Russian Scientists Create Hybrid AI System for Ultra-Fast Blood Diagnostics
A new software platform developed in Russia can identify white blood cell types in less than a second, combining expert logic with deep learning to deliver accuracy and transparency.
Researchers at Belgorod State National Research University have developed a unique hybrid system for automatic recognition of white blood cell types. The software analyzes microscopic images in under a second with 93% accuracy—outperforming existing solutions in both speed and versatility.
The key innovation lies in merging two approaches. The first uses expert rules—an algorithmic decision-making tool that mimics the logic of a human specialist—analyzing the shape, color, and structure of the cell. The second applies a deep-learning neural network that independently classifies cells into one of five categories: neutrophils, lymphocytes, monocytes, eosinophils, or basophils.
When both methods produce the same result, it’s automatically recorded by the system. In cases of discrepancy, the image is sent to a doctor for final review. This hybrid setup not only ensures high diagnostic accuracy but also maintains transparency, giving medical professionals the ability to oversee and adjust the analysis.
The scientists confirmed the method’s effectiveness on real-world medical data and plan to further improve its adaptability and precision for widespread clinical adoption.
