AI-Based Medical Document Summarization Using NLP

Abstract: Hospital have to handle a lot of paperwork these days. Doctors write down lots of notes labs make test reports. Patients get papers when they leave the hospital or clinic. All this information, about care is important but the thing is it is not easy to look at.Doctors have to spend a time reading these reports. Medical paperwork is a problem. Patients often go home with papers they do not really understand.

Hospitals and clinics have to find a way to make medical paperwork easier to deal with.This situation is something we have personally noticed and wanted to work on. In this paper, we look at how Artificial Intelligence and Natural Language Processing can be used to make medical documents easier to work with for everyone involved. We reviewed four recent research papers on this topic and based on what they showed, we are proposing a system that can take a medical report as input and give back two summaries — one written for the medical team and one written in plain language for the patient. The system also generates discharge instructions and includes a chatbot this allows users to ask questions about the content they have just read. We use models like BioBERT, FLAN-T5, GatorTronGPT, and Longformer, together with a RAG setup for accuracy and ICD/SNOMED checks to make sure the medical content is correct. The papers we looked at have results for ROUGE-L, BERTScore and clinical accuracy. This shows that a system, like this can work well and be helpful.

Keywords: Medical Document Summarization, Natural Language Processing, BioBERT, FLAN-T5, Retrieval-Augmented Generation, Clinical NLP, Discharge Summarization, Transformer Models, MIMIC-III, Healthcare AI.