Enhanced Pneumonia Detection in Chest X-Rays Through a Hybrid DL-ML Ensemble Model

Pneumonia is a respiratory infection and has occurred in more than 450 million people every year, causing about 2.5 million deaths, especially among infants, the aged, and people with compromised immune systems. It inflames the lungs by filling air sacs with fluid or pus, causing coughing, fever, and labored breathing. It is very essential to diagnose at the right time because pneumonia may be confused with pulmonary tuberculosis and seasonal infections, thus delaying the treatment and raising the risk of death. Currently, diagnostic methods rely on conventional image analysis models such as InceptionResNetV2, Xception, VGG16, ResNet50, and DenseNet121 in addition to transfer learning and standalone algorithms. However, these approaches lack the accuracy required for the detection of pneumonia. This study proposes a hybrid model combining deep learning, especially CNNs, with machine learning classifiers. It enhances specificity and sensitivity in diagnosing pneumonia from CXR images for earlier and more accurate detection. It could lead to a major decrease in mortality associated with pneumonia based on prompt effective treatment. Feature extraction from the CNN is married with machine learning classification to maximize the diagnostic power. This approach may provide an excellent framework from which future augmentations can focus on severity stratification and also more general types of thoracic diseases.