In recent years, DL-based image methods have become more popular in the fields of computer-based detection, diagnosis, and prognosis of respiratory diseases. One such use case is in Chest X-rays (CXR), which we use for the diagnosis and handling of pneumonia, a condition that can cause serious health problems if not detected early. However, understanding CXR images can be challenging due to hybrid and ambiguous signals. This is where an inventive image retrieval model comes into play. By providing not only similar images but also relevant clinical information, our approach can provide more meaningful insights compared to traditional diagnostic models. In this study, we introduce a new CXR image retrieval model specifically designed to detect pneumonia, utilising DML (Deep Metric Learning) methods. Unlike standard diagnostic methods, which typically map images directly to feature labels, the proposed model focuses on creating an optimized CXR image feature embedding space where images with similar content and labels are clustered together. Using inherent deep learning capabilities, particularly through the use of Siamese networks, our model detection accuracy has been improved compared to the traditional model having inferior approximation, ultimately asserting effective diagnoses in clinical environments.