Inspiration
Hearing from the challenges of rural medicine doctors speaking out online
What it does
Connects patients to specialists by utilizing a matching algorithm that best matches patient information to doctor speciality and history.
How we built it
We began with a simple web development phase that enabled the user to input data and format that data in a readable file; For the Retrieval Augmented Generation (RAG) model that matches the patient to the top 5 specialists, we utilized all-mpnet-base-v2 sentence transformer model to make text embeddings of the doctor database. Next, we utilized FAISS (similarity search algorithm) to compare the encoded text embedding of the patient data against the text embeddings of the doctors in the doctor database. A similarity score between the patient information and each doctor profile is given, and we rank them to output the top 5 specialists related to the patient profile. To interpret EHR data, we also utilized a second model called Bart Large CNN, which is a fine-tuned model dedicated to text summarization and information processing. This model will take the pdfs and/or text files as input, and it will summarize the information in the EHR as input for the RAG model to match with patients. In addition, we also made an SQL database to store the data of all the patients of each respective user.
Challenges we ran into
We had a challenge of finding a proper expert database consisting of the information that we needed. Additionally, our lack of UI/UX experience made developing the website model rather difficult so we decided to simplify such design using StreamLit.
Accomplishments that we're proud of
We are particularly proud of our implementation of the searching algorithm (FAISS) in combination with the RAG model in order to have a fully functioning prototype within 24 hours.
What we learned
We learned a significant amount about the actual problems that doctors, particularly those in rural areas face.
What's next for DocConnect
Improving the databases and connecting with healthcare insurance companies in order to provide DocConnect to those who need it.
Log in or sign up for Devpost to join the conversation.