Advancing Orthopaedics Diagnostic Intelligence: Deep Learning and Generative AI models for Fracture Identification and Dialogue-driven documentation and Decision Making

Accurate radiographic assessment is essential in diagnosing pediatric fractures to prevent misdiagnosis. Pediatric skeleton anatomy exhibits both uniqueness and age-related variability, which enhance the challenges in correct radiograph analysis and diagnostic decisions. Therefore, we propose artificial Intelligence models for radiograph annotation and fracture identification using deep learning and natural language processing to analyze radiographs for fracture detection. Secondly, physicians may overlook crucial questions to pose during the doctor-patient dialogue, which may lead to misdiagnosis and incorrect decision-making. Therefore, we aim to equip the system with Generative Pre-trained Transformer (GPT) models to extract information from the dialogue. Our system extracts the important information, identifies missing information, and auto-documents the extracted information. We will develop a user-friendly software “Ortho-Assistant” with functionalities of automatic radiograph assessment, automatic clinical documentation, and assistance in diagnosis and treatment decision-making. In KT activities, we will publish the outcomes in reputable journals, present at conferences and workshops, and train the undergraduate students.