An accurate deep learning model for wheezing in children using real world data.

Area of Science:

• Pulmonology
• Artificial Intelligence
• Medical Diagnostics

Background:

• Auscultation is a subjective method for diagnosing lung diseases, requiring significant expertise.
• Existing artificial intelligence (AI) models for respiratory sound analysis need performance improvements and better clinical relevance.

Purpose of the Study:

• To develop an improved deep learning model for detecting wheezing in pediatric patients using real-world clinical data.
• To enhance the accuracy and clinical applicability of AI in diagnosing pediatric respiratory conditions.

Main Methods:

• A prospective study collected respiratory sounds and clinical data from 76 pediatric patients.
• A deep learning model was implemented using a 34-layer residual network with a convolutional block attention module for audio and multilayer perceptron for tabular data.

Main Results:

• The proposed deep learning model achieved high performance metrics: 91.2% accuracy, 89.1% AUC, 94.4% precision, 81% recall, and 87.2% F1-score.
• The model demonstrated significant capability in accurately detecting wheeze sounds.

Conclusions:

• The developed deep learning model offers a high-performance solution for wheeze detection in children.
• This AI tool has the potential to significantly improve the accuracy of diagnosing respiratory diseases in clinical practice.