Content area
Abstract
Peripheral arterial disease (PAD) causes blockage of the arteries, altering the blood flow to the lower limbs. The blockage of the arteries, during a walk, can cause the individual with PAD to feel severe pain located in the lower limbs. This pain disappears when the individuals stop walking. A recommended treatment to this disease is supervised physical exercise, which can prevent or mitigate the effects of PAD.
In the recommended treatments for PAD, there are exercise-based programs that can be done in hospital settings or outpatient settings. The program in a hospital environment requires the direct supervision of health professionals, which requires frequent travel by patients to hospital centers, implying high human resources costs and patient travel.
The outpatient exercise program allows patients to take the walks proposed by health professionals in the community, being supervised using information and communication technologies.This program is an economical solution to the supervision of physical exercise for individuals with PAD. However, it can be hard to evaluate the patients’ progress based only on communication.For that, a way to evaluate gait metrics is need.
The main contribution of this research is to come up with a solution that allows automatic detection of the onset of claudication based on the analysis of data acquired with the patients’smartphones.
This dissertation started by analyzing the methods used in the literature in other to be able to get gait parameters from the inertial sensors available in the smartphone. After that, we had a more straightforward path to work on before looking into the models. First, we need to filter and rotate the data to reduce any noise caused by the system or the user. Then we need to extract the gait cycle, which is the basis for many parameters. After that, we create our feature dataset to select the best models to identify the onset of claudication.
The results show that we can indeed detect claudication onset, based on inertial sensors, with some gait parameters. More tests are needed to be done on real-world data. We finished with a model with 95.58% accuracy for our dataset.
