Lung infections, including tuberculosis, pneumonia, and COVID-19, remain significant global health challenge. During pandemic the number cases registered with COVID-19, tuberculosis (TB), and pneumonia were is millions. Identifying the difference between the three infectious diseases using clinical images is challenging for radiologists. There is a need for accurate and efficient diagnostic approach and reduce the dependency on radiologist to identify the lung infectious disease for better treatment. This study focuses on analysis of clinical bio-medical images of lung infectious diseases using computer aided devices along with deep learning (DL) models for identifying the four categories of chest X-ray (CXR) images. The study proposed a novel convolutional neural network (CNN) inspired from an inception-v3 and ResNet pre-trained DL model to identify the four categories of lung infectious disease CXR images. The proposed CNN architecture consists of multi-scale convolutional filter and residual skip connections to identify the small and large spatial features and avoid vanishing gradient in classifying the lung infectious CXR images with better performance The proposed model achieved an accuracy of 97.5% in identifying the TB, pneumonia, COVID-19, and healthy CXR images.