Inspiration
My brother is a fellow of child and adolescent psychiatry. He has up to 298 diagnoses to factor in when viewing a patient. He has a system for keeping track of criteria, but he came to me to automate it.
What it does
It takes in patient responses to questions and spits out a differential diagnosis, which is list of possible diagnoses that align with their symptomology.
How we built it
I built a "Character Guesser" AI with Python and Flask. The user thinks of a character from popular media, then the AI guesses what the answer could be, quickly. We're currently adapting this model to not guess characters, but diagnoses.
Challenges we ran into
The current challenge we are running into is building a model that replicates how a clinician thinks. Doctors do not use algorithms to make a diagnosis, but instead, intuition. Intuition is something that is very hard to replicate in code.
Accomplishments that we're proud of
We have one clinician (and more to come) who is excited to use this product. We're applying principles we learned in our classes to build a real world product!
What we learned
I have learned that you can never learn everything there is to know about coding. It is important to rely on other people to help build a project at scale.
What's next for DiffDiagnose
Translate our Character guesser model to a mental health disorder diagnostic model. Once we have that, the front end and back end will fall in place to support our algorithm!
Log in or sign up for Devpost to join the conversation.