Allergy Risk No Longer a Mystery! Big Data from 1,500 Children Creates Personalized Allergy Prediction ModelJun 24, 2024

Can the risk of childhood allergies be predicted? Dr. Yao Zong-jie from Linkou Chang Gung Memorial Hospital has led a cross-disciplinary national team to develop a clinical decision system called "Personalized Risk Assessment of Pediatric Allergic Diseases Using Artificial Intelligence." By analyzing demographic, clinical, genetic, and bioinformatics data from 1,500 school-aged children and employing machine learning strategies, this system can accurately predict children's allergy risks with an accuracy rate of up to 97%. 

Allergies, such as asthma, allergic rhinitis, and atopic dermatitis, are common chronic conditions in children. They not only affect children's health and quality of life but also impose a heavy burden on families and society. Current allergy risk assessments mainly rely on doctors' experience, lacking objective standards, and assessments can vary between doctors. Traditional methods also fail to fully account for individual differences, making it difficult to provide precise personalized risk predictions. 

Multifaceted Data Analysis for Personalized Allergy Prediction Model 

At the heart of this decision system is a deep learning model that integrates and analyzes various types of data, including numerical and categorical data. Initially, the system cleans and calibrates the raw data from 1,500 school-aged children to ensure accuracy. It then encodes demographic data, clinical risk factors, epidemiological risk factors, and genetic factors into a format analyzable by the machine. Through the deep learning model, the system learns and analyzes how these complex data influence prediction outcomes, thereby constructing a personalized allergy risk prediction model. This model not only considers the impact of individual risk factors but also comprehensively evaluates the interactions between various factors, leading to more precise allergy risk predictions for individuals. Currently, the model has an accuracy rate of 97% in predicting atopic allergies and has demonstrated excellent stability through ten-fold cross-validation, making it resilient to data variations. 

Dr. Yao Zong-jie stated that this decision system not only accurately predicts children's allergy risks but also provides interpretable risk assessment results, helping doctors understand the factors influencing allergy risks. This understanding allows for the development of more precise personalized treatment plans. He emphasized that the application of AI will improve the diagnosis and treatment quality of pediatric allergic diseases, reduce the risk of misdiagnosis and treatment failure, and lighten the burden on healthcare teams. 

Upgrading AI in Healthcare: Potential Expansion to Online Platforms 

The team has integrated the model with web technologies to create an online pediatric allergy prediction platform named KARATE, accessible via mobile devices. In the future, the team will continue to optimize the AI model by incorporating more diverse data and machine learning methods to build a comprehensive online prediction platform. This platform will further assist clinical doctors in accurate diagnosis, thereby enhancing the medical quality of pediatric allergic diseases. 

Resource (mandarin)