Search

Myoelectric control requires fast and stable identification of a movement from data recorded from a comfortable and straightforward system.We consider a new real-time pre-processing method applied to a single differential surface electromyogram (EMG): deconvolution, providing an estimation of the cumulative firings of motor units. A 2 channel-10 class finger movement problem has been investigated on 10 healthy subjects. We have compared raw EMG and deconvolution signals, as sources of information for two specific classifiers (based on either Support Vector Machines or k-Nearest Neighbours), with classical time-domain input features selected using Mutual Component Analysis.Using the proposed pre-processing technique, classification performances statistically improve. For example, the true positive rates of the best-tested configurations were 80.9% and 86.3% when using the EMG and its deconvoluted signal, respectively.Even considering the limited dataset and range of classification approaches investigated, our preliminary results indicate the potential usefulness of the deconvolution pre-processing.Deconvolution of EMG is a fast pre-processing that could be easily embedded in different myoelectric control applications.

Ultrasound (US) scans of inferior vena cava (IVC) are widely adopted by healthcare providers to assess patients’ volume status. Unfortunately, this technique is extremely operator dependent. Recently, new techniques have been introduced to extract stable and objective information from US images by automatic IVC edge tracking. However, these methods require prior interaction with the operator, which leads to a waste of time and still makes the technique partially subjective. In this paper, two deep learning methods, YOLO (You only look once) v4 and YOLO v4 tiny networks, commonly used for fast object detection, are applied to identify the location of the IVC and to recognise the either long or short axis view of the US scan. The output of these algorithms can be used to remove operator dependency, to reduce the time required to start an IVC analysis, and to automatically recover the vein if it is lost for a few frames during acquisition. The two networks were trained with frames extracted from 18 subjects, labeled by 4 operators. In addition, they were also trained on a linear combination of two frames that extracted information on both tissue anatomy and movement. We observed similar accuracy of the two models in preliminary tests on the entire dataset, so that YOLO v4 tiny (showing much lower computational cost) was selected for additional cross-validation in which training and test frames were taken from different subjects. The classification accuracy was approximately 88% when using original frames, but it reached 95% when pairs of frames were processed to also include information on tissue movements, indicating the importance of accounting for tissue motion to improve the accuracy of our IVC detector.

The execution or imagination of a movement is reflected by a cortical potential that can be recorded by electroencephalography (EEG) as Movement-Related Cortical Potentials (MRCPs). The identification of MRCP from a single trial is a challenging possibility to get a natural control of a Brain–Computer Interface (BCI). We propose a novel method for MRCP detection based on optimal non-linear filters, processing different channels of EEG including delayed samples (getting a spatio-temporal filter). Different outputs can be obtained by changing the order of the temporal filter and of the non-linear processing of the input data. The classification performances of these filters are assessed by cross-validation on a training set, selecting the best ones (adapted to the user) and performing a majority voting from the best three to get an output using test data. The method is compared to another state-of-the-art filter recently introduced by our group when applied to EEG data recorded from 16 healthy subjects either executing or imagining 50 self-paced upper-limb palmar grasps. The new approach has a median accuracy on the overall dataset of 80%, which is significantly better than that of the previous filter (i.e., 63%). It is feasible for online BCI system design with asynchronous, self-paced applications.

Chiari I malformation is characterized by the herniation of cerebellar tonsils below the foramen magnum. It is often accompanied by syringomyelia and neurosurgical management is still controversial. In fact, it is frequent that some symptomatic patients initially undergo bony decompression of the posterior fossa and need in a short time more invasive surgery with higher morbility (e.g., decompression of posterior fossa with dural plastic, with or without tonsillar coarctation) because of unsatisfactory results at MRI controls. This study proposes a machine learning approach (based on SVM classifier), applied to different morphometric indices estimated from sagittal MRI and some information on the patient (i.e., age and symptoms at diagnosis), to recognize patients with higher risk of syringomyelia and clinical deterioration. Our database includes 58 pediatric patients who underwent surgery treatment. A negative outcome at 1 year from the intervention was observed in 38% of them (accuracy of 62%). Our algorithm allows us to increase the accuracy to about 71%, showing it to be a valid support to neurosurgeons in refining the clinical picture.

The non-invasive estimation of right atrial pressure (RAP) would be a key advancement in several clinical scenarios, in which the knowledge of central venous filling pressure is vital for patients’ management. The echocardiographic estimation of RAP proposed by Guidelines, based on inferior vena cava (IVC) size and respirophasic collapsibility, is exposed to operator and patient dependent variability. We propose novel methods, based on semi-automated edge-tracking of IVC size and cardiac collapsibility (cardiac caval index—CCI), tested in a monocentric retrospective cohort of patients undergoing echocardiography and right heart catheterization (RHC) within 24 h in condition of clinical and therapeutic stability (170 patients, age 64 ± 14, male 45%, with pulmonary arterial hypertension, heart failure, valvular heart disease, dyspnea, or other pathologies). IVC size and CCI were integrated with other standard echocardiographic features, selected by backward feature selection and included in a linear model (LM) and a support vector machine (SVM), which were cross-validated. Three RAP classes (low < 5 mmHg, intermediate 5–10 mmHg and high > 10 mmHg) were generated and RHC values used as comparator. LM and SVM showed a higher accuracy than Guidelines (63%, 71%, and 61% for LM, SVM, and Guidelines, respectively), promoting the integration of IVC and echocardiographic features for an improved non-invasive estimation of RAP.

Self-organized criticality theory proved that information transmission and computational performances of neural networks are optimal in critical state. By using recordings of the spontaneous activity originated by dissociated neuronal assemblies coupled to Micro-Electrode Arrays (MEAs), we tested this hypothesis using Approximate Entropy (ApEn) as a measure of complexity and information transfer. We analysed 60 min of electrophysiological activity of three neuronal cultures exhibiting either sub-critical, critical or super-critical behaviour. The firing patterns on each electrode was studied in terms of the inter-spike interval (ISI), whose complexity was quantified using ApEn. We assessed that in critical state the local complexity (measured in terms of ApEn) is larger than in sub- and super-critical conditions (mean ± std, ApEn about 0.93 ± 0.09, 0.66 ± 0.18, 0.49 ± 0.27, for the cultures in critical, sub-critical and super-critical state, respectively—differences statistically significant). Our estimations were stable when considering epochs as short as 5 min (pairwise cross-correlation of spatial distribution of mean ApEn of 94 ± 5%). These preliminary results indicate that ApEn has the potential of being a reliable and stable index to monitor local information transmission in a neuronal network during maturation. Thus, ApEn applied on ISI time series appears to be potentially useful to reflect the overall complex behaviour of the neural network, even monitoring a single specific location.

Assessment of vascular size and of its phasic changes by ultrasound is important for the management of many clinical conditions. For example, a dilated and stiff inferior vena cava reflects increased intravascular volume and identifies patients with heart failure at greater risk of an early death. However, lack of standardization and sub-optimal intra- and inter- operator reproducibility limit the use of these techniques. To overcome these limitations, we developed two image-processing algorithms that quantify phasic vascular deformation by tracking wall movements, either in long or in short axis. Prospective studies will verify the clinical applicability and utility of these methods in different settings, vessels and medical conditions.

A deconvolution method is proposed for conduction block (CB) estimation based on two compound muscle action potentials (CMAPs) elicited by stimulating a nerve proximal and distal to the region in which the block is suspected. It estimates the time delay distributions by CMAPs deconvolution, from which CB is computed. The slow afterwave (SAW) is included to describe the motor unit potential, as it gives an important contribution in case of the large temporal dispersion (TD) often found in patients. The method is tested on experimental signals obtained from both healthy subjects and pathological patients, with either Chronic Inflammatory Demyelinating Polyneuropathy (CIDP) or Multifocal Motor Neuropathy (MMN). The new technique outperforms the clinical methods (based on amplitude and area of CMAPs) and a previous state-of-the-art deconvolution approach. It compensates phase cancellations, allowing to discriminate among CB and TD: estimated by the methods of amplitude, area and deconvolution, CB showed a correlation with TD equal to 39.3%, 29.5% and 8.2%, respectively. Moreover, a significant decrease of percentage reconstruction errors of the CMAPs with respect to the previous deconvolution approach is obtained (from a mean/median of 19.1%/16.7% to 11.7%/11.2%). Therefore, the new method is able to discriminate between CB and TD (overcoming the important limitation of clinical approaches) and can approximate patients’ CMAPs better than the previous deconvolution algorithm. Then, it appears to be promising for the diagnosis of demyelinating polyneuropathies, to be further tested in the future in a prospective clinical trial.

Ultrasound (US)-based measurements of the inferior vena cava (IVC) diameter are widely used to estimate right atrial pressure (RAP) in a variety of clinical settings. However, the correlation with invasively measured RAP along with the reproducibility of US-based IVC measurements is modest at best. In the present manuscript, we discuss the limitations of the current technique to estimate RAP through IVC US assessment and present a new promising tool developed by our research group, the automated IVC edge-to-edge tracking system, which has the potential to improve RAP assessment by transforming the current categorical classification (low, normal, high RAP) in a continuous and precise RAP estimation technique. Finally, we critically evaluate all the clinical settings in which this new tool could improve current practice.

The aging process is a multifaceted phenomenon that affects cognitive-affective and physical functioning as well as interactions with the environment. Although subjective cognitive decline may be part of normal aging, negative changes objectified as cognitive impairment are present in neurocognitive disorders and functional abilities are most impaired in patients with dementia. Electroencephalography-based brain–machine interfaces (BMI) are being used to assist older people in their daily activities and to improve their quality of life with neuro-rehabilitative applications. This paper provides an overview of BMI used to assist older adults. Both technical issues (detection of signals, extraction of features, classification) and application-related aspects with respect to the users’ needs are considered.

Understanding how different areas of the human brain communicate with each other is a crucial issue in neuroscience. The concepts of structural, functional and effective connectivity have been widely exploited to describe the human connectome, consisting of brain networks, their structural connections and functional interactions. Despite high-spatial-resolution imaging techniques such as functional magnetic resonance imaging (fMRI) being widely used to map this complex network of multiple interactions, electroencephalographic (EEG) recordings claim high temporal resolution and are thus perfectly suitable to describe either spatially distributed and temporally dynamic patterns of neural activation and connectivity. In this work, we provide a technical account and a categorization of the most-used data-driven approaches to assess brain-functional connectivity, intended as the study of the statistical dependencies between the recorded EEG signals. Different pairwise and multivariate, as well as directed and non-directed connectivity metrics are discussed with a pros–cons approach, in the time, frequency, and information-theoretic domains. The establishment of conceptual and mathematical relationships between metrics from these three frameworks, and the discussion of novel methodological approaches, will allow the reader to go deep into the problem of inferring functional connectivity in complex networks. Furthermore, emerging trends for the description of extended forms of connectivity (e.g., high-order interactions) are also discussed, along with graph-theory tools exploring the topological properties of the network of connections provided by the proposed metrics. Applications to EEG data are reviewed. In addition, the importance of source localization, and the impacts of signal acquisition and pre-processing techniques (e.g., filtering, source localization, and artifact rejection) on the connectivity estimates are recognized and discussed. By going through this review, the reader could delve deeply into the entire process of EEG pre-processing and analysis for the study of brain functional connectivity and learning, thereby exploiting novel methodologies and approaches to the problem of inferring connectivity within complex networks.

The influence of large vessels on near infrared spectroscopy (NIRS) measurement is generally considered negligible. Aim of this study is to test the hypothesis that changes in the vessel size, by varying the amount of absorbed NIR light, could profoundly affect NIRS blood volume indexes. Changes in haemoglobin concentration (tHb) and in tissue haemoglobin index (THI) were monitored over the basilic vein (BV) and over the biceps muscle belly, in 11 subjects (7 M – 4 F; age 31 ± 8 year) with simultaneous ultrasound monitoring of BV size. The arm was subjected to venous occlusion, according to two pressure profiles: slow (from 0 to 60 mmHg in 135 s) and rapid (0 to 40 mmHg maintained for 30 s). Both tHb and THI detected a larger blood volume increase (1.7 to 4 fold; p

Surface electromyogram (EMG) finds many applications in the non-invasive characterization of muscles. Extracting information on the control of motor units (MU) is difficult when using single channels, e.g., due to the low selectivity and large phase cancellations of MU action potentials (MUAPs). In this paper, we propose a new method to face this problem in the case of a single differential channel. The signal is approximated as a sum of convolutions of different kernels (adapted to the signal) and firing patterns, whose sum is the estimation of the cumulative MU firings. Three simulators were used for testing: muscles of parallel fibres with either two innervation zones (IZs, thus, with MUAPs of different phases) or one IZ and a model with fibres inclined with respect to the skin. Simulations were prepared for different fat thicknesses, distributions of conduction velocity, maximal firing rates, synchronizations of MU discharges, and variability of the inter-spike interval. The performances were measured in terms of cross-correlations of the estimated and simulated cumulative MU firings in the range of 0–50 Hz and compared with those of a state-of-the-art single-kernel algorithm. The median cross-correlations for multi-kernel/single-kernel approaches were 92.2%/82.4%, 98.1%/97.6%, and 95.0%/91.0% for the models with two IZs, one IZ (parallel fibres), and inclined fibres, respectively (all statistically significant differences, which were larger when the MUAP shapes were of greater difference).

A new algorithm is introduced to decompose interference surface electromyogram (EMG) recorded by a multi-channel system aligned to muscle fibers into propagating and non-propagating contributions. Muscle fiber conduction velocity (CV) is also estimated, reducing the bias induced by non-propagating components. The algorithm is fast and stable, as it is based on alignment and averaging procedures. Simulated signals (with different fat thickness, SNR, number of channels, epoch duration and force level) are used to test the algorithm. The median cross-correlation of simulated and estimated components were about 98% and 90%, for propagating and non-propagating terms, respectively. CV was estimated better than using a multi-channel maximum likelihood approach applied to double differential data (mean error of 0.08 versus 0.13 m/s), with a greater gain in case of thinner fat layer, low SNR and few channels. Example applications to experimental data are also shown (single motor unit action potential, M-wave and interference EMG). Propagating components reflect the travelling of action potentials along muscle fibers. Preliminary tests show that non-propagating contributions provide selective information on motor units firing statistics. The separation of interference EMG into propagating and non-propagating components opens new perspectives, e.g., in the study of synergies, common drive and myoelectric manifestations of fatigue.

Objectives Changes in pupil size are governed by the autonomic nervous system but may be systematically driven by voluntary shifting the gaze in depth. Thus, the pupil accommodative response (PAR) that accompanies voluntary gaze shifts from a far to a near target might be exploited as a simple human-computer interface (HCI), bypassing the somato-motor system. Here we aim to characterize PAR in quasi-natural conditions with low-cost equipment and test the possibility to use PAR as a binary communication tool. Methods Nineteen healthy subjects were instructed to voluntary switch the focus from a far to a near target upon presentation of an auditory cue. Three protocols addressed the effects of monocular/binocular vision, eye illuminance, duration of near vision, target texture and target brightness on PAR features. In a fourth protocol PAR was used to establish binary communication at different bit rates. Results PAR amplitude decreased with increasing eye illuminance and was modulated by monocular/binocular vision, duration of near vision or target texture. PAR amplitude was larger with a bright near target and a dark far target than vice-versa. However, PAR was always clearly present, regardless of the experimental manipulations. PAR-based communication performance achieved an accuracy of 100% at 10 bits/min and 96% at 15 bits/min. Significance Voluntary PAR is a robust signal, and can achieve a high communication speed when used as an HCI. This study provides a proof of concept for a PAR-based HCI, potentially useful to communicate with locked-in patients with preserved visual and autonomic functions.

The interference of artefacts with evoked scalp electroencephalogram (EEG) responses is a problem in event related brain computer interface (BCI) system that reduces signal quality and interpretability of user's intentions. Many strategies have been proposed to reduce the effects of non-neural artefacts, while the activity of neural sources that do not reflect the considered stimulation has been neglected. However discerning such activities from those to be retained is important, but subtle and difficult as most of their features are the same. We propose an automated method based on a combination of a genetic algorithm (GA) and a support vector machine (SVM) to select only the sources of interest. Temporal, spectral, wavelet, autoregressive and spatial properties of independent components (ICs) of EEG are inspected. The method selects the most distinguishing subset of features among this comprehensive fused set of information and identifies the components to be preserved. EEG data were recorded from 12 healthy subjects in a visual evoked potential (VEP) based BCI paradigm and the corresponding ICs were classified by experts to train and test the algorithm. They were contaminated with different sources of artefacts, including electromyogram (EMG), electrode connection problems, blinks and electrocardiogram (ECG), together with neural contributions not related to VEPs. The accuracy of ICs classification was about 88.5% and the energetic residual error in recovering the clean signals was 3%. These performances indicate that this automated method can effectively identify and remove main artefacts derived from either neural or non-neural sources while preserving VEPs. This could have important potential applications, contributing to speed and remove subjectivity of the cleaning procedure by experts. Moreover, it could be included in a real time BCI as a pre-processing step before the identification of the user’s intention.

This chapter provides an introduction to non-linear methods for the prediction of the concentration of air pollutants. We focused on the selection of features and the modelling and processing techniques based on the theory of Artificial Neural Networks, using Multi Layer Perceptrons and Support Vector Machines. Joint measurements of meteorological data and pollutants concentrations is useful in order to increase the number of parameters to be studied for the construction of mathematical air quality forecasting models and hence to improve forecast performances. Weather variables have a non-linear relationship with air quality, which can be captured by non-linear models such as Multi Layer Perceptrons and Support Vector Machines. Our analysis carries on the work already developed by the NeMeFo (Neural Meteo Forecasting) research project for meteorological data short-term forecasting (Pasero et al., 2004). The application provided in Section 4 illustrates how the theoretical methods for feature selection (Section 2) and data modelling (Section 3) can be implemented for the solution of a specific problem of air pollution forecast. The principal causes of air pollution are identified and the best subset of features (meteorological data and air pollutants concentrations) for each air pollutant is selected in order to predict its medium-term concentration (in particular for the PM10). The selection of the best subset of features was implemented by means of a backward selection algorithm which is based on the information theory notion of relative entropy. Multi Layer Perceptrons and Support Vector Machines constitute some of the most wide-spread statistical data-learning techniques to develop data-driven models. Their use is shown for the prediction problem considered. In conclusion, the final aim of this research is the implementation of a prognostic tool able to reduce the risk for the air pollutants concentrations to be above the alarm thresholds fixed by the law. The detection of meteorological and air pollutant data, the automatic selection of optimal descriptors of such data and the use of Multi Layer Perceptrons and Support Vector Machines are proposed as an efficient strategy to perform an accurate prediction of the time evolution of air pollutant concentration.

Aortic stiffness is an important clinical parameter for predicting cardiovascular events. Carotid–femoral pulse wave velocity (cf-PWV) has been proposed for performing this evaluation non-invasively; however, it requires dedicated equipment and experienced operators. We explored the possibility of measuring aortic stiffness using ultrasound scans of the abdominal aorta coupled with the Bramwell–Hill equation. Healthy subjects were investigated; measurements of cf-PWV were taken by arterial tonometry and aortic systo-diastolic pressure difference was estimated using a validated model. Pulsatility of an abdominal tract of aorta was assessed by automated processing of ultrasound scans. Through a Bland–Altmann analysis, we found large biases when estimating each parameter by applying the Bramwell–Hill equation to the measured values of the other two paramters (bias, ± 1.96 SD; PWV, about 2.1 ± 2.5 m/s; pulsatility, 12 ± 14%; pressure jump, 47 ± 55 mmHg). These results indicate that the two measures are not interchangeable, and that a large part of the bias is attributable to blood pressure estimation. Further studies are needed to identify the possible sources of bias between cf-PWV and aortic pulsatility.

Sensors 22(11), 4150 (2022). doi:10.3390/s22114150 special issue: "Special Issue "Advances in Bipolar and Array-Based Surface EMG: Detection, Interpretation and Teaching" / Special Issue Editors: Prof. Dr. Roberto Merletti, Guest Editor; Dr. Isabella Campanini; Guest Editor; Prof. Dr. Catherine Disselhorst-Klug, Guest Editor", Published by MDPI, Basel

The herniation of cerebellum through the foramen magnum may block the normal flow of cerebrospinal fluid determining a severe disorder called Chiari I Malformation (CM-I). Different surgical options are available to help patients, but there is no standard to select the optimal treatment. This paper proposes a fully automated method to select the optimal intervention. It is based on morphological parameters of the brain, posterior fossa and cerebellum, estimated by processing sagittal magnetic resonance images (MRI). The processing algorithm is based on a non-rigid registration by a balanced multi-image generalization of demons method. Moreover, a post-processing based on active contour was used to improve the estimation of cerebellar hernia. This method allowed to delineate the boundaries of the regions of interest with a percentage of agreement with the delineation of an expert of about 85%. Different features characterizing the estimated regions were then extracted and used to develop a classifier to identify the optimal surgical treatment. Classification accuracy on a database of 50 patients was about 92%, with a predictive value of 88% (tested with a leave-one-out approach).

A new approach is introduced for non-rigid registration of a pair of magnetic resonance images (MRI). It is a generalization of the demons algorithm with low computational cost, based on local information augmentation (by integrating multiple images) and balanced implementation. Specifically, a single deformation that best registers more pairs of images is estimated. All these images are extracted by applying different operators to the two original ones, processing local neighbors of each pixel. The following five images were found to be appropriate for MRI registration: the raw image and those obtained by contrast-limited adaptive histogram equalization, local median, local entropy and phase symmetry. Thus, each local point in the images is supplemented by augmented information coming by processing its neighbor. Moreover, image pairs are processed in alternation for each iteration of the algorithm (in a balanced way), computing both a forward and a backward registration. The new method (called balanced multi-image demons) is tested on sagittal MRIs from 10 patients, both in simulated and experimental conditions, improving the performances over the classical demons approach with minimal increase of the computational cost (processing time around twice that of standard demons). Specifically, a simulated deformation was applied to the MRIs (either original or corrupted by additive Gaussian or speckle noises). In all tested cases, the new algorithm improved the estimation of the simulated deformation (squared estimation error decreased by about 65% in the average). Moreover, statistically significant improvements were obtained in experimental tests, in which different brain regions (i.e., brain, posterior fossa and cerebellum) were identified by the atlas approach and compared to those manually delineated (in the average, Dice coefficient increased of about 6%). The conclusion is that a balanced method applied to multiple information extracted from neighboring pixels is a low cost approach to improve registration of MRIs.

The pulsatility of the inferior vena cava (IVC) reflects the volume status of patients. It can be investigated by ultrasounds (US), offering an important non-invasive tool supporting fluid management. However, the method has limitations attributable to many confounding factors, e.g., related to IVC movements and non-regular shapes. Short- or long-axis views have been used, both having advantages and limitations in counteracting such confounding factors, depending on the specific condition. The aim of this study is to investigate IVC pulsatility in the different directions on the transverse plane and to assess its variability. Moreover, different components of this pulsatility (induced by either respiratory or cardiac activity) are investigated. The method is tested on 10 healthy patients, with large variations across them of IVC section (mean diameters in the range 1 cm to 3 cm), shape and pulsatility (average caval index [CI] ranging from approximately 20% to 70%). The average coefficient of variation of the CI estimated on 10 different directions was 13% (21% and 20% for the respiratory and cardiac components, respectively), with a range that was approximately 50% of the mean CI across different directions (approximately the same for the 2 different components). The minimum and maximum CI were found close to the directions of maximum and minimum IVC diameter, respectively. The investigation of IVC dynamics in the entire cross-section is crucial to obtain a more repeatable and reliable characterization of IVC pulsatility. The calculation of a CI based on the "equivalent" diameter (proportional to the square root of the IVC cross-sectional area) is encouraged.

Slow biphasic complexes (SBC) have been identified in the EEG of patients suffering for inflammatory brain diseases. Their amplitude, location and frequency of appearance were found to correlate with the severity of encephalitis. Other characteristics of SBCs and of EEG traces of patients could reflect the grade of pathology. Here, EEG rhythms are investigated together with SBCs for a better characterization of encephalitis. EEGs have been acquired from pediatric patients: ten controls and ten encephalitic patients. They were split by neurologists into five classes of different severity of the pathology. The relative power of EEG rhythms was found to change significantly in EEGs labeled with different severity scores. Moreover, a significant variation was found in the last seconds before the appearance of an SBC. This information and quantitative indexes characterizing the SBCs were used to build a binary classification decision tree able to identify the classes of severity. True classification rate of the best model was 76.1% (73.5% with leave-one-out test). Moreover, the classification errors were among classes with similar severity scores (precision higher than 80% was achieved considering three instead of five classes). Our classification method may be a promising supporting tool for clinicians to diagnose, assess and make the follow-up of patients with encephalitis.

Surface electromyogram (EMG) has a relatively large pick-up volume, reflecting the activity of muscle tissue placed quite far from the electrodes. This could be beneficial when the global muscle activity is of interest, but it is a limitation when selective information is needed. The EMG from muscles that are neighbors of the one of interest is called crosstalk. Its interpretation, identification, quantification and removal have been the objectives of many works in the literature. However, it is still considered as an open problem, with effects that are difficult to predict. In this paper, the problem of crosstalk is discussed and the main literature is reviewed. Finally, a few recent techniques are introduced that are potentially relevant to quantify or reduce it.

Background In clinical practice, as stated in the ASE guidelines, the echocardiographic estimation of right atrial pressure (RAP) is based on the size of the inferior vena cava (IVC) and its inspiratory collapse. However, this method has proven to have limits of reliability and reproducibility. The use of a recently developed software that with a semi-automatic technique highlight the edges of the IVC could help to standardize the echocardiographic assessment of RAP. Aim of the study: The aim of the study was to assess feasibility and accuracy of a new semi-automated approach to estimate the RAP. Standard acquired echocardiographic images were processed with a semi-automatic technique, indexes related to the collapsibility of the vessel during inspiration (Caval Index, CI), during the whole respiratory cycle (Respiratory Caval Index, RCI) and through the heart cycle transmitted movements’ (Cardiac Caval Index (CCI) were derived (figure 1). Using these indexes, we developed two models: a) the Binary Tree Model (BTM), further divided in BTM3 and BTM5 (RAP estimated in 3 and 5 classes, respectively); b) the Regression Model (RM), further divided in RM linear (continuous model) and RM3 and RM5 (RAP estimated in 3 and 5 classes respectively). RAP assessed using these innovative techniques were compared with two standard estimation (SE) echocardiographic methods A and B. Direct RAP measurements obtained during a right heart catheterization (RHC), performed within 6 hours, were used as reference. Results 62 consecutive ‘all-comers’ patients that had a RHC were enrolled; 13 patients were excluded for technical reasons. Therefore 49 patients were included in this study (26 males and 23 females; mean age of 62.2 ± 15.2 years, 75.5% pulmonary hypertension, 34.7% severe left ventricular dysfunction and 51% right ventricular dysfunction). The two SE methods showed poor accuracy for RAP estimation (method A: ME = 51%, R2= 0.22; method B: ME = 69%, R2= 0.26). Instead, the new semi-automatic methods BTM3 and BTM5 based on parameters derived from IVC edge tracking (mean IVC diameter, CI, CCI and RCI) had a misclassification error of only 14% (R2 = 0.47) and 22% (R2 = 0.61), respectively, to classify RAP. The accuracy was lower for RM than BTM (RM3: ME = 61%, R2 = 0.39; RM5: ME = 55%, R2 = 0.39). However, the RM showed the lowest mean bias in estimating RAP: 0.23 [-8.34; 8.81] mmHg. Conclusions A multi-parametric approach using the new indexes, such as CCI and RCI, derived from a semi-automated edge tracking of the IVC is a promising tool for a more accurate estimation of RAP. This study proposes an innovative method for the non-invasive estimation of the RAP, which requires confirmation on larger population. Abstract P892 Figure 1

The echocardiographic estimation of right atrial pressure (RAP) is based on the size and inspiratory collapse of the inferior vena cava (IVC). However, this method has proven to have limits of reliability. The aim of this study is to assess feasibility and accuracy of a new semi-automated approach to estimate RAP. Standard acquired echocardiographic images were processed with a semi-automated technique. Indexes related to the collapsibility of the vessel during inspiration (Caval Index, CI) and new indexes of pulsatility, obtained considering only the stimulation due to either respiration (Respiratory Caval Index, RCI) or heartbeats (Cardiac Caval Index, CCI) were derived. Binary Tree Models (BTM) were then developed to estimate either 3 or 5 RAP classes (BTM3 and BTM5) using indexes estimated by the semi-automated technique. These BTMs were compared with two standard estimation (SE) echocardiographic methods, indicated as A and B, distinguishing among 3 and 5 RAP classes, respectively. Direct RAP measurements obtained during a right heart catheterization (RHC) were used as reference. 62 consecutive ‘all-comers’ patients that had a RHC were enrolled; 13 patients were excluded for technical reasons. Therefore 49 patients were included in this study (mean age 62.2 ± 15.2 years, 75.5% pulmonary hypertension, 34.7% severe left ventricular dysfunction and 51% right ventricular dysfunction). The SE methods showed poor accuracy for RAP estimation (method A: misclassification error, ME = 51%, R2 = 0.22; method B: ME = 69%, R2 = 0.26). Instead, the new semi-automated methods BTM3 and BTM5 have higher accuracy (ME = 14%, R2 = 0.47 and ME = 22%, R2 = 0.61, respectively). In conclusion, a multi-parametric approach using IVC indexes extracted by the semi-automated approach is a promising tool for a more accurate estimation of RAP.

Surface electromyogram (EMG) has a relatively large detection volume, so that it could include contributions both from the target muscle of interest and from nearby regions (i.e., crosstalk). This interference can prevent a correct interpretation of the activity of the target muscle, limiting the use of surface EMG in many fields. To counteract the problem, selective spatial filters have been proposed, but they reduce the representativeness of the data from the target muscle. A better solution would be to discard only crosstalk from the signal recorded in monopolar configuration (thus, keeping most information on the target muscle). An inverse modelling approach is here proposed to estimate the contributions of different muscles, in order to focus on the one of interest. The method is tested with simulated monopolar EMGs from superficial nearby muscles contracted at different force levels (either including or not model perturbations and noise), showing statistically significant improvements in information extraction from the data. The median over the entire dataset of the mean squared error in representing the EMG of the muscle under the detection electrode was reduced from 11.2% to 4.4% of the signal energy (5.3% if noisy data were processed); the median bias in conduction velocity estimation (from 3 monopolar channels aligned to the muscle fibres) was decreased from 2.12 to 0.72 m/s (1.1 m/s if noisy data were processed); the median absolute error in the estimation of median frequency was reduced from 1.02 to 0.67 Hz in noise free conditions and from 1.52 to 1.45 Hz considering noisy data.

Assessment of volume status is important to correctly plan the treatment of patients admitted and managed by cardiology, emergency and internal medicine departments. Non-invasive assessment of volume status by echography of the inferior vena cava (IVC) is a promising possibility, but its clinical use is limited by poor reproducibility of current standard procedures. We have developed new algorithms to extract reliable information from non-invasive IVC monitoring by ultrasound (US) imaging. Both long and short axis US B-mode video-clips were taken from 50 patients, in either hypo-, eu-, or hyper-volemic conditions. The video-clips were processed to extract static and dynamic indexes characterizing the IVC behaviour. Different binary tree models (BTM) were developed to identify patient conditions on the basis of those indexes. The best classifier was a BTM using IVC pulsatility indexes as input features. Its accuracy (78.0% when tested with a leave-one-out approach) is superior to that achieved using indexes measured by the standard clinical method from M-mode US recordings. These results were obtained with patients in conditions of normal respiratory function and cardiac rhythm. Further studies are necessary to extend this approach to patients with more complex cardio-respiratory conditions.

Background: Freezing of Gait (FOG) is one of the most disabling motor complications of Parkinson’s disease, and consists of an episodic inability to move forward, despite the intention to walk. FOG increases the risk of falls and reduces the quality of life of patients and their caregivers. The phenomenon is difficult to appreciate during outpatients visits; hence, its automatic recognition is of great clinical importance. Many types of sensors and different locations on the body have been proposed. However, the advantages of a multi-sensor configuration with respect to a single-sensor one are not clear, whereas this latter would be advisable for use in a non-supervised environment. Methods: In this study, we used a multi-modal dataset and machine learning algorithms to perform different classifications between FOG and non-FOG periods. Moreover, we explored the relevance of features in the time and frequency domains extracted from inertial sensors, electroencephalogram and skin conductance. We developed both a subject-independent and a subject-dependent algorithm, considering different sensor subsets. Results: The subject-independent and subject-dependent algorithms yielded accuracies of 85% and 88% in the leave-one-subject-out and leave-one-task-out test, respectively. Results suggest that the inertial sensors positioned on the lower limb are generally the most significant in recognizing FOG. Moreover, the performance impairment experienced when using a single tibial accelerometer instead of the optimal multi-modal configuration is limited to 2–3%. Conclusions: The achieved results disclose the possibility of getting a good FOG recognition using a minimally invasive set-up made of a single inertial sensor. This is very significant in the perspective of implementing a long-term monitoring of patients in their homes, during activities of daily living.

The electroencephalogram (EEG) of patients suffering from inflammatory diseases of the brain may show specific waveforms called slow biphasic complexes (SBC). Recent studies indicated a correlation between the severity of encephalitis and some features of SBCs, such as location, amplitude and frequency of appearance. Moreover, EEG rhythms were found to vary before the onset of an SBC, as if the brain was preparing to the discharge (actually with a slowing down of the EEG oscillation). Here, we investigate possible variations of EEG functional connectivity (FC) in EEGs from pediatric patients with different levels of severity of encephalitis. FC was measured by the maximal crosscorrelation of EEG rhythms in different bipolar channels. Then, the indexes of network patterns (namely strength, clustering coefficient, efficiency and characteristic path length) were estimated to characterize the global behavior when they are measured during SBCs or far from them. EEG traces showed statistical differences in the two conditions: clustering coefficient, efficiency and strength are higher close to an SBC, whereas the characteristic path length is lower. Moreover, for more severe conditions, an increase in clustering coefficient, efficiency and strength and a decrease in characteristic path length were observed in the delta–theta band. These outcomes support the hypothesis that SBCs result from the anomalous coordination of neurons in different brain areas affected by the inflammation process and indicate FC as an additional key for interpreting the EEG in encephalitis patients.

Objectives Respirophasic variation of inferior vena cava (IVC) size is affected by large variability with spontaneous breathing. This study aims at characterizing the dependence of IVC size on controlled changes in intrathoracic pressure. Methods Ten healthy subjects, in supine position, performed controlled isovolumetric respiratory efforts at functional residual capacity, attaining positive (5, 10, and 15 mmHg) and negative (−5, −10, and −15 mmHg) alveolar pressure levels. The isovolumetric constraint implies that equivalent changes are exhibited by alveolar and intrathoracic pressures during respiratory tasks. Results The IVC cross-sectional area equal to 2.88 ± 0.43 cm2 at baseline (alveolar pressure = 0 mmHg) was progressively decreased by both expiratory and inspiratory efforts of increasing strength, with diaphragmatic efforts producing larger effects than thoracic ones: −55 ± 15% decrease, at +15 mmHg of alveolar pressure (P

Chiari I Malformation (CM-I) is a disorder characterized by herniation of the cerebellar tonsils through the foramen magnum. The choice of the optimal surgical management of CM-I is still an open problem, as there are no guidelines for the most suitable approach in symptomatic patients. We assessed the possibility of predicting the optimal intervention on the basis of morphological properties of the brain, posterior fossa and cerebellum. The boundaries of these regions were estimated by non-rigid registration of test sagittal magnetic resonance images (MRI) over an atlas. Different features characterizing the estimated regions were extracted and used to develop Naive Bayes classifiers. Classification was tested on a database of 30 patients, which were either healthy or pathological, with either mild or severe manifestation of CM-I. The best cross-validated classifier selected input features related to the shape and relative dimension of different brain regions (specifically, cerebellum and posterior fossa), the possible presence and dimension of the hernia and the regularity of the boundary of the cerebellum. The best classification accuracy (tested with a leave-one-out approach) was about 83%. These promising results suggest investigating further the possibility of defining the surgical treatment of CM-I on the basis of morphometric parameters from sagittal MRIs.

Interference surface electromyogram (EMG) reflects many bioelectric properties of active motor units (MU), which are however difficult to estimate due to the asynchronous summation of their discharges. This paper introduces a deconvolution technique to estimate the cumulative firings of MUs. Tests in simulations show that the power spectral density of the estimated MU firings has a low-frequency peak corresponding to the mean firing rate of MUs in the detection volume of the recording system, weighted by the amplitudes of MU action potentials. The peak increases in amplitude and its centroid shifts to a higher frequency when MU synchronization is simulated (mainly due to the shift of discharges of large MUs). The peak is found even at high force levels, when such a contribution does not emerge from the EMG. This result is also confirmed in preliminary applications to experimental data. Moreover, the simulated cumulative firings of MUs are estimated with a correlation above 90% (considering frequency contributions up to 150 Hz), for all force levels. The method requires a single EMG channel, thus being feasible even in applied studies using simple recording systems. It may open many potential applications, e.g., in the study of the modulation of MU firing rate induced by either fatigue or pathology and in coherency analysis. Graphical Abstract Examples of application of the deconvolution (Deconv) algorithm and comparison with the cumulative firings and the cumulated weighted firings (CWF, i.e., each firing pattern is weighted by the root mean squared amplitude of the corresponding MU action potential). Portions of data are shown on the left, the power spectral densities (PSD) on the right (Welch method applied to 3 s of data, sub-epochs of 0.5 s, mean value removed from each of them, 50% of overlap). A) Simulated signal (50% of maximal voluntary contraction, MVC) with random MU firings. B) Simulated signal (50% MVC) with a level of synchronization equal to 10%. C) Experimental data from vastus medialis at 40% MVC (data decomposed by the algorithm of Holobar and Zazula, IEEE Trans. Sig. Proc. 2007; PSD of the cumulated firings almost identical to that of CWF, as few MUs were identified).

Inflammatory brain diseases have been associated to the presence of slow biphasic complexes (SBC)in the EEG. An automated method has been developed to identify them. A dataset of 128 EEGs was recorded from pediatric controls and patients showing encephalitis with different levels of severity. Experts assigned a severity score with 5 levels to each trace, considering both electrophysiological and clinical manifestations. The number, amplitude and location of SBCs were used to identify automatically the severity scores using a binary classification decision tree. True classifications have been obtained in the 64.1 % of cases (55.6% using a leave-one-out approach)and misclassifications were among close severity scores (about 80 % of accuracy was obtained considering 3 instead of 5 severity scores), indicating that SBCs may support the identification, assessment and follow-up of encephalitis.

Pulsatility of the inferior vena cava (IVC) provides information on volume status in healthy subjects and in many clinical conditions. The ultrasound (US) approach to estimating the caval index (CI) is not standardized, as it is operator dependent and vulnerable to measurement errors because of different factors, including movements of the IVC and non-uniform IVC pulsatility along its longitudinal axis. We propose and test in healthy subjects an innovative automated approach, which tracks the IVC movements registered in a B-mode US video clip and estimates the pulsatility of an entire portion of the vein rather than of a single arbitrary section. Large variations in CI estimates were observed along the longitudinal axis (in the worst case, CI ranged between 15% and 60%), indicating the importance of investigating a whole portion of the vessel.

The pulsatility of the inferior vena cava (IVC) reflects the volume status and central venous pressure of patients. The standard clinical indicator of IVC pulsatility is the caval index (CI), measured from ultrasound recordings. However, its estimation is not standardized and is vulnerable to artifacts, mostly because of IVC movements during respiration. Thus, we used a (recently patented) semi-automated method that tracks IVC movements and averages the CI across an entire section of the vein, which provides a more stable indication of pulsatility. This algorithm was used to estimate the CI, pulsatility indicators reflecting either respiratory or cardiac stimulation and the mean diameter of the IVC. These IVC indices, together with anthropometric information, were used as potential features to build an innovative model for the estimation of the right atrial pressure (RAP) recorded from 49 catheterized patients. An exhaustive search was carried out for the best among all possible models that could be obtained by using combinations of these features. The model with minimum estimation error (tested with a leave-one-out approach) was selected. This model estimated RAP with an error of about 3.6 ± 2.6 mm Hg (mean ± standard deviation); the error when using only operator measured variables, without software, was about 4.0 ± 2.5 mm Hg. These promising results underline the need for further study of our RAP estimation method on a larger data set.

The inferior vena cava (IVC) shows variations of cross section over time (pulsatility) induced by different stimulations (e.g., breathing and heartbeats). Pulsatility is affected by patients’ volume status and can be investigated by ultrasound (US) measurements. An index of IVC pulsatility based on US visualization and called caval index (CI) was proposed as a non-invasive indirect measurement of the volume status. However, its estimation is not standardized, operator dependent and affected by movements of the vein and non-uniform pulsatility. We introduced a software that processes B-mode US video clips to track IVC movements and estimate CI on an entire portion of the vein. This method is here compared to the standard approach in terms of repeatability of the estimated CI, reporting on the variability over different respiratory cycles, longitudinal IVC sections and intra-/inter-observers. Our method allows to reduce the variability of CI assessment, making a step toward its standardization.

Interference surface electromyogram (EMG) recorded from linear electrode arrays aligned to muscle fibres can be separated into propagating and non-propagating contributions. The first reflects the propagation of action potentials along muscle fibres. The non-propagating components are here shown to be related to the overall firing pattern of active motor units (MU). Indeed, in simulations, the power spectral density (PSD) of non-propagating components shows a low frequency peak corresponding to the mean firing rate, even when such a contribution is not visible in the PSD of the EMG (either monopolar or single differential configuration, either rectified or not). Moreover, it has a high correlation with the PSD of the cumulative firings of the MUs in the detection volume of the recording system. Applications to experimental data confirm that the low frequency peak is more evident for the non-propagating components than for the raw signals and is related to the MU firing frequency. Potential future applications are expected in the study of the MU control in different conditions (e.g., training, fatigue or pathology, inducing changes, or modulation of firing rate) and in the investigation of common synaptic inputs to motor neurons.

There are many cases in which the separation of different sources from single channel recordings is important, for example, in fluorescence spectral overlap compensation, electrical impedance signaling, intramuscular electromyogram decomposition or in the case of spike sorting of neuron potentials from microelectrode arrays (MEA). Focusing on the latter, the problem can be faced by identifying spikes emerging from the background and clustering into different groups, indicating the activity of different neurons. Problems are found when more spikes are superimposed in overlapped waveforms. We discuss the application of Biogeography-Based Optimization (BBO) to resolve this specific problem. Our algorithm is compared with three spike-sorting methods (SpyKING Circus, Common Basis Pursuit and Klusta), showing statistically better performance (in terms of F1 score, True Positive Rate—TPR and Positive Predictive Value—PPV) in resolving overlaps in realistic, simulated data. Specifically, BBO showed median F1, TPR and PPV of 100%, 100% and about 75%, respectively, considering a simulated noise with the same spectral density as the experimental one and a similar power with highly statistically significant improvements of at least two performance indexes over each of the other three tested algorithms.

Biomedical image processing is an interdisciplinary field [...]

Spontaneous activity of neural networks depends on their stage of development. Computational performances of a network increase when the maturation leads to a self-organized criticality. Thus, an increasing complexity in the behavior of the network is expected when it enters in this developmental stage, called critical state. We tested this hypothesis investigating with a Micro-Electrodes Array of 60 electrodes a neuronal culture that during maturation exhibited first a subcritical and then a critical state. We found that in the critical state the local complexity (measured in terms of Approximate Entropy) was larger than in subcritical conditions ($\mathbf{mean}\pm \mathbf{std}$, ApEn about $\mathbf{1.03}+\mathbf{0.10},\mathbf{0.77}+\mathbf{0.18}$ in critical and sub-critical states, respectively; differences statistically significant), but only if the embedding dimension is at least 3 and the tolerance is fixed (we considered it equal to 1 ms, which is close to the characteristic time of neural communications).

Objective Crosstalk can pose limitations to the applications of surface electromyogram (EMG). Its reduction can help in the identification of the activity of specific muscles. The selectivity of different spatial filters was tested in the literature both in simulations and experiments: their performances are affected by many factors (e.g. anatomy, conduction properties of the tissues and dimension/location of the electrodes); moreover, they reduce crosstalk by decreasing the detection volume, recording data that represent only the activity of a small portion of the muscle of interest. In this study, an alternative idea is proposed, based on a spatio-temporal filter. Approach An adaptive method is applied, which filters both in time and among different channels, providing a signal that maximally preserves the energy of the EMG of interest and discards that of nearby muscles (increasing the signal to crosstalk ratio, SCR). Main results Tests with simulations and experimental data show an average increase of the SCR of about 2 dB with respect to the single or double differential data processed by the filter. This allows to reduce the bias induced by crosstalk in conduction velocity and force estimation. Significance The method can be applied to few channels, so that it is useful in applicative studies (e.g. clinics, gate analysis, rehabilitation protocols with EMG biofeedback and prosthesis control) where limited and not selective information is usually available.

Non-linear analysis found many applications in biomedicine. Approximate entropy (ApEn) is a popular index of complexity often applied to biomedical data, as it provides quite stable indications when processing short and noisy epochs. However, ApEn strongly depends on parameters, which were chosen in the literature in wide ranges. This paper points out that ApEn depends on sampling rate of continuous time signals, embedding dimension, tolerance (under which a match is identified), epoch duration and low frequency trends. Moreover, contradicting results can be obtained changing parameters. This was found both in simulations and in experimental EEG. These limitations of ApEn suggest the introduction of an alternative index, here called modified ApEn, which is based on the following principles: oversampling is compensated, self-recurrences are ignored, a fixed percentage of recurrences is selected and low frequency trends are removed. The modified index allows to get more stable measurements of the complexity of the tested simulated data and EEG. The final conclusions are that, in order to get a reliable estimation of complexity using ApEn, parameters should be chosen within specific ranges, data must be sampled close to the Nyquist limit and low frequency trends should be removed. Following these indications, different studies could be more easily compared, interpreted and replicated. Moreover, the modified ApEn can be an interesting alternative, which extends the range of parameters for which stable indications can be achieved.

Background and ObjectiveHeartbeats Holter monitoring is important for the detection of arrhythmias and possible anomalies, which are predictive of cardiovascular risks and infections. Reducing the number of acquired samples is useful to save energy and memory, but a proper down-sampling schedule is needed to record all useful information.MethodAn adaptive algorithm for the non-uniform down-sampling of data is used to reduce the mean sampling frequency of ECG data. The acquired data are processed to extract RR rhythm and to classify the heartbeats among a set of possible types of arrhythmias. ResultsThe proposed method is tested in terms of the ability to estimate the heart rate and to classify the heartbeats from the MIT-BIH Arrhythmia data down-sampled below the Nyquist limit. The mean accuracy in identifying the heartbeats was over the 98% and the RMS error in estimating the RR time series was lower than the 1%. Variability, spectral and complexity indexes extracted from RR series were estimated with a mean error that was lower than 10%. Classification accuracy was above the 95%. ConclusionsAn adaptive method to down-sample ECG data is discussed. It can be useful to save energy and to reduce memory occupation, while still preserving important information on the heartbeats. Heartbeats monitoring is important to predict cardiovascular risks and infections.Down-sampling allows to save energy and memory.The discussed algorithm down-samples ECG to about 60Hz.RR series and its variability, spectrum and complexity can be investigated.Recorded ECG waveform can be classified among different types of arrhythmias.