White blood cell evaluation in haematological malignancies using a web-based digital microscopy platform

Area of Science:

• Hematology
• Digital Pathology
• Artificial Intelligence

Background:

• Digital microscopy offers advantages over traditional light microscopy for blood film analysis.
• Current digital systems face limitations in accessibility and manufacturer constraints.
• Web-based AI platforms present a potential solution for remote and flexible analysis.

Purpose of the Study:

• To evaluate the accuracy of a scanner-agnostic, web-based AI system for white blood cell (WBC) differentials.
• To assess the AI system's capability in identifying blast cells in hematological malignancies.
• To compare the AI system's performance against the gold standard of manual microscopy.

Main Methods:

• Digitized peripheral blood films (normal and abnormal) were analyzed using an online AI platform (Techcyte©).
• AI-performed WBC differentials were reviewed, and manual adjustments were made as needed.
• Results were correlated with manual microscopy, and sensitivity/specificity for blast identification were calculated.

Main Results:

• The AI system showed strong correlation (r > .8) with microscopy for normal cell types.
• AI performance on abnormal films (r = .50-.87) improved significantly with manual digital review (r > .95), except for immature granulocytes (r = .62).
• Blast identification sensitivity was 96% initially, improving to 99% after manual review; specificity increased from 25% to 84%.

Conclusions:

• The web-based platform facilitated remote analysis comparable to traditional microscopy.
• The AI software provided adequate WBC differentials for normal films.
• The AI system demonstrated high sensitivity for blast identification in malignant cases, further enhanced by manual review.