Mentis.ai

Introducing Mentis.ai, a program that can improve your wellbeing with deep insights about your mind with a neural network.

Comment

Inspiration

My inspiration for this project was the number of people in my family that struggle with neurodegenerative diseases like dementia. I wanted to learn more about dementia, the factors that cause it and the different ways that AI can be used in healthcare to prevent diseases like dementia.

What it does

The program I wrote first asks the user questions about their lifestyle and genetics, then a Keras neural network takes this data from the user and predicts if the user has a higher likelihood of developing dementia. I also have a different program that visualizes and explores the data from a dementia patient dataset to show the user how lifestyle and genetic factors influence the likelihood of developing dementia. The data analysis explores the connection between mental health and dementia, the connection between genetics and dementia, and which factors are the most influential on dementia risk. The predictions and the data visualization allow anyone to assess their future risk for dementia and gain awareness on different ways they can reduce their risk of dementia. My solution can help people earlier in their life take preventative measures to slow the progress of dementia. Whether it be encouraging someone to be more active or reduce their consumption of drugs and alcohol, this program can have both short term and long term benefits. Short term, the user can have an improved lifestyle, and long term, the user could slow the effects of dementia and the effects of other diseases as well.

How I built it

First, I started exploring a dataset in Kaggle about patients who either did or did not have dementia. The dataset had many features detailing different aspects of the patients' genetics and lifestyle choices. I used matplotlib to visualize and explore this data. I then found the features that had the greatest influence on dementia risk. I took the top 5 most influential features and decided to use them to make a neural network that predicts if someone is more likely to get dementia. I then built a neural network using Keras. Then I started tweaking the model and changing the number of layers, the number of nodes and the activation functions, to optimize the accuracy. Once the model was trained and tested, I coded a function that would ask the user questions about their lifestyle and genetics to gain data to input into the neural network.

Challenges I ran into

One challenge I ran into was that I have never coded a neural network before, and it was challenging at first to understand how to do it. I also was having trouble understanding how to improve the accuracy score, because for a long while the score was stuck at around 50% and wasn't changing. I had been using all the features in the dataset to train the model, but when I switched to only using the top 5 most influential factors, the accuracy score went up.

Accomplishments that I'm proud of

I'm proud of being able to make a neural network for the first time, with a 90% accuracy. I'm also proud of being able to work with such a large dataset and use visualization and statistics to get insights from the data.

What I learned

I learned more about Keras and neural networks. I also learned about dementia and the different factors that influence it.

What's next for Mentis.ai

In the future I would want to add a UI to improve the user experience. I would also want to ask more questions to the user about their lifestyle and genetics, since it is important to be able to gain a full picture on the user before making any predictions.

Built With

  • csv
  • kaggle
  • keras
  • matplotlib
  • numpy
  • python
  • sklearn
  • statistics
  • vscode
Share this project:

Updates