Enhancing Spirometry Reference Values Prediction: Exploring Neural Networks as Alternatives to Established Statistical Models

Spirometry is a widely used pulmonary function test that measures the volume and flow of air entering and leaving the lungs. It plays a crucial role in the diagnosis and monitoring of respiratory conditions such as asthma, chronic obstructive pulmonary disease (COPD), and cystic fibrosis. The interpretation of spirometry results often involves comparing the measured values to established reference values, which are computed based on equations involving demographic factors such as age, height, ethnicity, and gender. These equations are however developed for specific groups, leading to the need for separate equations for each ethnic group, gender, and age category as well as specific populations, leaving individuals from under-represented or unstudied groups without a representative reference point. The lack of universally applicable reference equations hinders both accurate diagnoses and comprehensive research on lung function. This paper focuses on the prediction of reference values for key spirometry parameters, including forced vital capacity (FVC), forced expiratory volume in one second (FEV1), and peak expiratory flow (PEF), using neural networks. Compared to traditional equations, neural networks offer a more inclusive and adaptable approach. By eliminating the need for separate equations for different groups, this method holds promise for improving the accuracy and consistency of spirometry reference values across diverse populations. Furthermore, the ability to continuously learn from new data allows the model to stay current with evolving trends. This study describes the deployment and evaluation of a neural network for spirometry reference value prediction, paving the way for further research on its potential impact in clinical settings.