Inspiration
This was our team’s first time at Hackalytics, and we wanted to tackle a condition that affects the most vulnerable in our society—Parkinson’s disease. Parkinson’s is a neurodegenerative disorder that currently has no cure, and its symptoms, including tremors, stiffness, dementia, and impaired movement, progressively worsen over time. With millions of individuals living with Parkinson's worldwide, it remains a challenge for both patients and caregivers to monitor and manage the disease effectively. Those with Parkinson’s experience a wide range of symptoms, from impaired mobility to dementia. These symptoms lead to severe, and oftentimes, fatal consequences. About 60% of individuals with Parkinson’s will fall every year. Of those who fall, individuals diagnosed with Parkinson’s are four times more likely to experience fractures. These falls are not only dangerous but can lead to a significant decline in health, quality of life, and independence, making it a critical area to address in Parkinson's care. Another key factor of Parkinson's is dementia. Approximately 80% of patients with Parkinson’s will have developed Dementia ten years into their diagnosis. According to the Alzheimer’s Association, 74% of people with dementia have wandered home from walking, bicycling, driving, or on public transportation. Of those found within two days, only 51% survived, longer than that and the percentage drops to lower than 20%. Given these alarming statistics, we realized that early detection through proactive AI and rapid responses were crucial to improving the safety and well-being of those living with Parkinson's.
What it does
Our product, LOCUS, is designed to enhance the safety, monitoring, and quality of life for individuals living with Parkinson's disease. It collects real-time data from wearable devices, such as accelerometers, and tracks the user's location. The software detects falls and monitors physical activity, identifying any significant deviations from normal behavior, such as wandering. When an emergency occurs, like a fall or wandering incident, LOCUS triggers an automated SMS system to notify the user’s contacts with their exact location. The information is also made available to local law enforcement and healthcare services for a rapid response. Overall, LOCUS helps provide constant updates and early detection of potential risks, ensuring timely intervention and support.
How we built it
We developed LOCUS, a software solution to improve the safety and monitoring of individuals with Parkinson’s disease. The system collects location data and accelerometer data from wearable devices to track physical activity and potential risks. MongoDB was chosen for efficient data storage, handling a variety of data like location, activity, and patient details. The accelerometer data was preprocessed using pandas and a standard scaler to prepare it for AI model training. We trained a Random Forest Classifier on the processed data to detect falls and differentiate activities, achieving 98% accuracy. Google APIs were integrated for geolocation and mapping, allowing real-time tracking of users' positions. The TeleSign API was used to automate SMS notifications, sending alerts about the user’s location to contacts and emergency services in case of an incident.
Challenges we ran into
Since we had such a wide range of data, we ended up using different clusters in MongoDB. This caused a lot of errors when merging and pushing onto GitHub. The SMS API’s had a lot of issues, the format of the phone numbers in our database made the SMS API repeatedly error.
Accomplishments that we're proud of
Our entire team is composed of first-years, and we are short a member, so just the fact that we were able to successfully complete a project in the timeframe. Furthermore, we found this idea to be novel and felt that we were able to effectively utilize many languages to accomplish our goal.
What we learned
This project was a combination of different tools and APIS. We had to collaborate to combine Google API, and Telesign API, and manage a wide range of data using Node.JS and MongoDB. This was also the first time we trained a machine learning model, and working out the issues was a rewarding experience.
What's next for Locus
Because of MongoDB’s restrictions on free accounts, we were capped on how much data we could hold. We want to scale our model so that it can process and predict information for thousands of people who have Parkinson’s. We plan to build an AI model that can determine based on previous routes/paths from individuals if they are wandering. Also, we want to implement infrastructure to directly receive and process data from electronic devices that are inputted into our model.
Log in or sign up for Devpost to join the conversation.