Inspiration

A big problem that exists and is most prevalent in cancer is the idea that people developed these diseases and conditions but never get it checked up on, because no one knows that they have one of these diseases until it is too late. The whole spectrum of neurodegenerative diseases matches this phenomenon. Diseases related to dementia have symptoms that are present way before the average time that someone suspects they have this disease. When shown the power of Alexa, that idea came to us that we could utilize Alexa, a product that is increasing in popularity especially amongst the older generations, to be able to utilize predictive analytics to foresee potential diseases or conditions. With neurodegenerative diseases, the symptoms are usually changes in the speaking patterns and habits that an individual has. Hence, we believed that utilizing this along with machine learning would be an amazing way to predict the likelihood of these devastating conditions.

What it does

This Alexa skill allows for analysis of speech attributes such as pauses, repeated words, and unintelligible words, which are attributes of speech known for being among the first victims of neurodegenerative diseases like Alzheimer's. With this information, we can monitor significant worsening of speech over time, notify users that they have exhibited symptoms of Alzheimer's and recommend that they see a doctor for a second opinion.

How we built it

Our projects had many parts. The analytical part of the project was utilizing Python and machine learning through scikit-learn to develop an algorithm given a lot of training data to predict if an individual has a form of dementia given speech to text translated transcripts. From there that input was received (somewhat) from Amazon Alexa where she would be able to respond to user prompts. This was done through a combination of node.js on the Amazon Developer platform and AWS Lambda to achieve this. The combination of the two allows for an analytical mechanism to interact over time with individuals to predict their health and well being; the use case for our project was predicting the presence of dementia given a certain confidence interval along with having some cool health-related features integrated into our Alexa Skill. This is what constituted of our project, Echo MyHealth.

Challenges we ran into

Because we aren’t Amazon, we could not create the dream implementation of the idea. Ideally, the machine learning algorithm could run passively upon any request without having to open the MyHealth skill. Instead, we created a proof of concept in the form of an Alexa Skill.

Accomplishments that we're proud of

Using the machine learning algorithm, we are able to accurately predict instances of Alzheimer’s with an accuracy of 70 percent, with nothing but vocal recordings. Along with that, our team had no experience with Javascript and limited machine learning experience. None of our team knew each other coming into the Hackathon and we created something super awesome and that we enjoyed doing.

What we learned

We learned so much so it was really awesome. We all learned aspects of node.js and javascript that none of us knew before. We were also exposed to creating Alexa Skills which was so exciting because of the power developers have.

What's next for Echo MyHealth

We did a lot in 36 hours but there was more we wanted to do. We wanted to better increase the interactions between our analytical half and our Alexa half. Moreover, we want to be able to implement more features that can help us increase the accuracy of our machine learning algorithm. Lastly, we also would love to make Echo MyHealth a more immersive platform with more functionality given more time.

Built With

Share this project:
×

Updates