Innovations in artificial intelligence to minimize diagnostic error - a comparison with human interpretation of chest radiographs in the clinical context: a scoping review

Introduction: Chest radiography remains the primary imaging modality for the evaluation of suspected pulmonary and mediastinal diseases, enabling the identification of conditions such as pulmonary nodules, pneumonia foci, fractures, and atelectasis, among others. In parallel, artificial intelligence (AI) has facilitated the development of advanced systems capable of highlighting suspicious areas on imaging studies and suggesting differential diagnoses. Objective: To evaluate the impact of implementing artificial intelligence (AI) in the interpretation of chest radiographs on minimizing diagnostic errors, compared to interpretations made solely by medical personnel. Methods: An exploratory systematic review was conducted using four databases: PubMed, LILACS, Clinical Key, and ScienceDirect. Twenty-three documents in English and Spanish, published between 2019 and 2024, were selected using the Rayyan reference management tool. Specific inclusion and exclusion criteria were applied to ensure relevance and quality. Results: Three thematic categories were described: “Pulmonary Nodules”, “Respiratory Pathologies”, and “Other Pathologies.” It was determined that the combined use of artificial intelligence and human interpretation increased sensitivity by 2–9% in diagnosing pathologies such as pulmonary nodules. Sensitivity for lung cancer, COVID-19, and pneumonia was reported at 91%, 75%, and 95%, respectively. Conclusions: The integration of artificial intelligence in the interpretation of chest radiographers enhances both sensitivity and specificity. Its value lies in complementing human expertise, thereby reducing diagnostic errors. AI also holds significant promise in clinical settings where access to specialized personnel is limited.