Core Concepts
Diet-ODIN bridges dietary patterns and opioid misuse detection through a novel framework.
Abstract
Abstract:
The opioid crisis in the US is a critical concern.
Medication-assisted treatment (MAT) has side effects triggering relapse.
Dietary nutrition intervention is crucial but under-explored.
Diet-ODIN establishes a dataset and framework for opioid misuse detection.
Introduction:
MAT is effective but has side effects leading to relapse.
Dietary nutrition plays a role in prevention and recovery of opioid misuse.
Diet-ODIN bridges dietary and health information for detection.
Related Work:
Heterogeneous Graph Neural Networks are effective in graph learning tasks.
Large Language Models enhance reasoning capabilities across domains.
Research on opioid misuse detection includes quantitative and qualitative approaches.
NHANES Dietary Graph:
Defines heterogeneous graphs, meta-paths, and neighborhoods.
Constructs NHANES Dietary Graph from food intake, habits, and drug usage data.
The Framework of Diet-ODIN:
NR-HGNN for Detecting Users with Opioid Misuse:
Shared dietary pattern learning with macro-level aggregation.
Individual dietary habit learning with micro-level aggregation.
Graph refinement with noise reduction for detection.
Bridging NR-HGNN and LLM for Interpretation:
Strategies to enhance domain-specific reasoning using prompts.
Statistical analysis panel to interpret key dietary patterns associated with opioid misuse.
Experiments:
NR-HGNN outperforms existing methods in detecting users with opioid misuse.
Stats
MAT is recognized as the most effective treatment for opioid misuse (MATはオピオイドの誤用に対する最も効果的な治療として認識されています)
10.1 million Americans reported misusing opioids in 2019 (2019年には1,010万人のアメリカ人がオピオイドを誤用していると報告されました)
An estimated 108,000 drug overdose deaths in the United States in 2021 (2021年にはアメリカで推定108,000件の薬物過剰摂取死亡が発生しました)
Quotes
"The results are truly inspiring. ... It really addresses roots of issues." (「結果は本当に刺激的です。... 問題の根源に真正面から取り組んでいます。」)