In the security and healthcare domains, using data from medical imaging for precise individual identification and verification has become increasingly vital. This necessity is particularly pronounced in scenarios where old and traditional identification techniqhes prove inadequate, such as during natural disasters or instances of physical impairment. Chest X-ray radiographs offer a promising avenue for reliable identification owing to their ability to capture distinctive anatomical features of the rib cage, lungs, and heart. Our research introduces a new approach to person identification utilizing chest X-ray radiographs, employing ResNet50 with spatial attention within a Siamese and Triplet network architecture. In contrast to conventional convolutional neural networks (CNNs), the inclusion of ResNet50 with spatial attention enhances the model’s capability to discern global dependencies within images, potentially facilitating the extraction of discriminative characteristics for identification purposes. This research contributes to the existing by presenting a tailored methodology explicitly crafted for processing chest X-ray images, thereby presenting a viable solution for reliable identification in security and medical contexts, particularly in emergency scenarios. To validate the efficacy of the proposed method and explore its practical applications, rigorous validation and testing procedures are imperative.