Centrifugal microfluidic platform with digital image analysis for parallel red cell antigen typing.

This study presents a portable multichannel microfluidic device for parallel and digital analysis of red cell antigen typing. A zigzag-shaped precise metering channel was designed for the simultaneous aliquoting of samples, which is independent of the volume of the predeposited blood-typing reagents in the reaction chambers. The entire assay protocol can be conducted using a sequential-step spinning protocol, which resembles that of conventional tube tests for blood typing; however, the manual procedure is largely reduced compared to that of conventional systems. After loading the samples, the disc is centrifuged in a defined program with five sequential steps, each of which can be completed in a few seconds. Through step-wise centrifugation, predeposited antibodies react with red blood cells, enabling the parallel identification of multiple red blood cell antigens without cross-contamination in 1 min. This is combined with gentle mixing to rapidly concentrate the agglutinates, making both visual and digital determination of agglutination straightforward. A customized image analysis algorithm for automatically determining the agglutination state was developed to complement this microfluidic system. The acquired image is processed after the test. The blood type is determined using a machine learning algorithm based on a histogram of oriented gradients (HOG) and support vector machines (SVM). This allows digital analysis to mirror the classical laboratory procedure for blood-type determination more accurately. The system was trained using a validated dataset of 150 blood samples, presenting 750 different agglutination patterns. The combination of SVM and HOG achieved 94.10% in the micro-weighted performance evaluation. This integrated microfluidic chip-based platform provides a "sample-in and answer out" demonstration for red blood cell typing, ensuring fast and reliable results because minimum manual steps are involved.