ANS2JSON: uma API baseada em IA, para extração de Informações da avaliação neurológica de pacientes com hanseníase

Abstract

The Simplified Neurological Assessment (SNA) is a mandatory and essential tool for monitoring patients with leprosy in Brazil. However, its physical format and manual completion represent a significant obstacle, as they confine crucial clinical data to paper records and hinder longitudinal analyses, large-scale research, and evidence-based health policymaking. This study introduces ANS2JSON, an innovative application that overcomes this barrier through a hybrid Artificial Intelligence architecture for the complete and automatic digitization of SNA forms. The proposed solution addresses the challenge of interpreting a complex form that combines textual data with visual diagrams by employing a two-stage pipeline: (1) a computer vision model (YOLOv8 + Random Forest) to detect and anatomically map sensitivity assessment points, and (2) an Intelligent Document Processing model to extract textual fields. The approach demonstrated outstanding effectiveness, achieving 97.5% precision in visual data interpretation and over 95% accuracy in extracting key clinical indicators. ANS2JSON, publicly available via the web, provides a robust tool for converting a vast repository of unstructured data into ready-to-use digital information, unlocking the potential of years of clinical records to advance both leprosy treatment and research. The application is accessible at: https://ans2json.dotlabbrazil.com.br .