Optimal Pneumonia Detection Using X-ray Images and Tabular Patient Data

Pneumonia is the most common lung infection and a major cause of child mortality. Detecting the disease early can lead to better diagnoses and more effective treatment. While chest X-rays are essential for diagnosing, interpreting the results became challenging and inconsistent. Here, we created a machine learning model that analyzes the chest X-ray pictures by combining convolutional neural networks with a decision tree algorithm applied to tabular patient data, containing age, gender, diabetes status, heart rate, respiratory rate, and nutrition levels. It categorizes the chest X-ray images into two main outcomes either the X-ray appears normal, or it shows signs of pneumonia, while the tabular patient data will enhance the predictions for efficiently diagnosing pneumonia. The CNN component will identify the radiological features that are indicators of pneumonia. The model was trained on a large dataset that includes $\mathbf{5, 2 1 6}$ X-ray images. The improvement we made is to create a multi-model framework that analyzes the X-ray images with the integration of tabular patient data, improving accuracy by integrating both patient data and imaging compared to models that use only imaging data.