Background and Aims

Detection of antinuclear antibodies (ANA) on HEp-2 cells by automated immunofluorescence assay (IFA) interpretation systems is the tool of choice for IFA automation. Machine-learning algorithms, a branch of artificial intelligence (AI), are considered the state-of-the-art in image classification. Especially deep learning algorithms based on convolutional neural networks (CNN) are regarded particularly powerful. The aim of the study was to create a complete automated AI-supported classification of ANA patterns according to revised ICAP nomenclature tree.

Methods

CNNs were established and trained for major ANA sub-classification tasks: a) metaphase recognition in DAPI, b) metaphase pattern, c) interphase pattern, d) pleomorphic pattern and e) cytoplasmic pattern. Images were obtained by an automated IFA interpretation system (akironNEO, Medipan, Germany).

Sets of pre-classified images were manually created for each task: a) training set for supervised learning b) validation set and c) test set for evaluation of prediction accuracy. Followed by merging rules of single model results to achieve final ICAP conform pattern result.

Results

Trained CNNs showed true classification rates above 95% for all models. The CNN were able to learn important features for major ANA sub-classification tasks. Merged results provided a classification model for revised ICAP nomenclature tree at competent-level that achieved high accuracy in consecutive routine samples.

Conclusions

Machine-learning algorithms add value to enhance the accuracy of image classification in automated IFA interpretation systems. Our approach showed efficient learning of competent-level pattern allowing full automation of ICAP nomenclature conform ANA classification.