Convolutional neural network-based automatic classification for incomplete antibody reaction intensity in solid phase...

Area of Science:

• Medical Diagnostics
• Artificial Intelligence in Healthcare
• Immunohaematology

Background:

• Accurate classification of incomplete antibody reaction intensity (IARI) is crucial for diagnosing haemolytic disease.
• Current manual classification methods for IARI in solid-phase Coombs tests are time-consuming and prone to variability.
• An automated, contactless approach is needed to improve IARI classification accuracy and efficiency.

Purpose of the Study:

• To develop and evaluate a deep ensemble learning model for the automatic classification of IARI.
• To compare the model's performance against manual classification by immunologists.
• To assess the impact of the model on classification accuracy and efficiency.

Main Methods:

• A deep ensemble learning model integrating five convolutional neural networks was developed.
• The model was trained and validated on a dataset of 1628 IARI images across five categories.
• Performance was evaluated based on accuracy, comparison with expert manual classification, and classification time.

Main Results:

• The deep ensemble model achieved high classification accuracies for all IARI categories (99.4%-100%).
• Model-assisted manual classification by immunologists showed a significant accuracy increase (average +6.1%).
• Automated classification time (0.094 s/image) was substantially faster than manual classification (5.528 s/image).

Conclusions:

• The proposed deep ensemble learning model offers a highly accurate and efficient solution for IARI classification.
• This technology can significantly improve the automation of haemolytic disease screening.
• The model assists human experts, enhancing diagnostic capabilities and potentially reducing errors.