Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Introduction: More than 65 million people worldwide suffer from COPD. An important aspect of the disease are exacerbations which have a significant impact on a patient’s health status and include substantial healthcare costs. Present-day management of COPD exacerbations aims at minimizing the negative impact of the current event and preventing future events. To improve current management, increasing attention is going to the prediction of acute exacerbations of COPD (AECOPD) by means of remote monitoring of COPD patients. This thesis explores the possibility of linking abnormal physical activity (measured as step count) to acute events using data from the Mr PAPP study. Objectives: Three objectives were identified: (i) developing a risk profile based on baseline characteristics such as pulmonary function tests, health questionnaire scores and 6-minute walk distance, (ii) developing an imputation method for step data imputation, and (iii) implementing three unsupervised anomaly detection methods for identification of abnormal physical activity indicative for acute events. Methods: Data from the Mr PAPP study, a three-month telecoaching program for patients with COPD, was used. This data set provided characteristics data and daily step count data. A risk profile, i.e. most important predictors of having an AECOPD during the study, was developed using logistic regression and Cox proportional hazards regression analysis. Abnormal physical activity was detected using (i) a moving window approach which used the moving mean and standard deviation for anomaly detection, (ii) a seasonal trend decomposition based on LOESS (STL) approach, and (iii) the Isolation Forest algorithm on transformed step data. Results and Conclusion: Pulmonary function test results and history of COPD exacerbations were the most important predictors of having an AECOPD during the Mr PAPP study. Both FEV1 and the number of exacerbations of the year prior to the study were found statistically significant in the logistic regression model and the Cox proportional hazards model. The developed imputation method performs well for imputation of small gaps in the step data but not for imputation of large gaps. Unsupervised point-anomaly detection methods such as the moving window approach and the STL approach seem less suitable for detection of abnormal physical activity prior to an exacerbation. An unsupervised anomaly detection algorithm applied to moving window extracted features is therefore preferred. For further research, this thesis suggests using a supervised classification algorithm for identification of abnormal physical activity in step data indicative for acute events.

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Worldwide, approximately 20% of the people suffers from an allergy. In Belgium, an estimated 10% of the population are allergic to tree pollen of the Alder, Birch and Hazel and an estimated 18% of Belgians are allergic to grass pollen. During the pollen season, these people regularly experience symptoms such as a running nose, sneezing and itching and watery eyes. In general, allergy sufferers report a lower quality of life. In order to improve their well-being, an efficient control of their allergy symptoms is crucial. There exist several smartphone applications that provide information about the current risk on allergic reactions and that allow patient to fill in a medical 'allergy' diary and they seem to help allergy sufferers to better manage their symptoms. However, there is much room for improvement since only a minority of the available apps targets specific skills such as management of acute symptoms while a majority only provide basic information about the allergy. A possibility to improve allergy sufferers' ability to manage their symptoms is the use of wearable technology such as smart watches or sports watches that can measure heart rate (HR). The RespirIT project collected data about heart rate, activity and acute allergic reactions of persons that are allergic to tree pollen by means of an innovative tracking system that consisted of a sports watch (HR), a smartphone (activity) and a smartphone application (acute allergic reactions). This data was used to study the relation between allergic reactions and heart rate. Since physical activity causes changes in heart rate, e.g. running increases heart rate compared to sitting, it needs to be taken into account to study the effect of allergic reactions on the heart rate. Thus, the possible heart rate changes due to allergic reactions need to be differentiated from those caused by physical activity. This is done by describing the heart rate in function of the activity by a mathematical model. If an allergic reaction would cause HR changes, the model would fail to simulate the HR accurately and larger differences between the modelled HR and measured HR (model errors) would occur. The results of this thesis were not able to show that allergic reactions cause HR changes in data collected by a sports watch. Additionally, more severe symptoms did not produce higher model errors, thus, the results did not provide evidence that more severe allergic reaction can be detected more accurately. However, there were several limitations of the dataset and methods that can be avoided in further research.