NurseGPT

Recently there has been a trend in healthcare towards Ai-based diagnostics, which utilize ML/DL to diagnose patients based on EHR, imaging, etc. But in order to use them, there has to be a complex interface that is difficult for most users to navigate, and in most cases, it’s designed in english, which limits their usability for non-english speakers. Our project aims to utilize the immense power of LLMs to facilitate the communication between the end-user (patient/ healthcare provider) and these large diagnostic tools. By using a simple chatbot. How it works: In its current form, the chatbot is connected to 2 machine learning classifiers (Cardiovascular disease risk / Maternal health risk). The basic use case is this: Users are asked to describe their symptoms in free form text. First message is interpreted and parsed by the LLM framework (langchain) and then checked if they’re sufficient. If not, the bot asks the user about the missing data points. When data is sufficient, it’s sent to classifiers. After the classification process, the result is sent back to the user in a properly formatted message. Who it is for: Main focus would be the telehealth market in africa, which is expected to reach $22 Billion in 2030 in the Saudi Arabia; UAE; South Africa region alone, The target market includes: personalized checkups, where users can learn about their risks. Chronic care: the bot would help with updates, medication adherence, and other care aspects. Healthcare providers, who would use this to help with the throughput of outpatient clinics in times of pressure on the system. Like what happened during the COVID-19 crisis. What the fully realized version would be: With the new release of GPT-4, this tool could be expanded to accept images (X-ray, CT, etc.) inside its context, which will enable the bot to comprehend the full medical history of a patient, this will enable greater question answering, and will be more useful as a personalized telehealth tool.