The Application of Artificial Intelligence-Based Bone Marrow Cell Analysis System in Pediatric Hematological Diseases

Area of Science:

• Hematology
• Medical Diagnostics
• Artificial Intelligence

Background:

• Clinical diagnosis of hematological diseases relies on bone marrow (BM) nucleated cell differential counts.
• Automated systems offer potential for improving diagnostic efficiency and accuracy.

Purpose of the Study:

• To evaluate an AI-based system (Morphogo) for automatic classification of bone marrow nucleated cells in pediatric hematological disease samples.
• To assess the accuracy, sensitivity, and time efficiency of Morphogo compared to manual methods.

Main Methods:

• Bone marrow aspirate smears from 213 pediatric patients were collected.
• Digital images of smears were acquired using automated scanning (×40 objective).
• Morphogo captured and classified nucleated cells from selected areas (×100 objective).

Main Results:

• Morphogo achieved >87.8% overall accuracy in pre-classifying nucleated cells.
• High average sensitivity and accuracy were observed in cell classification.
• Morphogo demonstrated positive correlations with manual categorization for various hematological diseases and reduced classification time.

Conclusions:

• Morphogo shows significant clinical potential for early screening of pediatric hematological diseases.
• The AI system is a reliable automated tool for differential counting and analysis of bone marrow nucleated cells.