- - -

- - -

- - -

- - -

-

Introduction:
Granulocyte immunology is a small but relevant field, especially for the detection of diseases like the (often under-diagnosed) transfusion-related acute lung injury (TRALI).
Guidelines for performing granulocyte immunology testing recommend screening using functional tests (GAT, granulocyte aggregation test), microscopy (GIFT, granulocyte immunofluorescence test) and flow cytometry (white blood cell/granulocyte immunofluorescence testing, Flow-WIFT/GIFT) as well as confirmatory testing using ELISA-based techniques.

Methods:
Mostly, granulocyte immunology testing is performed using separate tests for each assay, which is labour-intensive and time-consuming.
In addition, microscopical assays are always prone to subjectivities.
But with the recent advances in the field of image flow cytometry and computational image analysis using machine learning, multiple diagnostic steps could be joined together so that analysis steps could be automated.
In this project we present a method for automated, high-throughput image flow cytometry analysis for combining GIFT and Flow-WIFT/GIFT.
Comparisons were performed on specimen including allo- and auto-antibodies against HNA-1a, -1b, -1d, -2, -3a, -3b, -4a as well as HLA class I.

Results:
Outcomes between image flow cytometry and GIFT were greatly comparable (AUROC 0.984), although the resulting images had a lower quality when being compared to classical microscopy.
The use of machine learning tools also allowed to automate the generation of a proposed classification from the high-throughput image generation.
Explainable Artificial Intelligence (XAI) tools can also be used to give a human-understandable interpretation of the resulting images, which could later be evaluated by the technician (if needed).
In addition, the results from image flow cytometry were identical to the Flow-WIFT/GIFT assay (AUROC 1).
Furthermore, hands-on-time and turnaround-time could be greatly reduced.

Conclusion:
The use of image flow cytometry could be a major advancement in the field of granulocyte immunology.
In addition, the use of machine learning and XAI techniques allows not only to automate image classification tasks but also to propose human-understandable interpretations of the images.