|
Title:
|
ENHANCING PERSONALIZED DIABETES TREATMENT
WITH LARGE LANGUAGE MODELS AND
CHAIN-OF-THOUGHT REASONING |
|
Author(s):
|
Qi Sun, Xuekuan Fu and Chenyang Zhou |
|
ISBN:
|
978-989-8704-61-0 |
|
Editors:
|
Demetrios G. Sampson, Dirk Ifenthaler and Pedro Isaías |
|
Year:
|
2024 |
|
Edition:
|
Single |
|
Keywords:
|
Large Language Model, Chain-of-Thought, Diabetes Treatment, Data Augment |
|
Type:
|
Short |
|
First Page:
|
359 |
|
Last Page:
|
364 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
if you are a member please login 
|
|
Paper Abstract:
|
Diabetic patients often struggle to determine the appropriate dosage of short-acting insulin to effectively manage their
blood glucose levels. Miscalculating the insulin dose can lead to serious complications, such as hypoglycemia or
hyperglycemia. In this paper, we propose a novel approach for personalized insulin treatment using large language models
(LLMs), dubbed L4DT (Large Language Models for Diabetes Treatment). This method can be divided into two phases.
The first phase involves applying chain-of-thought reasoning for data augmentation, simulating so that L4DT could learn
the step-by-step thought processes that human experts would use to determine optimal insulin dosages. The second phase
focuses on training the LLM to obtain personalized insulin dosage recommendations. Our evaluation of the L4DT
demonstrates its expertise in insulin dosage prediction. On the MIMIC-IV dataset, the L4DT model achieves a mean
squared error of 4.55 and a mean absolute error of 2.01, outperforming existing approaches. This study not only enhances
the application of exploratory learning approaches in complex medical domains but also assesses the impact of exploratory
technologies like LLMs on diabetes treatment. The integration of technology and expertise in this model offers a reliable
reference for clinicians and a platform for continuous learning and expertise development in the field of diabetes
management. |
|
|
|
|
|
|
