Visualizing Parkinson's Disease
Goal:
Provide a dashboard for patients and medical professionals to keep track of a patient’s Parkinson’s Disease progression with data gathered from accelerometers and gyroscopes attached to a pair of shoes.
Product:
*Note: there is absolutely no PII shown below and client logos were scrubbed beforehand.
Admittedly there was a slight oversight of not centering the left accelerometer’s z-axis to zero.
With an accelerometer and gyroscope attached to a person’s shoes, we can track their movements to get a quick overview of any obvious walking abnormalities. In this case, my coworker attached this device to his own shoes and simulated gait freezes.
The red text was a desired feature that wasn’t implemented in time - the idea was to have alerts that would identify gaits that were considered outlier to the average shape of a person’s gait.
Taking the acceleration data from the devices we can dive further into the data by separating each cycle by gait and overlay those on top of each other to identify freezes or abnormal strides. There is one obvious abnormal stride during the 9th stride where acceleration seems to be delayed and stutter.
By using the timestamps within the dataset we can derive the time taken per stride, which allows us to calculate velocity by using the acceleration and time metrics. Average velocity was calculated after every stride and compared to the population average walking velocity of the patient’s gender and age group. If this average was not met within a specific threshold, there would be a visual cue to signal an issue.
Using both the acceleration and time, distance was also calculated per stride, which allowed for step analysis. The distance between each step was plotted - displaying a person’s normal step. Any steps which had too little distance between each other were to be raised as alerts.
Afterword:
This is to be taken as mere proof of concept as to what we can extract with limited data. There were obvious issues with the implementation of tracking a person’s movements to begin with, such as only having sensors attached to the shoes, instead of a network of sensors on multiple pivotal body parts such as the hip and knees.