Search

Purpose: To gain an understanding of the heterogeneous group of type 2 diabetes (T2D) patients, we aimed to identify patients with the homogenous long-term HbA1c trajectories and to predict the trajectory membership for each patient using explainable machine learning methods and different clinical-, treatment-, and socio-economic-related predictors. Patients and Methods: Electronic health records data covering primary and specialized healthcare on 9631 patients having T2D diagnosis were extracted from the North Karelia region, Finland. Six-year HbA1c trajectories were examined with growth mixture models. Linear discriminant analysis and neural networks were applied to predict the trajectory membership individually. Results: Three HbA1c trajectories were distinguished over six years: “stable, adequate” (86.5%), “improving, but inadequate” (7.3%), and “fluctuating, inadequate” (6.2%) glycemic control. Prior glucose levels, duration of T2D, use of insulin only, use of insulin together with some oral antidiabetic medications, and use of only metformin were the most important predictors for the long-term treatment balance. The prediction model had a balanced accuracy of 85% and a receiving operating characteristic area under the curve of 91%, indicating high performance. Moreover, the results based on SHAP (Shapley additive explanations) values show that it is possible to explain the outcomes of machine learning methods at the population and individual levels. Conclusion: Heterogeneity in long-term glycemic control can be predicted with confidence by utilizing information from previous HbA1c levels, fasting plasma glucose, duration of T2D, and use of antidiabetic medications. In future, the expected development of HbA1c could be predicted based on the patient’s unique risk factors offering a practical tool for clinicians to support treatment planning.

Artificial intelligence (AI) is increasingly being used to improve patient care and management. In this paper, we propose explainable AI (XAI) models for predicting severe hypoglycemia (SH) and diabetic ketoacidosis (DKA) episodes in adults with type 1 diabetes (T1D) and relapses in adults with relapsing-remitting multiple sclerosis (RRMS). We follow a three-step process in this study: (1) develop baseline machine learning (ML) models, (2) improve the models using ReliefF feature selection technique, and develop sex-stratified models, (3) explain the models and their results using SHapley Additive exPlanations (SHAP). We built six ML models (XGBoost, LightGBM, CatBoost, AdaBoost, random forest, and linear regression) for all scenarios. Applying the ReliefF feature selection led to improved model performance in predicting all outcomes compared to the baseline models. Additionally, sex-stratified models further improved the prediction of SH episodes and relapses. The F1 scores for predicting SH episodes in male and female patients were 84.07% and 84.95%, respectively, and the DKA prediction model achieved an F1 score of 78.67%. The proposed relapse prediction models outperformed existing models with F1 scores of 84.55% (males) and 76.11% (females), and ROCs of 70.26% (males) and 69.05% (females). Our results highlight the importance of considering sex differences, socioeconomic factors, and physical and mental health in medical outcome prediction. Boosting ML algorithms were found to be effective in detecting SH and DKA in T1D patients and relapses in RRMS patients compared to conventional tree-based ML and statistical models.

Clinical data analysis could lead to breakthroughs. However, clinical data contain sensitive information about participants that could be utilized for unethical activities, such as blackmailing, identity theft, mass surveillance, or social engineering. Data anonymization is a standard step during data collection, before sharing, to overcome the risk of disclosure. However, conventional data anonymization techniques are not foolproof and also hinder the opportunity for personalized evaluations. Much research has been done for synthetic data generation using generative adversarial networks and many other machine learning methods; however, these methods are either not free to use or are limited in capacity. This study evaluates the performance of an emerging tool named synthpop, an R package producing synthetic data as an alternative approach for data anonymization. This paper establishes data standards derived from the original data set based on the utilities and quality of information and measures variations in the synthetic data set to evaluate the performance of the data synthesis process. The methods to assess the utility of the synthetic data set can be broadly divided into two approaches: general utility and specific utility. General utility assesses whether synthetic data have overall similarities in the statistical properties and multivariate relationships with the original data set. Simultaneously, the specific utility assesses the similarity of a fitted model’s performance on the synthetic data to its performance on the original data. The quality of information is assessed by comparing variations in entropy bits and mutual information to response variables within the original and synthetic data sets. The study reveals that synthetic data succeeded at all utility tests with a statistically non-significant difference and not only preserved the utilities but also preserved the complexity of the original data set according to the data standard establish

Digital twin (DT) emerges as a key concept of the Industry 4.0 paradigm and beyond. However, the current literature lacks focus on humans and human activities as a part of complex system DTs. Acknowledging human aspects in DTs can enhance work performance, well-being, motivation, and personal development of professionals. This study examines emerging requirements for human digital twins (HDTs) in three use cases of industry–academia collaboration on complex systems. The results draw together the overall design problem and four design objectives for HDTs. We propose to combine the machine and human-related aspects of DTs and highlight the need for virtual-to-virtual interoperability between HDTs and machines alike. Furthermore, we outline differences between humans and machines regarding digital twinning by addressing human activities and knowledge-based behavior on systems. Design of HDTs requires understanding of individual professional characteristics, such as skills and information preferences, together with twinning between the physical and digital machine entities and interactions between the human and machine DTs. As the field moves toward including humans as a part of the DT concept, incorporating HDTs in complex systems emerges as an increasingly significant issue.

The aim of this paper is to identify the barriers that are specifically relevant to the use of Artificial Intelligence (AI)-based evidence in Central and Eastern European (CEE) Health Technology Assessment (HTA) systems. The study relied on two main parallel sources to identify barriers to use AI methodologies in HTA in CEE, including a scoping literature review and iterative focus group meetings with HTx team members. Most of the other selected articles discussed AI from a clinical perspective (n = 25), and the rest are from regulatory perspective (n = 13), and transfer of knowledge point of view (n = 3). Clinical areas studied are quite diverse—from pediatric, diabetes, diagnostic radiology, gynecology, oncology, surgery, psychiatry, cardiology, infection diseases, and oncology. Out of all 38 articles, 25 (66%) describe the AI method and the rest are more focused on the utilization barriers of different health care services and programs. The potential barriers could be classified as data related, methodological, technological, regulatory and policy related, and human factor related. Some of the barriers are quite similar, especially concerning the technologies. Studies focusing on the AI usage for HTA decision making are scarce. AI and augmented decision making tools are a novel science, and we are in the process of adapting it to existing needs. HTA as a process requires multiple steps, multiple evaluations which rely on heterogenous data. Therefore, the observed range of barriers come as a no surprise, and experts in the field need to give their opinion on the most important barriers in order to develop recommendations to overcome them and to disseminate the practical application of these tools.

Meditation is a practice that aims at self-inducing a state of calmed rest. In this work, we analyze physiological signals collected with wearable sensors to observe if meditation has a noticeable effect on the human body response and if this effect is inversely related to stress and can be detected using the same biosignals and similar features and methods. Our work is based on the extraction of statistical and physiological features and extends the models found in the literature by extracting 30 additional features related to heart rate variability. The results show that using wrist wearable devices, meditation periods can be distinguished from spontaneous rest with an accuracy of up to 86% accuracy.

This study introduces an ensemble-based personalized human activity recognition method relying on incremental learning, which is a method for continuous learning, that can not only learn from streaming data but also adapt to different contexts and changes in context. This adaptation is based on a novel weighting approach which gives bigger weight to those base models of the ensemble which are the most suitable to the current context. In this article, contexts are different body positions for inertial sensors. The experiments are performed in two scenarios: (S1) adapting model to a known context, and (S2) adapting model to a previously unknown context. In both scenarios, the models had to also adapt to the data of previously unknown person, as the initial user-independent dataset did not include any data from the studied user. In the experiments, the proposed ensemble-based approach is compared to non-weighted personalization method relying on ensemble-based classifier and to static user-independent model. Both ensemble models are experimented using three different base classifiers (linear discriminant analysis, quadratic discriminant analysis, and classification and regression tree). The results show that the proposed ensemble method performs much better than non-weighted ensemble model for personalization in both scenarios no matter which base classifier is used. Moreover, the proposed method outperforms user-independent models. In scenario 1, the error rate of balanced accuracy using user-independent model was 13.3%, using non-weighted personalization method 13.8%, and using the proposed method 6.4%. The difference is even bigger in scenario 2, where the error rate using user-independent model is 36.6%, using non-weighted personalization method 36.9%, and using the proposed method 14.1%. In addition, F1 scores also show that the proposed method performs much better in both scenarios that the rival methods. Moreover, as a side result, it was noted that the

In this article, regression and classification models are compared for stress detection. Both personal and user-independent models are experimented. The article is based on publicly open dataset called AffectiveROAD, which contains data gathered using Empatica E4 sensor and unlike most of the other stress detection datasets, it contains continuous target variables. The used classification model is Random Forest and the regression model is Bagged tree based ensemble. Based on experiments, regression models outperform classification models, when classifying observations as stressed or not-stressed. The best user-independent results are obtained using a combination of blood volume pulse and skin temperature features, and using these the average balanced accuracy was 74.1% with classification model and 82.3% using regression model. In addition, regression models can be used to estimate the level of the stress. Moreover, the results based on models trained using personal data are not encouraging showing that biosignals have a lot of variation not only between the study subjects but also between the session gathered from the same person. On the other hand, it is shown that with subject-wise feature selection for user-independent model, it is possible to improve recognition models more than by using personal training data to build personal models. In fact, it is shown that with subject-wise feature selection, the average detection rate can be improved as much as 4%-units, and it is especially useful to reduce the variance in the recognition rates between the study subjects.

This study presents incremental learning based methods to personalize human activity recognition models. Initially, a user-independent model is used in the recognition process. When a new user starts to use the human activity recognition application, personal streaming data can be gathered. Of course, this data does not have labels. However, there are three different ways to obtain this data: non-supervised, semi-supervised, and supervised. The non-supervised approach relies purely on predicted labels, the supervised approach uses only human intelligence to label the data, and the proposed method for semi-supervised learning is a combination of these two: It uses artificial intelligence (AI) in most cases to label the data but in uncertain cases it relies on human intelligence. After labels are obtained, the personalization process continues by using the streaming data and these labels to update the incremental learning based model, which in this case is Learn++. Learn++ is an ensemble method that can use any classifier as a base classifier, and this study compares three base classifiers: linear discriminant analysis (LDA), quadratic discriminant analysis (QDA), and classification and regression tree (CART). Moreover, three datasets are used in the experiment to show how well the presented method generalizes on different datasets. The results show that personalized models are much more accurate than user-independent models. On average, the recognition rates are: 87.0% using the user-independent model, 89.1% using the non-supervised personalization approach, 94.0% using the semi-supervised personalization approach, and 96.5% using the supervised personalization approach. This means that by relying on predicted labels with high confidence, and asking the user to label only uncertain observations (6.6% of the observations when using LDA, 7.7% when using QDA, and 18.3% using CART), almost as low error rates can be achieved as by using the supervised approach, i

In this study, importance of user inputs is studied in the context of personalizing human activity recognition models using incremental learning. Inertial sensor data from three body positions are used, and the classification is based on Learn++ ensemble method. Three different approaches to update models are compared: non-supervised, semi-supervised and supervised. Non-supervised approach relies fully on predicted labels, supervised fully on user labeled data, and the proposed method for semisupervised learning, is a combination of these two. In fact, our experiments show that by relying on predicted labels with high confidence, and asking the user to label only uncertain observations (from 12% to 26% of the observations depending on the used base classifier), almost as low error rates can be achieved as by using supervised approach. In fact, the difference was less than 2%-units. Moreover, unlike non-supervised approach, semisupervised approach does not suffer from drastic concept drift, and thus, the error rate of the non-supervised approach is over 5%-units higher than using semi-supervised approach.

The recognition of complex and subtle human behaviors from wearable sensors will enable next-generation human-oriented computing in scenarios of high societal value (e.g., dementia care). This will require large-scale human activity corpuses and much improved methods to recognize activities and the context in which they occur. This workshop deals with the challenges of designing reproducible experimental setups, running large-scale dataset collection campaigns, designing activity and context recognition methods that are robust and adaptive, and evaluating systems in the real world. We wish to reflect on future methods, such as lifelong learning approaches that allow open-ended activity recognition. This year HASCA will welcome papers from participants to the Second Sussex-Huawei Locomotion and Transportation Recognition Competition and Open Lab Nursing Activity Recognition Challenge in special sessions.

Stress detection is becoming a popular field in machine learning and this study focuses on recognizing stress using the sensors of commercially available smartwatches. In most of the previous studies, stress detection is based on partly or fully on electrodermal activity sensor (EDA). However, if the final aim of the study is to build a smartwatch application, using EDA signal is problematic as the smartwatches currently in the market do not include sensor to measure EDA signal. Therefore, this study surveys what sensors the smartwatches currently in the market include, and which of them 3rd party developers have access to. Moreover, it is studied how accurately stress can be detected user-independently using different sensor combinations. In addition, it is studied how detection rates vary between study subjects and what kind of effect window size has to the recognition rates. All of the experiments are based on publicly available WESAD dataset. The results show that, indeed, EDA signal is not necessary when detecting stress user-independently, and therefore, commercial smartwatches can be used for recognizing stress when the used window length is big enough. However, it is also noted that recognition rate varies a lot between the study subjects.

Data are crucial to support decision making. If data have low veracity, decisions are not likely to be sound. Internet of Things (IoT) generates big data with inaccuracy, inconsistency, incompleteness, deception, and model approximation. Enhancing data veracity is important to address these challenges. In this article, we summarize the key characteristics and challenges of IoT, which influence data processing and decision making. We review the landscape of measuring and enhancing data veracity and mining uncertain data streams. Moreover, we propose five recommendations for future development of veracious big IoT data analytics that are related to the heterogeneous and distributed nature of IoT data, autonomous decision-making, context-aware and domain-optimized methodologies, data cleaning and processing techniques for IoT edge devices, and privacy preserving, personalized, and secure data management.

The recognition of complex and subtle human behaviors from wearable sensors will enable next-generation human-oriented computing in scenarios of high societal value (e.g., dementia care). This will require large-scale human activity corpuses and much improved methods to recognize activities and the context in which they occur. This workshop deals with the challenges of designing reproducible experimental setups, running large-scale dataset collection campaigns, designing activity and context recognition methods that are robust and adaptive, and evaluating systems in the real world. We wish to reflect on future methods, such as lifelong learning approaches that allow open-ended activity recognition. Unique this year, HASCA will welcome papers from participants to the Sussex-Huawei Locomotion and Transportation Recognition Competition in a special session.

This study introduces OpenHAR, a free Matlab toolbox to combine and unify publicly open data sets. It provides an easy access to accelerometer signals of ten publicly open human activity data sets. Data sets are easy to access as OpenHAR provides all the data sets in the same format. In addition, units, measurement range and labels are unified, as well as, body position IDs. Moreover, data sets with different sampling rates are unified using downsampling. What is more, data sets have been visually inspected to find visible errors, such as sensor in wrong orientation. OpenHAR improves re-usability of data sets by fixing these errors. Altogether OpenHAR contains over 65 million labeled data samples. This is equivalent to over 280 hours of data from 3D accelerometers. This includes data from 211 study subjects performing 17 daily human activities and wearing sensors in 14 different body positions.

Our article “Recognizing human activities user-independently on smartphones based on accelerometer data” was published in the International Journal of Interactive Multimedia and Artificial Intelligence (IJIMAI) in 2012. In 2018, it was selected as the most outstanding article published in the 10 years of IJIMAI life. To celebrate the 10th anniversary of IJIMAI, in this article we will introduce what has happened in the field of human activity recognition and wearable sensor-based recognition since 2012, and especially, this article concentrates on introducing our work since 2012.

In this study, the aim is to personalize inertial sensor databased human activity recognition models using incremental learning. At first, the recognition is based on user-independent model. However, when personal streaming data becomes available, the incremental learning-based recognition model can be updated, and therefore personalized, based on the data without user-interruption. The used incremental learning algorithm is Learn++ which is an ensemble method that can use any classifier as a base classifier. In fact, study compares three different base classifiers: linear discriminant analysis (LDA), quadratic discriminant analysis (QDA) and classification and regression tree (CART). Experiments are based on publicly open data set and they show that already a small personal training data set can improve the classification accuracy. Improvement using LDA as base classifier is 4.6 percentage units, using QDA 2.0 percentage units, and 2.3 percentage units using CART. However, if the user-independent model used in the first phase of the recognition process is not accurate enough, personalization cannot improve recognition accuracy.

This work investigates how context should be taken into account when performing continuous authentication of a smartphone user based on touchscreen and accelerometer readings extracted from swipe gestures. The study is conducted on the publicly available HMOG dataset consisting of 100 study subjects performing pre-defined reading and navigation tasks while sitting and walking. It is shown that context-specific models are needed for different smartphone usage and human activity scenarios to minimize authentication error. Also, the experimental results suggests that utilization of phone movement improves swipe gesture-based verification performance only when the user is moving.

The migraine is a chronic, incapacitating neurovascular disorder, characterized by attacks of severe headache and autonomic nervous system dysfunction. Among the working age population, the costs of migraine are 111 billion euros in Europe alone. The early detection of migraine attacks would reduce these costs, as it would shorten the migraine attack by enabling correct timing when taking preventive medication. In this article, whether it is possible to detect migraine attacks beforehand using wearable sensors is studied, and t preliminary results about how accurate the recognition can be are provided. The data for the study were collected from seven study subjects using a wrist-worn Empatica E4 sensor, which measures acceleration, galvanic skin response, blood volume pulse, heart rate and heart rate variability, and temperature. Only sleep time data were used in this study. A novel method to increase the number of training samples is introduced, and the results show that, using personal recognition models and quadratic discriminant analysis as a classifier, balanced accuracy for detecting attacks one night prior is over 84%. While this detection rate is high, the results also show that balance accuracy varies greatly between study subjects, which shows how complicated the problem actually is. However, at this point, the results are preliminary as the data set contains only seven study subjects, so these do not cover all migraine types. If the findings of this article can be confirmed in a larger population, it may potentially contribute to early diagnosis of migraine attacks.

Migraine is a poorly understood disease and it is estimated that 15% of people in Europe alone are affected by it. In this study skin electrodermal activity (EDA) signals of left and right wrists were collected from a migraine patient to see if the asymmetry associated with EDA signals will affect to migraine detection based on wearable sensors. In the study, nighttime EDA storm epochs were detected and visual inspection on total time of EDA storm epochs and timing of EDA storms between wrists were done. Also filtered EDA signals of nights that preceded migraine attacks were visually checked. According to the results, EDA measurements from one wrist are enough to detect changes before a migraine attack because the EDA asymmetry measured between wrists might play not a significant role in migraine prediction.

Background: The majority of young people do not meet the recommendations on physical activity for health. New innovative ways to motivate young people to adopt a physically active lifestyle are needed. Objective: The study aimed to study the feasibility of an automated, gamified, tailored Web-based mobile service aimed at physical and social activation among young men. Methods: A population-based sample of 496 young men (mean age 17.8 years [standard deviation 0.6]) participated in a 6-month randomized controlled trial (MOPO study). Participants were randomized to an intervention (n=250) and a control group (n=246). The intervention group was given a wrist-worn physical activity monitor (Polar Active) with physical activity feedback and access to a gamified Web-based mobile service, providing fitness guidelines, tailored health information, advice of youth services, social networking, and feedback on physical activity. Through the trial, the physical activity of the men in the control group was measured continuously with an otherwise similar monitor but providing only the time of day and no feedback. The primary outcome was the feasibility of the service based on log data and questionnaires. Among completers, we also analyzed the change in anthropometry and fitness between baseline and 6 months and the change over time in weekly time spent in moderate to vigorous physical activity. Results: Mobile service users considered the various functionalities related to physical activity important. However, compliance of the service was limited, with 161 (64.4%, 161/250) participants visiting the service, 118 (47.2%, 118/250) logging in more than once, and 41 (16.4%, 41/250) more than 5 times. Baseline sedentary time was higher in those who uploaded physical activity data until the end of the trial (P=.02). A total of 187 (74.8%, 187/250) participants in the intervention and 167 (67.9%, 167/246) in the control group participated in the final measurements. There were

In this work, it was tested if a computer game and wearable sensors could be used to motivate young men to increase their everyday activity. The game itself was planned so that the possibility to play the game and success in it was dependent of how much activity the subject achieved during his everyday life. A pilot version of a computer game, called Space Pioneer, was developed for this purpose while the activity measures were collected using three type of sensors: pedometer, Polar Active activity monitor or mobile phone application. The results were varying; although many subjects did not even try the game or disliked it, the awareness of subjects’ own physical condition as well as daily activity information were found out to be motivating factors in the target group.

In this study, a novel method to obtain user-dependent human activity recognition models unobtrusively by using the sensors of a smartphone is presented. The recognition consists of two models: sensor fusion-based user-independent model for data labeling and single sensor-based user-dependent model for final recognition. The functioning of the presented method is tested with human activity data set, including data from accelerometer and magnetometer, and with two classifiers. Comparison of the detection accuracies of the proposed method to traditional user-independent model shows that the presented method has potential, in nine cases out of ten it is better than the traditional method, but more experiments using different sensor combinations should be made to show the full potential of the method.

Inertial sensors are devices that measure movement, and therefore, when they are attached to a body, they can be used to measure human movements. In this thesis, data from these sensors are studied to recognize human activities user-independently. This is possible if the following two hypotheses are valid: firstly, as human movements are dissimilar between activities, also inertial sensor data between activities is so different that this data can be used to recognize activities. Secondly, while movements and inertial data are dissimilar between activities, they are so similar when different persons are performing the same activity that they can be recognized as the same activity. In this thesis, pattern recognition -based solutions are applied to inertial data to find these dissimilarities and similarities, and therefore, to build models to recognize activities user-independently. Activity recognition within this thesis is studied in two contexts: daily activity recognition using mobile phones, and activity recognition in industrial context. Both of these contexts have special requirements and these are considered in the presented solutions. Mobile phones are optimal devices to measure daily activity: they include a wide range of useful sensors to detect activities, and people carry them with them most of the time. On the other hand, the usage of mobile phones in active recognition includes several challenges; for instance, a person can carry a phone in any orientation, and there are hundreds of smartphone models, and each of them have specific hardware and software. Moreover, as battery life is always as issue with smartphones, techniques to lighten the classification process are proposed. Industrial context is different from daily activity context: when daily activities are recognized, occasional misclassifications may disturb the user, but they do not cause any other type of harm. This is not the case when activities are recognized in industrial context, Tiivistelmä Liikettä mittaavista antureista, kuten kiihtyvyysantureista, saatavaa tietoa voidaan käyttää ihmisten liikkeiden mittaamiseen kiinnittämällä ne johonkin kohtaan ihmisen kehoa. Väitöskirjassani tavoitteena on opettaa tähän tietoon perustuvia käyttäjäriippumattomia malleja, joiden avulla voidaan tunnistaa ihmisten toimia, kuten käveleminen ja juokseminen. Näiden mallien toimivuus perustuu seuraavaan kahteen oletukseen: (1) koska henkilöiden liikkeet eri toimissa ovat erilaisia, myös niistä mitattava anturitieto on erilaista, (2) useamman henkilön liikkeet samassa toimessa ovat niin samanlaisia, että liikkeistä mitatun anturitiedon perusteella nämä liikkeet voidaan päätellä kuvaavan samaa toimea. Tässä väitöskirjassa käyttäjäriippumaton ihmisten toimien tunnistus perustuu hahmontunnistusmenetelmiin ja tunnistusta on sovellettu kahteen eri asiayhteyteen: arkitoimien tunnistamiseen älypuhelimella sekä toimintojen tunnistamiseen teollisessa ympäristössä. Molemmilla sovellusalueilla on omat erityisvaatimuksensa ja -haasteensa. Älypuhelimien liikettä mittaavien antureihin perustuva tunnistus on haastavaa esimerkiksi siksi, että puhelimen asento ja paikka voivat vaihdella. Se voi olla esimerkiksi laukussa tai taskussa, lisäksi se voi olla missä tahansa asennossa. Myös puhelimen akun rajallinen kesto luo omat haasteensa. Tämän vuoksi tunnistus tulisi tehdä mahdollisimman kevyesti ja vähän virtaa kuluttavalla tavalla. Teollisessa ympäristössä haasteet ovat toisenlaisia. Kun tarkoituksena on tunnistaa esimerkiksi työvaiheiden oikea suoritusjärjestys kokoamislinjastolla, yksikin virheellinen tunnistus voi aiheuttaa suuren vahingon. Teollisessa ympäristössä tavoitteena onkin tunnistaa toimet mahdollisimman tarkasti välittämättä siitä kuinka paljon virtaa ja tehoa tunnistus vaatii. Väitöskirjassani kerrotaan kuinka nämä erityisvaatimukset ja -haasteet voidaan ottaa huomioon suunniteltaessa malleja ihmisten toimien tunnistamiseen. Väitöskirjassani esiteltyjä uusia menet

We are increasingly in situations of divided attention, subject to interruptions, and having to deal with an abundance of information. Our cognitive load changes in these situations of divided attention, task interruption or multitasking; this is particularly true for older adults. To help mediate our finite attention resources in performing cognitive tasks, we have to be able to measure the real-time changes in the cognitive load of individuals. This paper investigates how to assess real-time cognitive load based on psycho-physiological measurements. We use two different cognitive tasks that test perceptual speed and visio-spatial cognitive processing capabilities, and build accurate models that differentiate an individual’s cognitive load (low and high) for both young and older adults. Our models perform well in assessing load every second with two different time windows: 10 seconds and 60 seconds, although less accurately for older participants. Our results show that it is possible to build a realtime assessment method for cognitive load. Based on these results, we discuss how to integrate such models into deployable systems that mediate attention effectively.