알러지 스트립 자동 분석을 위한 영상처리 기술 연구

Abstract

This study presents an advanced image processing algorithm for analyzing Multiple Allergen Simultaneous Test (MAST) strips, addressing the limitations of manual interpretation methods. The algorithm is designed to ana- lyze LG Chem's 108-allergen MAST strips, capable of simultaneously processing 15 strips in a single image under varying lighting conditions. Key to the algorithm's robustness is a novel statistical analysis method for grade classification that effectively minimizes light interference. The method utilizes a calibration strip to estab- lish baseline measurements and employs pixel intensity comparisons between test and background areas. This approach allows for accurate result recognition across five levels: negative, weak positive, moderate positive, strong positive, and very strong positive. The algorithm incorporates adaptive thresholding and morphological operations for precise strip detection and segmentation. Crucially, it performs individual analyses on each detection area, calculating pixel intensity means (PIM) for both test and background regions. By comparing the PIM differences against calibrated thresholds, the system can accurately clas- sify results even under non-uniform lighting conditions. Experimental results demonstrated 100% accu- racy across various strip positions and lighting scenarios, showcasing the algorithm's resilience against environmental factors such as ambient light variations and structural obstructions. This high accuracy was maintained even when strips were placed in areas with significantly different light intensities, prov- ing the effectiveness of the statistical approach in mitigating light interference. While the study was lim- ited to laboratory conditions, the algorithm's ability to adapt to varying light conditions suggests its potential for improving MAST strip analysis in clinical settings. This automated approach significantly enhances the efficiency and consistency of allergy diagnostics, offering a solution that balances accuracy with practicality in real-world clinical applications.