Identification of patient deterioration by means of physical activity changes in patients with COPD using Fitbits
With COPD, the less a patient does, the less they are capable of doing. With deteriorating lungs, respiratory disorders, shortness of breath, a patient can only do limited physical activity before they feel exhausted. simple tasks like cooking or grocery shopping can exhaust them. It might seem impossible for a patient with COPD to get any physical exercise at all, but it is extremely necessary that they get a certain amount of exercise every day. Exercising helps in managing the disease to a great extent. Low or decreasing physical activity can be closely associated with deteriorating lung function and will most likely lead to hospitalization. With today’s technology, it has become very convenient to track a person’s physical movement every second of the day. The research carried out at KU Leuven involves COPD patients equipped with smart bands to track their physical activity. The work highlighted in this master’s thesis involves extracting the data from smart bands and using Machine Learning algorithms to predict the need of hospitalization of any patient by analysing their physical activity. Three approaches were presented, supervised classification with and without handling class imbalance in all the three approaches. The results were not satisfactory in the first approach, whereas in the second and the third approach the results were slightly better. This work thereby shows that this problem of predicting the need of hospitalization can be solved using those approaches with necessary optimizations.
