Related Experiment Video
Updated: Sep 13, 2025
Author Spotlight: Impact of Intergenic Interactions on Disease-Identifying Dark Biomarkers
Published on: March 1, 2024
A Multi-Stage Framework for Kawasaki Disease Prediction Using Clustering-Based Undersampling and Synthetic Data
Heng-Chih Huang1, Chuan-Sheng Hung1, Chun-Hung Richard Lin1
1Department of Computer Science and Engineering, National Sun Yat-sen University, Kaohsiung 80424, Taiwan.
Insights
This study introduces an AI framework to improve early detection of Kawasaki disease (KD) using routine lab tests. The AI model effectively balances high sensitivity and specificity for practical clinical screening.
Area of Science:
- Pediatric Rheumatology
- Artificial Intelligence in Medicine
- Diagnostic Accuracy
Background:
- Kawasaki disease (KD) is a critical pediatric vasculitis with challenging diagnosis due to varied symptoms.
- Delayed diagnosis of KD can lead to severe cardiovascular complications.
- AI leveraging routine laboratory tests offers a potential solution for early KD detection.
Purpose of the Study:
- To develop and validate a multi-stage AI framework for enhanced early detection of Kawasaki disease.
- To address the challenge of class imbalance in AI models for rare diseases like KD.
- To improve the diagnostic accuracy and clinical utility of AI in identifying KD.
Main Methods:
- A multi-stage AI framework incorporating clustering-based undersampling, data augmentation, and stacking ensemble learning was developed.
- The model was trained and tested on a large dataset (n=74,641) from Chang Gung Memorial Hospital (CGMH).
- External validation was performed on an independent dataset (n=1582) from Kaohsiung Medical University Hospital (KMUH) to ensure generalizability.
Main Results:
- The AI model achieved 95% recall with 97.5% specificity and a 53.6% F1-score on the internal test set (CGMH).
- On the external validation set (KMUH), the model demonstrated 74.7% specificity and a 23.4% F1-score at 95% recall.
- The framework successfully maintained high specificity while prioritizing sensitivity, indicating practical predictive performance.
Conclusions:
- The proposed AI framework shows significant promise for improving the early detection of Kawasaki disease.
- The model's ability to balance sensitivity and specificity makes it a valuable tool for real-world KD screening.
- This AI-driven approach can aid clinicians in overcoming diagnostic challenges associated with Kawasaki disease.
Abstract:
Kawasaki disease (KD) is a rare yet potentially life-threatening pediatric vasculitis that, if left undiagnosed or untreated, can result in serious cardiovascular complications. Its heterogeneous clinical presentation poses diagnostic challenges, often failing to meet classical criteria and increasing the risk of oversight. Leveraging routine laboratory tests with AI offers a promising strategy for enhancing early detection. However, due to the extremely low prevalence of KD, conventional models often struggle with severe class imbalance, limiting their ability to achieve both high sensitivity and specificity in practice. To address this issue, we propose a multi-stage AI-based predictive framework that incorporates clustering-based undersampling, data augmentation, and stacking ensemble learning. The model was trained and internally tested on clinical blood and urine test data from Chang Gung Memorial Hospital (CGMH, n = 74,641; 2010-2019), and externally validated using an independent dataset from Kaohsiung Medical University Hospital (KMUH, n = 1582; 2012-2020), thereby supporting cross-institutional generalizability. At a fixed recall rate of 95%, the model achieved a specificity of 97.5% and an F1-score of 53.6% on the CGMH test set, and a specificity of 74.7% with an F1-score of 23.4% on the KMUH validation set. These results underscore the model's ability to maintain high specificity even under sensitivity-focused constraints, while still delivering clinically meaningful predictive performance. This balance of sensitivity and specificity highlights the framework's practical utility for real-world KD screening.
More Related Videos
09:47Author Spotlight: Advancing Alzheimer's Research – Exploring Early Detection and Multi-Omics Approaches
Published on: December 15, 2023
05:47Evidence-based Knowledge Synthesis and Hypothesis Validation: Navigating Biomedical Knowledge Bases via Explainable AI and Agentic Systems
Published on: June 13, 2025
Related Concept Videos
01:20Cluster Sampling Method
To choose a cluster sample, divide the population into clusters (groups) and then randomly select some of the clusters. All the members from these clusters are in the cluster sample. For example, if you randomly sample four departments from your...
01:06Model Approaches for Pharmacokinetic Data: Distributed Parameter Models
The distributed parameter models are specifically designed to account for variations and differences in some drug classes. This model is particularly useful for assessing regional concentrations of anticancer or...
01:24One-Compartment Open Model: Wagner-Nelson and Loo Riegelman Method for ka Estimation
On...
00:59Model-Independent Approaches for Pharmacokinetic Data: Noncompartmental Analysis
One important characteristic of noncompartmental analyses is that drug exposure increases proportionally with increasing doses. This...