MindLite

Patient - Home
Doctor - Manage Patient
Home Screen
Doctor - AI/ML Scoring
Patient - Games
Register
Login

DEVPOST PROJECT

Inspiration

Our inspiration for this project was our professor who had to retire from her teaching career due to Alzheimer. Alzheimer's begins damaging the brain 10 to 20 years before the first symptom appears. We are diagnosing people after the damage. That's the crisis. That is why we built MindLite an AI-powered, clinically informed, web-based platform that catches cognitive decline before it becomes visible.

What it does

The patients log in and play five short games — Family Recognition, where they identify faces and relationships. Pattern Recognition, spotting sequences. Memory Match, pairing symbols for speed and recall. Word Recall, reproducing words after a delay. And a Reaction Time test. After at least 10 days of data, our system generates a Z-Score — a statistical measure of how far a person's performance deviates from their own baseline. Not from a population average. From their own normal. If the Z-Score drops suddenly or trends downward gradually — Smart Alerts fire automatically to the assigned neurologist and the patient's guardian. Catch it early. Act early.

How we built it

We built the front end on node.js the backend on python and the ml on Google Collab by using random forest regressor.

Challenges we ran into

As newbies, our biggest obstacle was selecting the right tech stack because it all felt right initially but after working on each one we realised their flaws.. The biggest challenge we ran into was getting dataset for training our model and also reading through various research papers to just get the right games for our model .

Accomplishments that we're proud of

By far the best accomplishments we are proud of are being able to present this to our professor who retired and listening "you guys did it hahahah" from her, truly makes all the efforts worth it.

What we learned

We learned a hell lot of stuff new API's, new models, integrating everything but more than that we learned how to work as a team being fellow coders, how to assign bits and pieces to everyone and then integrate all of it.

What's next for MINDLITE

Our next destination is going to be refining our approach and training a better model on crowdsourced data by letting people use it. Next, we plan to integrate wearables combining cognitive scores with sleep and heart-rate data. We'll add speech biomarkers analyzing voice recordings for pauses and vocabulary patterns. And we'll deploy directly to senior centers and rural communities across India.