Search

BACKGROUND Decision-support systems for the suggestion of optimal treatments for individual patients with back pain are currently insufficiently accurate to use for clinical application. Most of the input provided to train these systems is based on patient reported outcome measures. However, with the appearance of electronic patient records, additional qualitative data on reasons for referrals, and patients’ goals become available for support systems. There are currently no decision support tools that cover a wide range of biopsychosocial factors including referral letter information to help clinicians triage patients to the optimal LBP-treatment. OBJECTIVE The objective of this study was to investigate the added value of including qualitative data from electronic patient records on the accuracy of a quantitative decision-support system for patients with low back pain. METHODS A retrospective study in a clinical cohort of Dutch patients with low back pain was carried out. Patients filled out a baseline questionnaire about demographics, pain, disability, work status, quality of life, medication, psychosocial functioning, comorbidity, history and duration of pain. Referral reasons and patient request for help (patient goals) were extracted via natural language processing (NLP) and enriched the dataset. For decision support, these data were considered as independent factors for the choice to triage to neurosurgery, anesthesiology, rehabilitation or minimal intervention. Support Vector Machine, Boosted Decision Tree and Multiclass Neural Network (NN) models are trained for two conditions: with and without consideration of referral letter content. The models’ accuracies were evaluated via F1 Scores, and Confusion Matrices to predict treatment-path (out of 4 paths) with and without the additional referral parameters. RESULTS Data of 1608 patients were evaluated. The evaluation indicates that two referral reasons from the referral letter (for an anesthesiology and a rehabilitation intervention) increased the F1-Score accuracy by up to 19.5% for the triaging. The results were confirmed by the confusion matrices. CONCLUSIONS This study indicates that data-enriching by adding NLP-based extraction of the content of referral letters increases the model accuracy of decision support systems in suggesting optimal treatments for individual patients with low back pain. Overall model accuracies were considered low and insufficient for clinical application.

A continuous monitoring of the physical strength and mobility of elderly people is important for maintaining their health and treating diseases at an early stage. However, frequent screenings by physicians are exceeding the logistic capacities. An alternate approach is the automatic and unobtrusive collection of functional measures by ambient sensors. In the current publication, we show the correlation among data of ambient motion sensors and the well-established mobility assessments Short-PhysicalPerformance-Battery, Tinetti and Timed Up & Go. We use the average number of motion sensor events as activity measure for correlation with the assessment scores. The evaluation on a real-world dataset shows a moderate to strong correlation with the scores of standardised geriatrics physical assessments.

Smart support systems for the recognition of Activities of Daily Living (ADLs) can help elderly people live independently for longer improving their standard of living. Many machine learning approaches have been proposed lately for Human Activity Recognition (HAR), including elaborated networks that contain convolutional, recurrent, and attentive layers. The ubiquity of wearable devices has provided an increasing amount of time-series data that can be used for such applications in an unobtrusive manner. But there are not many studies on the performance of the attention-based Transformer model in HAR, especially not for complex activities such as ADLs. This work implements and evaluates the novel self-attention Transformer model for the classification of ADLs and compares it to the already well-established approach of recurrent Long-Short Term Memory (LSTM) networks. The proposed method is a two-level hierarchical model, in which atomic activities are initially recognized in the first step and their probability scores are extracted and utilized for the Transformer-based classification of seven more complex ADLs in the second step. The Transformer is used at the second step to classify seven ADLs. Our results show that the Transformer model reaches the same performance and even outperforms LSTM networks cleary in the subject-dependent configuration (73.36 % and 69.09 %), while relying only on attention-mechanism to depict global dependencies between input and output without the need to use any recurrence. The proposed model was tested using two different segment lengths, indicating its effectiveness in learning long-range dependencies of shorter actions in complex activities.

A continuous monitoring of the physical strength and mobility of elderly people is important for maintaining their health and treating diseases at an early stage. However, frequent screenings by physicians are exceeding the logistic capacities. An alternate approach is the automatic and unobtrusive collection of functional measures by ambient sensors. In the current publication, we show the correlation among data of ambient motion sensors and the well-established mobility assessments Short-Physical-Performance-Battery, Tinetti and Timed Up & Go. We use the average number of motion sensor events as activity measure for correlation with the assessment scores. The evaluation on a real-world dataset shows a moderate to strong correlation with the scores of standardised geriatrics physical assessments., 15 pages, 1 figure, 16 tables

Background It is important to identify the relevant parameters of physical performance to prevent early functional decline and to prolong independent living. The aim of this study is to describe the development of physical performance in a healthy community-dwelling older cohort aged 70+ years using comprehensive assessment over two years and to subsequently identify the most relevant predictive tests for physical decline to minimize assessment. Methods Physical performance was measured by comprehensive geriatric assessment. Predictors for the individual decline of physical performance by Principal Component and k-means Cluster Analysis were developed, and sensitivity and specificity determined accordingly. Results 251 subjects (Ø 75.4 years) participated in the study. Handgrip strength was low in 21.1%. The follow-up results of tests were divergent. Handgrip strength [− 16.95 (SD 11.55)] and the stair climb power test (power) [− 9.15 (SD 16.84)] yielded the highest percentage changes. Four most relevant tests (handgrip strength, stair climb power time, timed up & go and 4-m gait speed) were identified. A predictor based on baseline data was determined (sensitivity 82%, specificity 96%) to identify subjects characterized by a high degree of physical decline within two years. Discussion Although the cohort of older adults is heterogeneous, most of the individuals in the study exhibited high levels of physical performance; only a few subjects suffered a relevant decline within the 2-year follow-up. Four most relevant tests were identified to predict relevant decline of physical function. Conclusion In spite of ceiling effects of the geriatric assessment in high-performers, we assume that it is possible to predict an individual’s risk of physical decline within 2 years with four tests of a comprehensive geriatric assessment.

A continuous monitoring of the physical strength and mobility of elderly people is important for maintaining their health and treating diseases at an early stage. However, frequent screenings by physicians are exceeding the logistic capacities. An alternate approach is the automatic and unobtrusive collection of functional measures by ambient sensors. In the current publication, we show the correlation among data of ambient motion sensors and the wellestablished mobility assessment Short-Physical-Performance-Battery and Tinetti. We use the average number of motion sensor events for correlation with the assessment scores. The evaluation on a real-world dataset shows a moderate to strong correlation with the scores of standardised geriatrics physical assessments.

The The 2nd Nurse Care Activity Recognition Challenge Using Lab and Field Data addresses the important issue about care and the need for assistance systems in the nursing profession like automatic documentation systems. Data of 12 different care activities were recorded with an accelerometer attached to the right arm of the nurses. Both, laboratory and field data were taken into account. The task was to classify each activity based on the accelerometer data. We participated as team Gudetama in the challenge. We trained a Random Forest classifier and achieved an accuracy of 61.11% on our internal test set.

Dyslexia is a widespread specific learning disorder, which can have a particularly negative influence on the learning success of children. Early detection of dyslexia is the foundation for early intervention, which is the key to reduce the adverse effects of dyslexia, e.g., bad school grades. In this paper, we present the prototype of a puzzle app, which we explicitly designed with the human-centred design (HCD) process for dyslexia screening in pre-reader using new indicators related to motor skills to ensure users needs for the data collection to apply machine learning prediction. The app records the telemetry of the gaming sequence in order to derive future statements about the prevalence of dyslexia based on the telemetry data. The high-fidelity prototype was evaluated with a five-user test usability study with five German-speaking child-parent pairs. The results show how young children and parents are interacting with new games, and how new applications (web and mobile technologies) which are used for online experiments, could be developed. The usability of the prototype is suitable for the target group with only minor limitations.

This article introduces the architecture of a Long-Short-Term Memory network for classifying transportation-modes via Smartphone data and evaluates its accuracy. By using a Long-Short-Term-Memory Network with common preprocessing steps such as normalisation for classification tasks a F1-Score accuracy of 63.68\% was achieved with an internal test dataset. We participated as Team 'GanbareAM' in the 'SHL recognition challenge'., 5 pages, 6 figures, 2 tables, ubicomp19

The Ovako Working posture Assessment System (OWAS) is a commonly used observational assessment method for determining the risk of work-related musculoskeletal disorders. OWAS claims to be suitable in the application for untrained persons but there is not enough evidence for this assumption. In this paper, inter-rater (inter-observer) reliability (agreement) is examined down to the level of individual postures and categories. For this purpose, the postures of 20 volunteers have been observed by 3 varying human raters in a laboratory setting and the inter-rater agreement against reference values was determined. A high agreement of over 98%(κ=0.98) was found for the postures of the arms but lower agreements were found for posture classification of the legs (66-97%,κ=0.85) and the upper body (80-96%,κ=0.85). No significant difference was found between raters with and without intense prior training in physical therapy. Consequently, the results confirm the general reliability of the OWAS method especially for raters with non-specialized background but suggests weaknesses in the reliable detection of a few particular postures.

Cardiopulmonary resuscitation (CPR) is one of the most critical emergency interventions for sudden cardiac arrest. In this paper, a robust sinusoidal model-fitting method based on a Evolution Strategy inspired algorithm for CPR quality parameters -- naming chest compression frequency and depth -- as measured by an inertial measurement unit (IMU) attached to the wrist is presented. The proposed approach will allow bystanders to improve CPR as part of a continuous closed-loop support system once integrated into a smartphone or smartwatch application. By evaluating the model's precision with data recorded by a training mannequin as reference standard, a variance for the compression frequency of $\pm 2.22$ compressions per minute (cpm) has been found for the IMU attached to the wrist. It was found that this previously unconsidered position and thus, the use of smartwatches is a suitable alternative to the typical placement of phones in hand for CPR training., Comment: 26 pages. arXiv admin note: text overlap with arXiv:1809.07692

The stair climbing test (SCT) is a standard geriatric assessment to measure lower-limb strength being one of the essential components of physical function. To detect functional decline as early as possible, regular assessments of mobility, balance, and strength are necessary. Inertial measurement units (IMU) are a promising technology for flexible and unobtrusive measurements of the SCTs. We introduce an automated assessment via IMUs in a study of 83 participants aged 70–87 (75.64 ± 4,17) years. The activity of stair ascending has been automatically classified via a k-nearest-neighbor classifier and the performance was evaluated regarding the power. Therefore, we considered both, stair climb average power and peak power. Stair ascending was correctly classified in 93% of the cases with a mean deviation of 2.35% of the average power value in comparison to conventional measurements. Additionally, we showed the medical sensitivity of our system regarding the detection of transitions towards the frail status in controlled conditions and also confirmed the general suitability of automated stair climb analyses in unsupervised home-assessments.

Driving is a complex and cognitively demanding task. It is essential to assess the cognitive state of the driver in order to design cognitive technical systems that can adapt to different driver cognitive states. Our research attempts to assess these states using functional Near Infrared Spectroscopy (fNIRS) by measuring brain activity in a virtual reality driving simulator. However, the fNIRS brain activation measurements could be influenced by muscle activity and we wanted to investigate this phenomenon. For this, we designed a paradigm with two conditions (listen, teeth clench) which show a significant contrast in the influence of muscle activity. We observed that the muscle hemodynamic response can show a higher magnitude of signal change compared to brain hemodynamic response. The muscle hemodynamic response showed an increase in deoxygenated hemoglobin (HbR) whereas the brain hemodynamic response showed a decrease in HbR. Moreover, the dynamics of the brain and muscle hemodynamic response differed. The brain response showed the same latency for oxygenated hemoglobin (HbO) and HbR while the muscle HbR response had a slower latency compared to HbO. We concluded that the fNIRS brain activation measurements could indeed be influenced by muscle activity. We were also able to determine some characteristics of the muscle hemodynamic response.

Indicators for a decrement in vigilance are a slowdown in reaction times and an increase in alpha power in the electroencephalogram in posterior regions of the brain. Transcranial alternating current stimulation (tACS) is a neuropsychological technique that has been found to interact with intrinsic brain oscillations and is able to enhance cognitive and behavioral performance. Recent studies show that tACS in the gamma frequency range (30–80 Hz) is able to downregulate amplitudes in the alpha frequency range (8–12 Hz), in accordance to the effect referred to as cross-frequency coupling, where intrinsic alpha and gamma waves modulate each other. We applied 40 Hz gamma-tACS to the visual cortex during a vigilance experiment and investigated if stimulation improves reaction times and error rates with time-on-task. In our sham controlled experiment, participants completed two blocks of 30 minutes duration while performing the same visual two-choice task. The first block was used as BASELINE. A statistical analysis with a linear mixed model revealed a significantly lower increase of modeled reaction times over time in the INTERVENTION-block of the tACS-group as compared with their BASELINE-block whereas there was no significant change between the BASELINE- and INTERVENTION-block for the SHAM-group. Error rates did not differ between groups. This paper indicates that gamma-tACS can enhance performance in vigilance tasks as it significantly decreased the slowdown of reaction times in our study.

In this paper, we show how the XML dialect SKAML (Skeletal Assessment Markup Language) can be used to use data from one or more Motion Capture systems to perform human posture assessments with multiple assessment methods. We show an implementation example using an inertial measuring suit and both OWAS and REBA assessment methods. SKAML makes it possible to implement classifiers for a Motion Capture system once and adapt the classifier by-configuration to various ergonomics assessment methods. We anticipate our work as help for researchers and developers that implement new assessment methods or motion capture systems.

This paper presents a new mobile near-infrared functional spectroscopy (fNIRS) device, with digital detectors that can be placed anywhere on the head and fit into standard caps to measure cortical brain activation. The device's functionality was evaluated in two steps, i.e. first, by means of simple pulse measurements and second, in a motor cortex study with nine subjects. In this study, the subjects had to alternate between right and left hands while using hand-held strength trainers. While the signals from the mobile prototype were not yet stable enough across all channels to perform analysis such as statistical parametric mapping, it was able to measure significant brain activation changes over the area of the motor cortex with the mobile prototype when the contralateral hand was activated in four subjects. In contrast, the device was yet unable to measure ipsilateral activities. The problems encountered and possible methods to improve signal acquisition are discussed at the end of the paper.

Cardiopulmonary resuscitation (CPR) is alongside electrical defibrillation the most crucial countermeasure for sudden cardiac arrest, which affects thousands of individuals every year. In this paper, we present a novel approach including sinusoid models that use skeletal motion data from an RGB-D (Kinect) sensor and the Differential Evolution (DE) optimization algorithm to dynamically fit sinusoidal curves to derive frequency and depth parameters for cardiopulmonary resuscitation training. It is intended to be part of a robust and easy-to-use feedback system for CPR training, allowing its use for unsupervised training. The accuracy of this DE-based approach is evaluated in comparison with data of 28 participants recorded by a state-of-the-art training mannequin. We optimized the DE algorithm hyperparameters and showed that with these optimized parameters the frequency of the CPR is recognized with a median error of $\pm 2.9$ compressions per minute compared to the reference training mannequin., Final paper, 22 pages

To initiate appropriate interventions and avoid physical decline, comprehensive measurements are needed to detect functional changes in elderly people at the earliest possible stage. The established Timed UpGo (TUG) test takes little time and, due to its standardized and easy procedure, can be conducted by elderly people in their own homes without clinical guidance. Therefore, cheap light barriers (LBs) and force sensors (FSs) are well suited ambient sensors that could easily be attached to existing (arm)chairs to measure and report TUG times in order to identify functional decline. We validated the sensitivity of these sensors in a clinical trial with 100 elderlies aged 58-92 years with a mean of 74 (±6.78) years by comparing the sensor-based results with standard TUG measurements using a stopwatch. We further evaluated the accuracy enhancement when calibrating the algorithm via a mixed linear model. With calibration, the LBs achieved a root mean square error (RMSE) of 0.83 s, compared to 1.90 s without, and the FSs achieved 0.90 s compared to 2.12 s without. The suitability of measuring accurate TUG times with each of the ambient sensors and of measuring TUG regularly in the homes of elderly people could be confirmed.

Driving is a complex and cognitively demanding task. It is important to assess the cognitive state of the driver in order to develop cognitive technical systems that can adapt to different cognitive states of the driver. For this purpose, we have developed a mobile functional near-infrared spectroscopy (mofNIRS) prototype. This paper describes the improvements of this mobile prototype with freely placeable optodes on a subject’s head and the results of an evaluation study. We conducted a motor cortex experiment with four subjects, whereby the mobile prototype was mounted on the right hemisphere and a commercial, stationary fNIRS on the left hemisphere above the motor cortex area. One data set had to be discarded due to incorrect synchronization between both systems. The results of the remaining three subjects are presented and discussed in this paper. Here, we report the results from the time-series and Statistical Parametric Mapping (SPM) analyses, which shows t-values with high differentiability of the Results. Furthermore, both analysis methods show comparable results between the commercial system and the mobile prototype.

To identify the functional decline as related to aging, geriatric assessments are an established instrument. Within such assessments, the functional ability is evaluated and consists of the three major components: strength, mobility, and balance. Counter movement jumps (CMJ) are well-suited to test these three essential elements of functional ability within a single assessment item. Since common balance measures have been shown to be significantly prone to algorithmic and technical variations, a robust alternative method is required. Thus, we introduce a model-based approach for balance and strength analyses, where the human lower extremities are modeled as an oscillating system during the phase of landing and recovery after a vertical jump. In the System and Control Technology, a transfer function of an oscillating system is described by a second-order delay element (PT2-element), which is characterized by the parameters natural frequency and damping. We analyze the jumps of 30 participants (70–87 years) regarding their jump phases and the mentioned parameters. A linear correlation between jump power and jump height, which are sensitive indicators of the muscle performance and the strength could be confirmed. While a correlation between jump power and spring constant could be observed, a significant relationship between the balance ability and natural frequency could not be identified.

Gait analysis is often supported by technology. Due to limitations in optical systems, such as limited measurement volumes and the requirement of a laboratory environment, low-cost inertial measurement unit (IMU) based motion capture systems might be better suited for gait analysis since they involve no spatial limitations and are flexibly applicable. In this paper, we investigate if low-cost IMU-based motion capture suits are an adequate alternative for clinical gait analysis in terms of accuracy of the determination of joint flexions and gait parameters. For this reason, we developed a gait analysis system and a gait analysis algorithm, which detects joint positions based on the Joint Coordinate System and determines knee, hip, and ankle flexions, as well as spatiotemporal parameters such as the number of steps, cadence, step duration and step length, and the specific gait phases. We evaluated and validated the IMU-based system in comparison to camera-based measurements (as gold standard) with three different healthy adult subjects. The evaluation indicates that the full-body motion capture system achieves a high degree of precision (0.86) and recall (0.98) in the recognition of gait cycles. The harmonic mean F0.15 of the two factors precision and recall is on average 0.96 and the mentioned temporal gait parameters can be determined with an error below 10 ms. The mean deviation in the determination of joint angles amounts 1.35° ± 2°. Consequently, the article at hand indicates that low-cost IMU-based motion capture suits are an accurate alternative for gait analysis.

Early detection of changes in mobility associated with functional decline can increase the therapeutic success by prolonging self-determined living. To get an unbiased and high frequently status of the physical performance of the persons at risk, unsupervised assessments of their functional abilities should ideally take place in their homes. Thus, we have developed a minimized unsupervised technical assessment of physical performance in domestic environments. By conducting an exploratory factor analysis, based on the results of 79 study participants with a minimum age of 70 years, we could clarify that common assessment items mainly represent three key parameters of functional performance "mobility and endurance", "strength" and "balance". Consequently, we identified a minimal set of assessment items that is suitable for home-assessments and that, since covering all three parameters, is able to generate clinical meaningful and relevant insights about the functional status. Regarding the parameter mobility, we developed a technical assessment of physical performance for domestic environments, which utilizes short distance walk times assessed via ambient presence sensors as an indicator for potential functional decline. In a field trial over ten months with 20 participants with a mean age of 84.25 years, we could confirm the general feasibility of our approach and the proposed system.

Healthy aging is a core societal aim especially regarding the demographic change. But with aging, functional decline can occur and this is a major challenge for health care systems. For the evaluation of the health of older adults and the identification of early changes associated with functional and cognitive decline, clinical geriatric assessments are a well-established approach. Ideally, the assessments should take place at home of the older adults or even in their daily life, to get an unbiased functional status. Therefore, we introduce a technology supported geriatric assessment as an intermediate step to a home-assessment or in future to sensor-based-assessments in daily life. Beside various ambient sensors, a sensor belt is used during the assessments and for 1 week in the participants’ daily life. We discuss the suitability of our measuring devices for an ambient home-assessment and evaluate the sensors in comparison to valid measurements. Thereby, we show that light barrier measurements achieve a high sensitivity and a good correlation to manual measurements through study nurses or physical therapists.

This article introduces a gesture-based user interface for hospital beds. This interface enables caregivers to focus on their patients and have both hands available for mobilizing and transferring them. Gestures are detected either via static (96% sensitivity) or dynamic gestures (67.5% sensitivity) and might be corrected by an extra repetition. Once gestures are correctly detected, caregivers can trigger bed movements via a foot switch as a hands-free operation, which as well functions as a dead man button. The evaluation of the usability through interviews with caregivers highlighted the system’s general applicability, but as well some future challenges that have to be solved in order to achieve a system for every-day use.