DESIGN AND EVALUATION OF AN AI-ASSISTED MEDICAL IMAGING SYSTEM FOR PULMONARY NODULE DIAGNOSIS

Abstract

The advancement of artificial intelligence (AI) technology has not only transformed social and industrial production but has also accelerated progress in the medical domain. Currently, the use of AI in healthcare is rapidly expanding, with medical imaging being a key area of application. This study focuses on integrating AI with a medical imaging–assisted diagnosis system to address the limitations and inaccuracies of conventional manual diagnosis in the detection of pulmonary nodules. By applying established principles of image segmentation, a comprehensive design and optimisation of an AI-driven diagnostic framework was developed to enhance precision in pulmonary nodule assessment. To evaluate its performance, 200 patient cases comprising 231 nodules, confirmed by pathology or with stability over more than two years of follow-up, were analysed. The AI system detected 881 true nodules with a sensitivity of 99.10% (881/889), whereas radiologists identified 385 true nodules with a sensitivity of 43.31% (385/889). Moreover, the AI system demonstrated markedly higher sensitivity in detecting non-calcified nodules compared to radiologists (99.01% vs. 43.30%, P < 0.001), with statistically significant differences. These findings highlight the potential of AI-based medical imaging systems to substantially improve diagnostic accuracy, reduce missed detections, and support radiologists in clinical decision-making.