Evaluation of artificial intelligence-assisted morphological analysis for platelet count estimation.

Area of Science:

• Hematology
• Medical technology
• Artificial intelligence in diagnostics

Background:

• Platelet count estimation is crucial for diagnosing and managing various medical conditions.
• Current methods may have limitations in accuracy, particularly with abnormal samples.

Purpose of the Study:

• To evaluate the performance of platelet count estimation using artificial intelligence (AI) on the MC-80 digital morphology analyzer.
• To compare AI-based estimations with traditional methods and flow cytometry.

Main Methods:

• The MC-80 analyzer utilizes two AI principles for platelet count estimation: PLT/RBC ratio (PLT-M1) and estimate factor (PLT-M2).
• 977 samples with varying platelet counts were analyzed, with 271 subjected to immunoassay.
• Platelet counts from the MC-80 analyzer were compared against a hematology analyzer (PLT-I, PLT-O) and flow cytometry (PLT-IRM).

Main Results:

• Both PLT-M1 and PLT-M2 showed minimal deviation and strong correlation with flow cytometry (PLT-IRM), outperforming standard hematology counts (PLT-I).
• AI estimations demonstrated robust correlation even with interfering factors like large platelets or RBC fragments.
• The AI methods provided higher accuracy for platelet transfusion decisions, particularly for samples with low platelet values.

Conclusions:

• The MC-80 digital morphology analyzer's AI-driven platelet count estimation offers high accuracy.
• The reviewed AI results effectively confirm suspicious platelet counts, enhancing diagnostic confidence.