abstract={A child dies of pneumonia every 43 seconds. In 2017, Pneumonia killed more than 808 000 children under the age of 5, or around 2,000 every day, accounting for 15% of all deaths of children under 5 years. Pneumonia kills more children than any other infectious disease. People at-risk for pneumonia also include adults over the age of 65 and people with preexisting health problems. Globally, an estimated 450 million cases of pneumonia are recorded each year, resulting in nearly 4 million deaths. Almost all of these deaths are preventable. Chest X-rays are primarily used for the diagnosis of this disease. However, even for a trained radiologist, it is a challenging task to examine chest X-rays. There is a need to improve the diagnosis accuracy. This study proposes an efficient dual-path deep learning framework for automated pneumonia detection from chest X-ray images to support clinical decision-making. The model employs two parallel streams: one processing lung-segmented images and the other using full chest X-ray images. Each stream concatenates the results and goes to another pipeline consists of an ensemble of ResNet18, DenseNet121, and InceptionV3 architectures, with final predictions combined via ensemble averaging. Evaluated on the PneumoniaCXR dataset comprising 33,547 chest X-ray images across three classes (Normal, Pneumonia, and No Opacity/No Normal), the proposed dual-path model achieved a test accuracy of 96% on over 6,000 held-out images. These results demonstrate that the approach significantly outperforms individual baseline models and offers a reliable, interpretable tool to assist radiologists in improving diagnostic accuracy and speed, especially in high-burden and low-resource environments.},
0 commit comments