Optimizing Pneumonia Identification in Chest X-rays: ResNet-18 and the Role of Diverse Activation Functions

Identifying pneumonia from chest X-ray images is essential for clinical diagnosis and the development of treatment plans. Deep learning methods have shown potential in automating this procedure, but obstacles remain to be overcome to improve precision and resilience. This paper presents a unique methodology that combines the Rectified Tangent Activation (RTA) Activation function utilizing the ResNet-18 architecture for pneumonia detection in chest X-ray images. The RTA activation function is a novel approach that combines the relu and Hyperbolic Tangent (Tanh) to overcome the constraints of traditional activation functions. This approach introduces non-linearity and tangent aspects to the function. We enhance the ResNet-18 architecture by substituting the relu with the RTA activation function. Subsequently, we perform experiments on a dataset comprising pneumonia chest X-ray pictures. The findings of our study indicate that the ResNet-18 model, when equipped with the RTA activation function, outperforms traditional methods in terms of precision, robustness, and generalizability. Including RTA improves the model's capacity to comprehend intricate patterns and reduce interference, resulting in more dependable pneumonia identification. This study contributes to medical image analysis by providing novel AFs and utilizing cutting-edge deep-learning architectures to enhance pneumonia diagnosis. The suggested methodology shows potential for improving clinical decision-making and patient care in identifying and treating pneumonia.