Search

Abstract This data descriptor describes the Roessingh Research & Development-MyLeg database for activity prediction (MyPredict), containing three data sets. These data sets contain data from 55 able-bodied subjects, mean age 24 ± 2 years, measured in 85 measurement sessions. Measurement sessions consisted of trials containing sitting, standing, overground walking, stair ascent, stair descent, ramp ascent, ramp descent, walking on uneven terrain and walking in simulated confined spaces. Subjects were measured using eight inertial measurement units in combination with different types of sEMG. Recorded kinematics consisted of joint angles, sensor accelerations, angular velocity, orientation and virtual marker positions. sEMG was recorded using bipolar sEMG, multi-array sEMG or a combination of both. All data showed excellent correlation with other online available data sets. The data reported in this descriptor forms a solid basis for research into myoelectric pattern recognition, myoelectric control development and electromyography to be used in data-driven applications.

Esophageal cancer can be treated effectively with esophagectomy; however, the postoperative complication rate is high. In this paper, we study to what extent machine learning methods can predict anastomotic leakage and pneumonia up to two days in advance. We use a dataset with 417 patients who underwent esophagectomy between 2011 and 2021. The dataset contains multimodal temporal information, specifically, laboratory results, vital signs, thorax images, and preoperative patient characteristics. The best models scored mean test set AUROCs of 0.87 and 0.82 for leakage 1 and 2 days ahead, respectively. For pneumonia, this was 0.74 and 0.61 for 1 and 2 days ahead, respectively. We conclude that machine learning models can effectively predict anastomotic leakage and pneumonia after esophagectomy.

Abstract Background Patient-reported outcome measurements (PROMs) are commonly used in clinical practice to support clinical decision making. However, few studies have investigated machine learning methods for predicting PROMs outcomes and thereby support clinical decision making. Objective This study investigates to what extent different machine learning methods, applied to two different PROMs datasets, can predict outcomes among patients with non-specific neck and/or low back pain. Methods Using two datasets consisting of PROMs from (1) care-seeking low back pain patients in primary care who participated in a randomized controlled trial, and (2) patients with neck and/or low back pain referred to multidisciplinary biopsychosocial rehabilitation, we present data science methods for data prepossessing and evaluate selected regression and classification methods for predicting patient outcomes. Results The results show that there is a potential for machine learning to predict and classify PROMs. The prediction models based on baseline measurements perform well, and the number of predictors can be reduced, which is an advantage for implementation in decision support scenarios. The classification task shows that the dataset does not contain all necessary predictors for the care type classification. Overall, the work presents generalizable machine learning pipelines that can be adapted to other PROMs datasets. Conclusion This study demonstrates the potential of PROMs in predicting short-term patient outcomes. Our results indicate that machine learning methods can be used to exploit the predictive value of PROMs and thereby support clinical decision making, given that the PROMs hold enough predictive power

To enable virtual reality exposure therapy (VRET) that treats anxiety disorders by gradually exposing the patient to fear using virtual reality (VR), it is important to monitor the patient's fear levels during the exposure. Despite the evidence of a fear circuit in the brain as reflected by functional near-infrared spectroscopy (fNIRS), the measurement of fear response in highly immersive VR using fNIRS is limited, especially in combination with a head-mounted display (HMD). In particular, it is unclear to what extent fNIRS can differentiate users with and without anxiety disorders and detect fear response in a highly ecological setting using an HMD. In this study, we investigated fNIRS signals captured from participants with and without a fear of height response. To examine the extent to which fNIRS signals of both groups differ, we conducted an experiment during which participants with moderate fear of heights and participants without it were exposed to VR scenarios involving heights and no heights. The between-group statistical analysis shows that the fNIRS data of the control group and the experimental group are significantly different only in the channel located close to right frontotemporal lobe, where the grand average oxygenated hemoglobin Δ[HbO] contrast signal of the experimental group exceeds that of the control group. The within-group statistical analysis shows significant differences between the grand average Δ[HbO] contrast values during fear responses and those during no-fear responses, where the Δ[HbO] contrast values of the fear responses were significantly higher than those of the no-fear responses in the channels located towards the frontal part of the prefrontal cortex. Also, the channel located close to frontocentral lobe was found to show significant difference for the grand average deoxygenated hemoglobin contrast signals. Support vector machine-based classifier could detect fear responses at an accuracy up to 70% and 74% in subject-dependent and subject-independent classifications, respectively. The results demonstrate that cortical hemodynamic responses of a control group and an experimental group are different to a considerable extent, exhibiting the feasibility and ecological validity of the combination of VR-HMD and fNIRS to elicit and detect fear responses. This research thus paves a way toward the a brain-computer interface to effectively manipulate and control VRET.

Pattern recognition in EMG-based control systems suffer from increase in error rate over time, which could lead to unwanted behavior. This so-called concept drift in myoelectric control systems could be caused by fatigue, sensor replacement and varying skin conditions. To circumvent concept drift, adaptation strategies could be used to retrain a pattern recognition system, which could lead to comparable error rates over multiple days. In this study, we investigated the error rate development over one week and compared three adaptation strategies to reduce the error rate increase. The three adaptation strategies were based on entropy, on backward prediction and a combination of backward prediction and entropy. Ten able-bodied subjects were measured on four measurement days while performing gait-related activities. During the measurement electromyography and kinematics were recorded. The three adaptation strategies were implemented and compared against the baseline error rate and against adaptation using the ground truth labels. It can be concluded that without adaptation the baseline error rate increases significantly from day 1 to 2, but plateaus on day 2, 3 and 7. Of the three tested adaptation strategies, entropy based adaptation showed the smallest increase in error rate over time. It can be concluded that entropy based adaptation is simple to implement and can be considered a feasible adaptation strategy for lower limb pattern recognition.

At vital moments in professional soccer matches, penalties were often missed. Psychological factors, such as anxiety and pressure, are among the critical causes of the mistakes, commonly known as choking under pressure. Nevertheless, the factors have not been fully explored. In this study, we used functional near-infrared spectroscopy (fNIRS) to investigate the influence of the brain on this process. An in-situ study was set-up (N = 22), in which each participant took 15 penalties under three different pressure conditions: without a goalkeeper, with an amiable goalkeeper, and with a competitive goalkeeper. Both experienced and inexperienced soccer players were recruited, and the brain activation was compared across groups. Besides, fNIRS activation was compared between sessions that participants felt anxious against sessions without anxiety report, and between penalty-scoring and -missing sessions. The results show that the task-relevant brain region, the motor cortex, was more activated when players were not experiencing performance anxiety. The activation of task-irrelevant areas was shown to be related to players experiencing anxiety and missing penalties, especially the prefrontal cortex (PFC). More particularly, an overall higher activation of the PFC and an increase of PFC lateral asymmetry were related to anxious players and missed penalties, which can be caused by players' worries about the consequences of scoring or missing the penalty kicks. When experienced players were feeling anxious, their left temporal cortex activation increased, which could be an indication that experienced overthink the situation and neglect their automated skills. Besides, the left temporal cortex activation is higher when inexperienced players succeeded to score a penalty. Overall, the results of this study are in line with the neural efficiency theory and demonstrate the feasibility and ecological validity to detect neurological clues relevant to anxiety and performance from fNIRS recordings in the field.

A lot of research has been done on the detection of mental workload (MWL) using various bio-signals. Recently, deep learning has allowed for novel methods and results. A plethora of measurement modalities have proven to be valuable in this task, yet studies currently often only use a single modality to classify MWL. The goal of this research was to classify perceived mental workload (PMWL) using a deep neural network (DNN) that flexibly makes use of multiple modalities, in order to allow for feature sharing between modalities. To achieve this goal, an experiment was conducted in which MWL was simulated with the help of verbal logic puzzles. The puzzles came in five levels of difficulty and were presented in a random order. Participants had 1 h to solve as many puzzles as they could. Between puzzles, they gave a difficulty rating between 1 and 7, seven being the highest difficulty. Galvanic skin response, photoplethysmograms, functional near-infrared spectrograms and eye movements were collected simultaneously using LabStreamingLayer (LSL). Marker information from the puzzles was also streamed on LSL. We designed and evaluated a novel intermediate fusion multimodal DNN for the classification of PMWL using the aforementioned four modalities. Two main criteria that guided the design and implementation of our DNN are modularity and generalisability. We were able to classify PMWL within-level accurate (0.985 levels) on a seven-level workload scale using the aforementioned modalities. The model architecture allows for easy addition and removal of modalities without major structural implications because of the modular nature of the design. Furthermore, we showed that our neural network performed better when using multiple modalities, as opposed to a single modality. The dataset and code used in this paper are openly available.

Full-body motion capture typically requires sensors/markers to be placed on each rigid body segment, which results in long setup times and is obtrusive. The number of sensors/markers can be reduced using deep learning or offline methods. However, this requires large training datasets and/or sufficient computational resources. Therefore, we investigate the following research question: “What is the performance of a shallow approach, compared to a deep learning one, for estimating time coherent full-body poses using only five inertial sensors?”. We propose to incorporate past/future inertial sensor information into a stacked input vector, which is fed to a shallow neural network for estimating full-body poses. Shallow and deep learning approaches are compared using the same input vector configurations. Additionally, the inclusion of acceleration input is evaluated. The results show that a shallow learning approach can estimate full-body poses with a similar accuracy (~6 cm) to that of a deep learning approach (~7 cm). However, the jerk errors are smaller using the deep learning approach, which can be the effect of explicit recurrent modelling. Furthermore, it is shown that the delay using a shallow learning approach (72 ms) is smaller than that of a deep learning approach (117 ms).