COMPARATIVE ANALYSIS OF THE CAPABILITIES OF TWO ARTIFICIAL INTELLIGENCE MODELS FOR CHEST X-RAY IMAGE INTERPRETATION

Abstract

The automated interpretation of chest X-rays with the help of artificial intelligence (AI) is gaining increasing importance in medical diagnostics. Technologies based on machine learning and deep neural networks demonstrate the potential to assist radiologists by accelerating the diagnostic process and reducing subjective errors. However, the effectiveness of different AI models varies depending on the algorithms, training data used, and methods of analysis. This pilot study conducts a comparative analysis between two AI models – ChatGPT and X-Ray Insight, a GPT model specifically developed to assist in the interpretation of medical images and support doctors in diagnostics. Their ability to interpret chest X-rays is evaluated, along with an examination of their strengths and limitations in real clinical scenarios. A standardized database of radiographs is used, including normal findings, pneumonia, pulmonary infiltrates, and other pathological conditions. The evaluation of the models includes diagnostic accuracy, clarity and correctness of interpretations, the usefulness of the provided explanations, and recommendations for further examinations. The obtained results are compared with expert radiological assessments. A paired t-test revealed no significant differences in Findings Description and Differential Diagnosis, but X-Ray Insight performed significantly better in recommending the Next Steps, suggesting stronger recommendations for further management. The study highlights the challenges and potential of generative and image-based AI models in radiographic diagnostics. Further adaptation and training of the models are necessary to achieve clinical reliability and effectiveness.