Update README.md

README.md

@@ -7,10 +7,15 @@ tags:
 ---
 
 
-# NOTE
-
-Notable generation Of patient Text summaries through an Efficient approach based on direct preference optimization (DPO)
+# NOTE: Notable generation Of patient Text summaries through Efficient approach based on direct preference optimization
 
+The discharge summary (DS) is a crucial document in the patient journey, as it encompasses all events from multiple visits, medications, varied imaging/laboratory tests, surgery/procedures, and admissions/discharge. 
+Providing a summary of the patient’s progress is crucial, as it significantly influences future care and planning. 
+Consequently, clinicians face the laborious and resource-intensive task of manually collecting, organizing, and combining all the necessary data for a DS. 
+Therefore, we propose NOTE, which stands for “Notable generation Of patient Text summaries through an Efficient approach based on direct preference optimization (DPO)”. 
+NOTE is based on MIMIC-III and summarizes a single hospitalization of a patient. Patient events are sequentially combined and used to generate a DS for each hospitalization. 
+To demonstrate the practical application of the developed NOTE, we provide a web page-based demonstration software. In the future, we will aim to deploy the software available for actual use by clinicians in hospital.
+NOTE can be utilized to generate various summaries not only discharge summaries but also throughout a patient's journey, thereby alleviating the labor-intensive workload of clinicians and aiming for increased efficiency.
 
 
 ## Model Description
@@ -117,13 +122,24 @@ evaluation strategy | step | step
 eval step | 10 | 5 
 
 
-## Applicability in medicine
+### Experimental setup
+- **Ubuntu 20.04 LTS**
+- **2 NVIDIA GeForce RTX 3090 GPUs**
+- **Python**: 3.8.10
+- **Pytorch**:2.0.1+cu118
+- **Transformer**:4.35.2
 
 
+## Applicability in medicine
 
-## Limitations
+Our NOTE has generated a sequential dataset by considering both table data and text data, which can aid in patient-specific report generation. 
+Furthermore, it suggests that within medical institutions where medical data import/export is challenging, it is possible to self-tune and utilize LLMs, proposing this through demo software.
 
+## Limitations
 
+The model was limited in character count for comparison with the existing T5 model, but it is planned to be expanded in future research.
+Additionally, further research on prompting engineering is needed due to it producing different results with the same instructions.
+Most metrics for evaluating summarization and generation tasks were somewhat challenging to apply to our study, and while we attempted to address this through the ChatGPT4 Assistant API, future research will be based on feedback from clinicians.