Presentation Information
[C-10_C-11-02]Multi-level classification of drainage images based on CIE L*a*b* color space
◎△Shoto kanatsu1, Iori Kojima, Shiori Kojima, Yohichiro Kojima (1. Hokkaido University of Science)
Keywords:
Image,Classification
Continuous monitoring of postoperative drainage fluid is essential for the early detection of complications such as internal bleeding and infection. However, drainage fluid color is currently assessed by visual inspection, resulting in subjective evaluations with limited reproducibility. Accurate color measurement is further hindered by specular reflections from semi-transparent drainage tubes and the continuous transition of drainage fluid color from bloody to serous.
In this study, we investigated an objective image-based approach for quantitative drainage fluid color assessment. Smartphone images were analyzed in the CIE L*a*b* color space after suppressing high-intensity specular reflections using a spatial masking technique. The extracted color features enabled multi-stage classification of drainage fluid color. Notably, pale-colored samples with nearly identical lightness (L*) were successfully distinguished using the b* component and chroma (C*ab). These results indicate that the proposed method provides a more objective and discriminative assessment than conventional visual inspection and has potential to support postoperative monitoring and nursing practice. Future work will focus on temporal ΔE analysis and the development of an automatic multi-stage classification system.
In this study, we investigated an objective image-based approach for quantitative drainage fluid color assessment. Smartphone images were analyzed in the CIE L*a*b* color space after suppressing high-intensity specular reflections using a spatial masking technique. The extracted color features enabled multi-stage classification of drainage fluid color. Notably, pale-colored samples with nearly identical lightness (L*) were successfully distinguished using the b* component and chroma (C*ab). These results indicate that the proposed method provides a more objective and discriminative assessment than conventional visual inspection and has potential to support postoperative monitoring and nursing practice. Future work will focus on temporal ΔE analysis and the development of an automatic multi-stage classification system.