DEVELOPMENT OF AN AUTOMATED INTERPRETATION SYSTEM FOR SEROLOGIC HLA TYPING USING THE CDC METHOD

Abstract

Complement-dependent cytotoxicity (CDC) testing is a widely used method for the classification and matching of human leucocyte antigens (HLA) in organ transplantation. However, manual interpretation of microplate images is time-consuming and error-prone, and an automated method is needed. This study focuses on pre-processing of CDC-scale microtiter plate images to facilitate machine learning-based cytotoxicity assessment and HLA antigen recognition. Pre-processing includes image normalization (scaling, brightness), well segmentation based on circular pattern recognition with OpenCV, greyscale adjustment to improve contrast and visibility of absorbing colors, and removal of noise and thresholds. The developed pipeline provides consistent image quality and achieves 75% accuracy for well segmentation throughout the dataset. These pre-processing steps form the basis of an automated evaluation system that minimizes human error sources and improves diagnostic efficiency. Further work will include feature extraction and classification modelling for cytotoxicity assessment.