Pilot Study on the Performance of a New System for Image Based Analysis of Peripheral Blood Smears on Normal Samples

Area of Science:

• Hematology
• Medical Image Analysis
• Artificial Intelligence in Medicine

Background:

• Manual review of peripheral blood smears is time-consuming.
• Automated systems for blood smear analysis are gaining traction.
• Accurate differential white blood cell counts are crucial for diagnosis.

Purpose of the Study:

• To evaluate the performance of the Shonit™ AI-based blood smear analysis system.
• To assess the accuracy of white blood cell classification and differential counting.
• To explore the potential for estimating red blood cell and platelet indices from images.

Main Methods:

• Analysis of 100 normal peripheral blood smear samples using the Shonit™ system.
• Comparison of Shonit™ results with an automated hematology analyzer.
• Evaluation metrics included classification accuracy and differential count correlation.

Main Results:

• White blood cell classification specificity exceeded 97.90% and precision exceeded 93.90% for all five normal types.
• White blood cell differential counts showed correlation within known inter-analyzer variability.
• The system demonstrated potential for deriving quantifiable metrics from blood smear images.

Conclusions:

• The Shonit™ AI system shows significant promise for automated peripheral blood smear analysis.
• The system achieves high accuracy in white blood cell classification and differential counting.
• Future enhancements should focus on analyzing abnormal blood samples.