In psychiatric practice, therapists work with many patients, and each one has a dedicated file where their progress is documented. For private therapy, the standard protocol usually includes weekly sessions, pharmacological treatments, and, occasionally, social initiatives to aid the patient’s well-being.
This project aims to develop an AI agent that assists psychiatrists throughout the therapy process. The AI will have access to patient files, which include the doctor’s notes and recorded dialogues. The agent will be able to:
- Provide reports on the medications a patient is taking.
- Report specific events in the patient’s life (as noted in the files).
- Compare the condition of one patient with another.
- Offer suggestions on the most appropriate medications.
- Recommend the best next course of action based on the patient’s current condition and evidence.
Resources:
- Model: GPT-3.5 integrated with Azure Cognitive Services (RAG)
- Programming Language: Python
- Fine-Tuning: Dataset from Hugging Face: PsycholexEval, though additional refinement is necessary.
This project combines natural language processing with clinical data to provide meaningful insights and therapeutic recommendations, helping doctors improve patient care through AI assistance.