Implicit dimension measurement for automated cross-sectional inspection of multi-lumen medical catheters

초록

Ensuring the cross-sectional shape integrity of medical catheters is a necessity for their safe and effective clinical functionality. Although visual inspection technologies have advanced rapidly, automated inspection of catheter tubes remains challenging due to the complex, deformable structures resulting from extrusion processes and the inherent properties of the materials. In this study, we introduce the Implicit Dimension Measurement (IDiM) framework, which combines rule-based expertise with a data-driven Endpoint Alignment Model (EAM). By parameterizing cross-sectional dimensions using carefully defined endpoints and reference points, IDiM robustly infers key geometric features—even under moderate deformations. We validate its measurement accuracy on multi-lumen catheters (2-, 3-, and 4-lumen) through a high-resolution imaging setup deployed on an actual production line. Experimental results demonstrate measurement precision within five pixels of inter-annotator deviation, comparable to that of human inspectors, along with reliable detection of severe deformation cases via an anomaly detection approach. These findings highlight the practical feasibility of IDiM for high-fidelity shape inspection in medical manufacturing and suggest its broader applicability to other industries requiring precise dimensional verification. © 2025 The Authors

키워드