Establishment and validation of intelligent review rules for blood cell analysis based on the gradient boosting...

Area of Science:

• Hematology
• Artificial Intelligence
• Medical Diagnostics

Background:

• Laboratory efficiency in hematology is crucial for timely patient diagnosis.
• Current review rules may lead to high false positive rates, increasing workload.
• Automated screening methods are needed to optimize laboratory workflows.

Purpose of the Study:

• To develop artificial intelligence (AI)-based intelligent review rules for screening blood samples.
• To reduce the review rate and improve the work efficiency of hematology laboratories.
• To ensure the quality of whole blood analysis by minimizing errors.

Main Methods:

• Utilized 10,212 venous blood samples from four hospitals for rule development and validation.
• Employed the gradient boosting decision tree (GBDT) algorithm to establish intelligent review rules.
• Compared AI-generated rules against 41 International Consensus Group for Haematology Review (ICGHR) rules, analyzing false positive and negative rates.

Main Results:

• Developed 26 intelligent review rules using the GBDT algorithm from 9000 samples.
• On the validation set, AI rules achieved a lower false positive rate (14.60%) and review rate (27.55%) compared to ICGHR rules.
• The false negative rate of AI rules (1.89%) was comparable to ICGHR rules (0.90%) and met the <5% requirement, with no critical cells missed.

Conclusions:

• AI-generated intelligent review rules effectively decrease false positive and review rates in hematology.
• These AI methods ensure diagnostic quality by preventing missed critical hematologic cells.
• The AI-driven approach enhances laboratory efficiency, making it suitable for medical institutions.