Optimizing Pneumonia Detection Accuracy Through CNN-Based Analysis of Patients' Data
Khalid Aljuaid, Ziyad Albaqami, Hossam Meshref
Abstract
Pneumonia detection through chest X-ray imaging is a critical aspect of modern healthcare, requiring models that are not only accurate but also capable of delivering rapid and reliable results. Deep learning techniques, especially Convolutional Neural Networks (CNNs), have shown great potential in automating this diagnosis; however, there remain challenges in balancing inference speed, accuracy, and overall generalizability. Existing models such as ResNet50 and VGG16 have yielded mixed outcomes. In this study, we introduce an innovative hybrid approach that integrates structured patient records with the YOLO11n-cls deep learning model to enhance pneumonia detection. This method combines key patient features-including age, oxygen saturation levels, and lung capacity-with YOLO's advanced image processing capabilities. The YOLO11n-cls model, customized for classification tasks, was trained and evaluated on chest X-ray datasets. The proposed model delivered 92.3% accuracy, 99.82% prediction confidence, and a fast inference speed of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$9 \sim$</tex> 10 milliseconds per image. It also achieved 92.7% precision, <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 5. 1 \%}$</tex> recall, and an F1-score of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 3. 9 \%}$</tex>. Compared to established models such as ResNet50 and FA-Net, our approach maintains competitive diagnostic accuracy while significantly improving processing speed, making it particularly suitable for real-time clinical use.