial games, video games, exergames, serious gaming, rehabilitation games, PlayStation, Nintendo, Wii, Wii Fit, Xbox, and Kinect. The search was limited to peer-reviewed English journals. The beginning of the search time frame was not restricted and the end of the search time frame was 31 December 2015. Only randomized controlled trial, cohort, and observational studies evaluating the effect of VGs on physical rehabilitation were included in the review. A total of 4728 abstracts were screened, 275 were fully reviewed, and 126 papers were eventually included. The following information was extracted from the selected studies: device type, number and type of patients, intervention, and main outcomes. The integration of VGs into physical rehabilitation has been tested for various pathological conditions, including stroke, cerebral palsy, Parkinson's disease, balance training, weight loss, and aging. There was large variability in the protocols used (e.g. number of sessions, intervention duration, outcome measures, and sample size). The results of this review show that in most cases, the introduction of VG training in physical rehabilitation offered similar results as conventional therapy. Therefore, VGs could be added as an adjunct treatment in rehabilitation for various pathologies to stimulate patient motivation. VGs could also be used at home to maintain rehabilitation benefits.

The range of applications of electromyography-based gesture recognition has increased over the last years. A common problem regularly encountered in literature is the inadequate data availability. Data augmentation, which aims at generating new synthetic data from the existing ones, is the most common approach to deal with this data shortage in other research domains. In the case of surface electromyography (sEMG) signals, there is limited research in augmentation methods and quite regularly the results differ between available studies. In this work, we provide a detailed evaluation of existing (i.e., additive noise, overlapping windows) and novel (i.e., magnitude warping, wavelet decomposition, synthetic sEMG models) strategies of data augmentation for electromyography signals. A set of metrics (i.e., classification accuracy, silhouette score, and Davies&ndash, Bouldin index) and visualizations help with the assessment and provides insights about their performance. Methods like signal magnitude warping and wavelet decomposition yield considerable increase (up to 16%) in classification accuracy across two benchmark datasets. Particularly, a significant improvement of 1% in the classification accuracy of the state-of-the-art model in hand gesture recognition is achieved.

In recent years, Deep Learning methods have been successfully applied to a wide range of image and speech recognition problems highly impacting other research fields. As a result, new works in biomedical engineering are directed towards the application of these methods to electromyography-based gesture recognition. In this paper, we present a brief overview of Deep Learning methods for electromyography-based hand gesture recognition along with an analysis of a modified simple model based on Convolutional Neural Networks. The proposed network yields a 3% improvement on the classification accuracy of the basic model, whereas the analysis helps in understanding the limitations of the model and exploring new ways to improve the performance.

There is a link between breathing and balance and posture. When the inspiratory loads are increased by pathologies, there is a decrease of postural control. The increase of the inspiratory load on respiratory muscles is a common feature in various chronic pulmonary pathologies. Consequently, the balance of those patients is likely affected., SCOPUS: ar.j, info:eu-repo/semantics/published

NeuroEvolution (NE) is a powerful method that uses Evolutionary Algorithms (EAs) for learning Artificial Neural Networks (ANNs). However, NE's performance is determined by the definition of dozens of parameters that guide the search of the EAs. In this study we apply automatic algorithm configuration for tuning the parameters of a NE method in an offline matter. The tuned NE method is then used to evolve the weights, topology and activation functions of ANNs while performing feature selection and its performance is compared to the case of using default parameters. We show that tuning the parameters results in NE methods able to solve the problems with 100% accuracy in significantly less generations.

The development of portable haematology analysers receives increased attention due to their deployability in resource-limited or emergency settings. Lens-free in-line holographic microscopy is one of the technologies that is being pushed forward in this regard as it eliminates complex and expensive optics, making miniaturisation and integration with microfluidics possible. On-chip flow cytometry enables high-speed capturing of individual cells in suspension, giving rise to high-throughput cell counting and classification. To perform a real-time analysis on this high-throughput content, we propose a fast and robust framework for the classification of leukocytes. The raw data consists of holographic acquisitions of leukocytes, captured with a high- speed camera as they are flowing through a microfluidic chip. Three different types of leukocytes are considered: granulocytes, monocytes and T-lymphocytes. The proposed method bypasses the reconstruction of the holographic data altogether by extracting Zernike moments directly from the frequency domain. By doing so, we introduce robustness to translations and rotations of cells, as well as to changes in distance of a cell with respect to the image sensor, achieving classification accuracies up to 96.8%. Furthermore, the reduced computational complexity of this approach, compared to traditional frameworks that involve the reconstruction of the holographic data, allows for very fast processing and classification, making it applicable in high-throughput flow cytometry setups.

Purpose: Cerebral palsy (CP) leads to various clinical signs mainly induced by muscle spasticity and muscle weakness. Among these ones impaired balance and posture are very common. Traditional physical therapy exercise programs are focusing on this aspect, but it is difficult to motivate patients to regularly perform these exercises, especially at home without therapist supervision. Specially developed serious games (SG) could therefore be an interesting option to motivate children to perform specific exercise for balance improvement. Method: Ten CP children participated in this study. Patients received four sessions of SG included into conventional therapy (1 session of 30 min a week during 4 weeks). Trunk control and balance were assessed using Trunk Control Motor Scale (TCMS) before and after interventions. Results: Children presented a significant improvement in TCMS global score after interventions [37.6 (8.7) and 39.6 (9.5) before and after interventions, respectively, p = 0.04]. Conclusion: SG could therefore be an interesting option to integrate in the conventional treatment of CP children.Implication for RehabilitationCerebral palsy (CP) leads to balance issues.Rehabilitation exercises are not performed (enough) at home.Serious games (SG) could increase patients’ motivation.SG increase balance control of CP children. Cerebral palsy (CP) leads to balance issues. Rehabilitation exercises are not performed (enough) at home. Serious games (SG) could increase patients’ motivation. SG increase balance control of CP children.

Feature Selective Neuroevolution of Augmenting Topologies (FS-NEAT) and Feature Deselective Neuroevolution of Augmenting Topologies (FD-NEAT) are two popular methods for optimizing the topology and the weights of Artificial Neural Networks (ANNs) simultaneously with performing feature selection. However, no study exists that systematically investigates their performance on the exclusive-or (XOR) problem with increasing complexity. Moreover, it is unknown whether the choice of a different initial topology of the ANNs would influence the performance of the two algorithms. For this reasons, this paper investigates the performance of FD-NEAT and FS-NEAT in terms of accuracy, number of generations required for their convergence to the optimal solution and their ability of selecting the relevant features in artificial datasets with irrelevant features. The comparisons are performed based on hypothesis tests (Wilcoxon rank sum test, p

Contains fulltext : 179531.pdf (Publisher’s version ) (Open Access) Automatic detection of pulmonary nodules in thoracic computed tomography (CT) scans has been an active area of research for the last two decades. However, there have only been few studies that provide a comparative performance evaluation of different systems on a common database. We have therefore set up the LUNA16 challenge, an objective evaluation framework for automatic nodule detection algorithms using the largest publicly available reference database of chest CT scans, the LIDC-IDRI data set. In LUNA16, participants develop their algorithm and upload their predictions on 888 CT scans in one of the two tracks: 1) the complete nodule detection track where a complete CAD system should be developed, or 2) the false positive reduction track where a provided set of nodule candidates should be classified. This paper describes the setup of LUNA16 and presents the results of the challenge so far. Moreover, the impact of combining individual systems on the detection performance was also investigated. It was observed that the leading solutions employed convolutional networks and used the provided set of nodule candidates. The combination of these solutions achieved an excellent sensitivity of over 95% at fewer than 1.0 false positives per scan. This highlights the potential of combining algorithms to improve the detection performance. Our observer study with four expert readers has shown that the best system detects nodules that were missed by expert readers who originally annotated the LIDC-IDRI data. We released this set of additional nodules for further development of CAD systems. 13 p.

Aging is associated with an increased risk of falling. This lead to a decrease of autonomy and deleterious consequences in case of fall. Rehabilitation exercises must be performed in order to maintain optimal balance. Since a few year video games are used into conventional rehabilitation to motivate patients to regularly do those exercises. Since the motion performed by the patients can be recorded during the games it offers new possibilities to perform evaluation and follow-up of the patients. The aim of this study was to determine if the data collected during specially developed balance rehabilitation exercises using the Microsoft Kinect™ sensor and the Wii Balance Board™ are correlated with the functional and clinical status of elderly patients.

The shoulder complex is one of the most movable joints of the body. Clinicians do not have tools to assess three-dimensional (3D) shoulder range of motion (RoM) during complex motions. The Kinect™ sensor, a cost-effective markerless motion capture system (MLS), offers interesting possibilities in clinical functional analysis and rehabilitation. However, to our best knowledge, no study has investigated the use of a Kinect™ sensor to analytically evaluate 3D motion of the shoulder complex during functional tasks. Therefore, the aim of this study was to assess the capacities of this device and to determine its limitations in order to understand whether or not it can be used in daily clinics to assess functional shoulder motions. Twenty subjects participated in this study; three functional motions were simultaneously recorded with a Marker Based System (MBS) and MLS. 3D motion analysis of the upper limb using MLS coupled to a post-processing optimization procedure can thus be achieved with the Kinect™ sensor.

In the last decade, technological advances in the gaming industry have allowed the marketing of hardware for motion and balance control that is based on technological concepts similar to scientific and clinical equipment. Such hardware is attractive to researchers and clinicians for specific applications. However, some questions concerning their scientific value and the range of future potential applications have yet to be answered. This article attempts to present an objective analysis about the pros and cons of using such hardware for scientific and clinical purposes and calls for a constructive discussion based on scientific facts and practical clinical requests that are emerging from application fields.

To date, the major dilemma concerning frailty is the lack of a standardized language regarding its operationalization. Considering the demographic challenge that the world is facing, standardization of frailty identification is indeed the first step in tackling the burdensome consequences of frailty. To demonstrate this diversity in frailty assessment, the available frailty instruments have been linked to the International Classification of Functioning, Disability, and Health (ICF): a standardized and hierarchically coded language developed by World Health Organization regarding health conditions and their positive (functioning) and negative (disability) consequences. A systematic review on frailty instruments was carried out in PubMed, Web of Knowledge, and PsycINFO. The items of the identified frailty instruments were then linked to the ICF codes. 79 original or adapted frailty instruments were identified and categorized into single (n = 25) and multidomain (n = 54) groups. Only 5 frailty instruments (indexes) were linked to all 5 ICF components. Whereas the ICF components Body Functions and Activities and Participation were frequently linked to the frailty instruments, Body Structures, Environmental and Personal factors were sparingly represented mainly in the multidomain frailty instruments. This review highlights the heterogeneity in frailty operationalization. Environmental and personal factors should be given more thought in future frailty assessments. Being unambiguous, structured, and neutral, the ICF language allows comparing observations made with different frailty instruments. In conclusion, this systematic overview and ICF translation can be a cornerstone for future standardization of frailty assessment.

Product-oriented policies gain more and more attention in environmental policies. In this framework, we studied cases (here presented only housing and computers and paper) in a life cycle perspective in order to determine their environmental impacts (in terms of GHG emissions, resource use and waste production) over the period 1990–2010. We also evaluated their potential for improvement up to 2010 by applying product-oriented strategies. We identified clear technical (theoretical) potential for improvement, especially for housing, packaging and computers and paper in offices. We then carried out a Delphi-based survey among stakeholders in order to investigate potential measures.

This paper describes the quality of experience (QoE) of a gaming platform developed for the physical rehabilitation of seriously impaired children. Most literature on QoE assessment in such serious games focuses on the difficult issue of assessing the patient's experience, ignoring the therapist's perspective. This paper rather investigates the experience of the therapist in interacting with such a platform. Therapists were asked to use three different user interfaces for configuring the games to the needs of their patients and their preferences were evaluated. Results indicate a preference for a configuration tool with 3D visualization and manipulation over a 2D version. Moreover, a tablet application allowed to setup the exercises much faster than a desktop application, due to the natural way of 3D manipulation with finger gestures.

SummaryIntroduction: This article is part of the Focus Theme of Methods of Information in Medicine on “Methodologies, Models and Algorithms for Patients Rehabilitation”. Objectives: This paper presents a camera based method for identifying the patient and detecting interactions between the patient and the therapist during therapy. Detecting interactions helps to discriminate between active and passive motion of the patient as well as to estimate the accuracy of the skeletal data. Methods: Continuous face recognition is used to detect, recognize and track the patient with other people in the scene (e.g. the therapist, or a clinician). We use a method based on local binary patterns (LBP). After identifying users in the scene we identify interactions between the patient and other people. We use a depth map/point cloud for estimating the distance between two people. Our method uses the association of depth regions to user identities and computes the minimal distance between the regions. Results: Our results show state-of-the-art performance of real-time face recognition using low-resolution images that is sufficient to use in adaptive systems. Our proposed approach for detecting interactions shows 91.9% overall recognition accuracy what is sufficient for applications in the context of serious games. We also discuss limitations of the proposed method as well as general limitations of using depth cameras for serious games. Conclusions: We introduced a new method for frame-by-frame automated identification of the patient and labeling reliable sequences of the patient’s data recorded during rehabilitation (games). Our method improves automated rehabilitation systems by detecting the identity of the patient as well as of the therapist and by detecting the distance between both over time.

Recently, efficient image descriptors have shown promise for image classification tasks. Moreover, methods based on the combination of multiple image features provide better performance compared to methods based on a single feature. This work presents a simple and efficient approach for combining multiple image descriptors. We first employ a Naive-Bayes Nearest-Neighbor scheme to evaluate four widely used descriptors. For all features, “Image-to-Class” distances are directly computed without descriptor quantization. Since distances measured by different metrics can be of different nature and they may not be on the same numerical scale, a normalization step is essential to transform these distances into a common domain prior to combining them. Our experiments conducted on a challenging database indicate that z-score normalization followed by a simple sum of distances fusion technique can significantly improve the performance compared to applications in which individual features are used. It was also observed that our experimental results on the Caltech 101 dataset outperform other previous results.

Summary The present study shows the results and methodology applied to the study of the identification of priority product categories for Belgian product and environmental policy. The main goal of the study was to gather insight into the consumption of products in Belgium and their related life-cycle environmental impacts. The conclusions of this project on the product categories with major environmental contributions can be used to start up working groups involving stakeholders and initiate detailed product studies on the impact reduction potential that could be achieved by means of implementing product policy measures. Several ways of assessing product category environmental impacts and the effects of policy measures have been developed; ‘bottom-up’ or ‘market-life-cycle assessment’ is one of these, and we tried this approach for the situation in Belgium. Simplified life-cycle assessment (LCA) studies were conducted for representative average products within each function-based product category and the results were multiplied with market statistics. Using this approach, we found that building construction, building occupancy, and personal transport are among the major categories for Belgium. The major drawbacks of this approach are the system-level limitations and the existence of a broad spectrum of nonharmonized methods and datasets from which a sound preliminary selection had to be made. Consequently, the retrieval and selection of data was very time consuming and due to this we had to accept some major limitations in the study design. Nevertheless, the study has contributed to the development of a methodology for market-LCA and elements that can be picked up in currently ongoing and future work. The study concludes that to improve the feasibility and acceptance of this type of study there is a need for the development of a harmonized methodology on market-LCA, policy-relevant impact indicators as well as a harmonized and stakeholder-agreed-upon LCA databases.

Environmental effects of economic activities are ultimately driven by consumption, via impacts of the production, use, and waste management phases of products and services ultimately consumed. Integrated product policy (IPP) addressing the life-cycle impacts of products forms an innovative new generation of environmental policy. Yet this policy requires insight into the final consumption expenditures and related products that have the greatest life-cycle environmental impacts. This review article brings together the conclusions of 11 studies that analyze the life-cycle impacts of total societal consumption and the relative importance of different final consumption categories. This review addresses in general studies that were included in the project Environmental Impacts of Products (EIPRO) of the European Union (EU), which form the basis of this special issue. Unlike most studies done in the past 25 years on similar topics, the studies reviewed here covered a broad set of environmental impacts beyond just energy use or carbon dioxide (CO2) emissions. The studies differed greatly in basic approach (extrapolating LCA data to impacts of consumption categories versus approaches based on environmentally extended input-output (EEIO) tables), geographical region, disaggregation of final demand, data inventory used, and method of impact assessment. Nevertheless, across all studies a limited number of priorities emerged. The three main priorities, housing, transport, and food, are responsible for 70% of the environmental impacts in most categories, although covering only 55% of the final expenditure in the 25 countries that currently make up the EU. At a more detailed level, priorities are car and most probably air travel within transport, meat and dairy within food, and building structures, heating, and (electrical) energy-using products within housing. Expenditures on clothing, communication, health care, and education are considerably less important. Given the very different approaches followed in each of the sources reviewed, this result hence must be regarded as extremely robust. Recommendations are given to harmonize and improve the methodological approaches of such analyses, for instance, with regard to modeling of imports, inclusion of capital goods, and making an explicit distinction between household and government expenditure. © 2006 by the Massachusetts Institute of Technology and Yale University.

This paper presents a novel classification approach for microcalcifications extracted from core biopsy tissue samples digitized using micro-CT, a high-resolution 3D imaging modality. Microcalcifications are tiny spots of calcium that may occur in the female breast. Although they are common in healthy woman, they are often an early sign of breast cancer. In case of suspiciousness, a biopsy is conducted and the extracted tissue is pathologically analyzed for the presence of cancer cells. The calcifications themselves however are mostly not analyzed. As a result, there is a ground truth for the tissue samples but not for the individual calcifications. Therefore, this paper presents an approach where the calcifications are first classified according to shape-based clusters before predicting a label for the tissue samples.

Body tracking sensors have become an integral part of the physical therapy not only as a motivational tool used by games but also as a diagnostic instrument. Skeletal tracking helps to analyze and quantify human motion and thus can provide tangible results from therapy sessions. Markerless skeletal tracking with depth sensing cameras represents currently the most popular approach mainly due to low cost cameras that use a model based approach to recognize a human skeleton. The model based approach works in many scenarios but faces limitations as well. This paper presents a method for identifying the patient and detecting interactions between the patient and the therapist. Identifying interactions helps to discriminate between active and passive motion of the patient as well as to estimate the accuracy of the skeletal data. Our experiments show the state-of-the-art performance of real-time face recognition from a low resolution images that is sufficient to use in adaptive systems. We also compare the performance of our interaction detection method with two other approaches (markerless and marker based approach) and shows its superior performance.

The Kinect™ (Microsoft™, Redmond, WA) sensor, originally developed for gaming purposes, may have interesting possibilities for other fields such as posture and motion assessment. The ability of the Kinect sensor to perform biomechanical measurements has previously been studied and shows promising results. However, interday repeatability of the device is still not known.This study assessed the intra- and interday repeatability of the Kinect sensor compared with a standard stereophotogrammetric device during posture assessment for measuring segment lengths. Forty subjects took part in the study. Five motionless captures were performed in one session to assess posture. Data were simultaneously recorded with both devices.Similar intraclass correlations coefficient (ICC) values were found for intraday (ICC=0.94 for the Kinect device and 0.98 for the stereophotogrammetric device) and interday (ICC=0.88 and 0.87, respectively) repeatability.Results of this study suggest that a cost-effective, easy-to-use, and portable single markerless camera offers the same repeatability during posture assessment as an expensive, time-consuming, and nontransportable marker-based device.

The recent availability of the Kinect™ sensor, a low-cost Markerless Motion Capture (MMC) system, could give new and interesting insights into ergonomics (e.g. the creation of a morphological database). Extensive validation of this system is still missing. The aim of the study was to determine if the Kinect™ sensor can be used as an easy, cheap and fast tool to conduct morphology estimation. A total of 48 subjects were analysed using MMC. Results were compared with measurements obtained from a high-resolution stereophotogrammetric system, a marker-based system (MBS). Differences between MMC and MBS were found; however, these differences were systematically correlated and enabled regression equations to be obtained to correct MMC results. After correction, final results were in agreement with MBS data (p = 0.99). Results show that measurements were reproducible and precise after applying regression equations. Kinect™ sensors-based systems therefore seem to be suitable for use as fast and reliable tools to estimate morphology. Practitioner Summary: The Kinect™ sensor could eventually be used for fast morphology estimation as a body scanner. This paper presents an extensive validation of this device for anthropometric measurements in comparison to manual measurements and stereophotogrammetric devices. The accuracy is dependent on the segment studied but the reproducibility is excellent.

The recent availability of the Kinect™ sensor, a cost-effective markerless motion capture system (MLS), offers interesting possibilities in clinical functional analysis and rehabilitation. However, neither validity nor reproducibility of this device is known yet. These two parameters were evaluated in this study. Forty-eight volunteers performed shoulder abduction, elbow flexion, hip abduction and knee flexion motions; the same protocol was repeated one week later to evaluate reproducibility. Movements were simultaneously recorded by the Kinect (with Microsoft Kinect SDK v.1.5) MLS and a traditional marker-based stereophotogrammetry system (MBS). Considering the MBS as reference, discrepancies between MLS and MBS were evaluated by comparing the range of motion (ROM) between both systems. MLS reproducibility was found to be statistically similar to MBS results for the four exercises. Measured ROMs however were found different between the systems.

Low Back Pain (LBP) is one of the most frequent pathologies related to musculoskeletal disorders. Patients suffering from LBP must perform rehabilitation exercises in order to avoid chronic disorder. At the beginning these exercises are performed with physiotherapists during rehabilitation. A major part of the treatment need to be performed at home by the patient himself. Serious Games (SG) could be used to increase patients’ motivation and to be sure that patients are performing these exercises and most important that they are doing it in the right way. A specific SG was created for LBP patients. Motions performed during the game are recorded and parameters (range of motion, speed, coupled motion) are processed. The aim of this feasibility study was to evaluate the intra and inter day reproducibility of this system to evaluate whether or not it can be used in daily practice for following the evolution of the patient and for scoring the severity of disorders.

Iris recognition has gained a lot of popularity for the last decades. Mainly a method based on binary iris templates found its way to real world use due to its simplicity, stability and reliability. The principle is that the unique iris structure is encoded to the bit code templates that are sufficient for high accuracy recognition. Encoding is performed by filtering a preprocessed iris image and storing only the phase information of the response to the filters. For years researchers used the 2D Gabor filters or their modifications, because these filters proved to provide the most reliable features. Despite the high recognition accuracy, the use of 2D Gabor filters faces a problem of spoofing. Recent studies show that the encoding process can be reverted and a spoofed iris can be obtained only based on the iris code. In this paper, we propose an efficient feature extraction method for iris recognition based on convolution kernels, learned from a database of irises. We show that the proposed method reaches state-of-the-art performance and can prohibit attackers from generating spoofed irises if the optimized convolution kernel is safely stored.