AI-Based Protocol Assistant in Pelvic Floor Ultrasonography: Development and Impact on Workflow Adherence and Efficiency

Abstract

Background: Pelvic floor ultrasonography (PFUS) is a highly operator-dependent imaging modality in which variability in image acquisition, procedural sequencing, and documentation may reduce diagnostic reliability and reproducibility. Although standardized protocols are available, consistent real-time implementation remains challenging. Artificial Intelligence (AI) has increasingly been explored as a workflow-support tool to improve standardization and reduce operator dependency in medical imaging. Objective: This study aimed to develop and evaluate an AI-based Protocol Assistant to support structured workflow execution in pelvic floor ultrasonography. Methods: This study used a Research and Development approach with a modified 4D model limited to the Define, Design, and Develop stages. The system was designed to translate standardized PFUS protocols into stepwise procedural guidance integrated with workflow control and structured documentation. Feasibility was assessed through expert validation using Aiken’s V coefficient. Initial functional evaluation employed a one-group pretest-posttest design involving 50 sonographers performing examinations without and with system assistance. Outcomes included protocol adherence and examination time. Data were analyzed descriptively and with the Wilcoxon Signed-Rank Test. Results: The system demonstrated very high content validity, with Aiken’s V values of at least 0.98 across all evaluated components. Protocol adherence improved significantly, with protocol loss decreasing from 16.7% without system assistance to 0.0% with the Protocol Assistant (p < 0.05). Examination time was also significantly reduced after implementation (p < 0.05). Conclusion: The AI-based Protocol Assistant showed strong preliminary feasibility as a workflow-support system for PFUS, improving protocol adherence and examination efficiency. Further studies are needed to assess diagnostic accuracy, reproducibility, and generalizability across clinical settings.

Ähnliche Arbeiten