Early-Detection of Austim using AI

When Austim(ASD) goes undetected it hinders an individual from reaching their full potential. Hence I created an AI+Autism Early Diagnosis tool as a fast and easy way to predict ASD early in children.

Comment

Inspiration

I was inspired to make a difference, after going to rural areas of Southeastern Asia, and learning about the lack of early diagonisis for autism spectrum disorder in children especially in rural areas, in terms of economic and access care.

What it does

Autism spectrum disorder(ASD) is a neurological disorder caused by differences in the brain. The most common way to detect autism in children is ADI-R. It helps doctors and their patients detect autism and begin treatment. This effective method can cost upwards of 5,000 USD for families. This is more than an entire month's salary to most people in developing countries. To reduce the financial burden while still having the same benefits as an ADI-R, I plan to create an AI model that can detect autism early in children.

How I built it

The backend of the code was developed using python, and the sklearn library. The frontend end was developing using streamlit.

Challenges I ran into

Initially I didn't know which sklearn model was better, so I trusted the model that performed better in the last few rounds. Each mistake divides the trust for the model into A and B. “A” happens when I have the error of the model predicting "Autistic" instead of "Non-Autistic". “B” happens when I have the "Non-Autistic " instead of "Autistic". If i believe A is less important than error B, I set error B to have a higher penalty score. Then I finalized the weights depending on the weight error. In this case, having the model predict autism in a non-autistic person is better than an autistic person as a non autistic. By adopting this strategy, I choose the most accurate model.

Accomplishments that I'm proud of

Learning the Streamlit platform was a new hurdle for me to overcome. This was also my first project where I coded both the front and backend.

What I learned

I learned how to use machine learning algorithms, and deploy it to an App.

What's next for AI+Autism Early Diagnosis tool

I will like to consider text written by the individual combined with the AI model I created. I also like to consider the way a person talks, as a way to predict Autism.

Built With

Share this project:

Updates