For over a century, diagnostic experts have used manual examination of peripheral blood smears as the gold standard for diagnosing hematological disorders. Morphological evaluation provides key insights into conditions ranging from malignancies to infections. Although laboratories widely use automated analyzers, experts still rely on manual differential counts [1] for abnormal or flagged samples. This is recommended by international guidelines.
However, the landscape is shifting. Clinical laboratories now face higher test volumes, rising costs, and fewer expert morphologists. Annual microscopy is increasingly labor-intensive. It requires highly trained staff and is subject to significant observer variability. Furthermore, limited expert review outside standard hours can delay decisions. These challenges drive the adoption of AI and digital imaging, transforming blood analysis.
From Glass Slides to Full-Field Digital Imaging
Early digital morphology systems, such as the CellaVision [2], introduced automated image capture and pre-classification of blood cells. Studies have shown that these systems achieve accuracy similar to manual microscopy in routine diagnostic settings. They also improve standardization and efficiency [3,4]. The International Council for Standardization in Hematology (ICSH) recognized its growing importance. The council recommended integrating it into laboratory workflows to enhance reproducibility and quality assurance.
Next-generation platforms like Scopio Labs now offer full-field imaging. This enables visualization of the entire blood smear at high resolution [5]. In contrast, traditional analyzers review about 200 white blood cells per sample. Full-field systems analyze up to 1,000 [5,6], covering the complete smear. This advancement boosts diagnostic sensitivity and enables confident cell classification.
These AI systems use deep learning, trained on large, annotated datasets of blood cell images [7]. They automate image acquisition, digitization, cell identification, and data transfer to lab systems no human input needed for normal samples. Unlike conventional systems, full-field imaging digitizes thousands of fields per slide, reducing sampling bias and enabling high-throughput, reproducible morphology.
Diagnostic Performance and Clinical Impact
AI-assisted systems consistently demonstrate solid agreement with manual differential counts across all major leukocyte populations. When a normal blood smear with 200 manually counted cells is compared with a digital count of over 1,000 cells [5], concordance remains high. Statistical stability also improves. The larger sample size improves precision and eliminates the variability often encountered in serial manual differentials.
AI-based morphology excels in detecting rare events. Manual evaluation of 200 cells yields a sensitivity of about 0.5%. This can miss low-burden diseases such as minimal residual leukemia or monoclonal B-cell lymphocytosis [8]. Automated systems review thousands of cells, enabling rare findings to be reproduced. This prompts earlier diagnosis and targeted confirmatory tests, such as flow cytometry. Improved detection leads to better patient management and reduced costs.
Discrimination between reactive and neoplastic lymphocytes has long challenged morphologists. Even experienced experts often disagree when they examine the same blood smear. Their disagreements reflect the inherent subjectivity of visual assessment. As a solution, AI models trained on diverse reference material double-check agreement levels and maintain consistency between samples and institutions. Similarly, blast quantification is critical for the diagnosis and monitoring of acute leukemia. It can achieve accuracy equal to or better than human inter-observer agreement. This strengthens the role of AI morphology in clinical decision-making.
Workflow improvement is a key operational benefit. Modern systems reduce blood smear review time by up to 60%, allowing skilled morphologists to focus on verification and interpretation [10]. Automated pre-classification speeds case throughput and reduces fatigue errors. Most slides are processed automatically, with experts reviewing only flagged or complex cases, while still benefiting from digital access. Compared to traditional workflows, where manual review rates are 9-15% [8,9], this yields much greater productivity and consistency.
Digital morphology enables real-time remote access to high-resolution images. This assists during off-hours or at smaller institutions lacking in-house expertise, and supports collaborative diagnostics and training across laboratory networks.
Limitations and the Road Ahead
Despite progress, limitations persist. Poorly prepared slides reduce AI accuracy, and current algorithms lack training on rare or novel cases. Continuous model validation is needed as datasets grow, and implementation challenges, such as data integration, compliance, and cost, need to be addressed. AI systems best serve as decision-support to augment, not replace, human expertise [11].
Researchers develop deep learning models with broader generalizability, improve recognition of rare or mixed-cell populations, and integrate these models with multi-omics and patient data. Together, these innovations deliver real-time diagnostics within a fully digital workflow [12].
The move from glass slides to AI-enabled full-field digital microscopy is the field’s most significant change since the advent of flow cytometry. This shift brings automation, analytical rigor, and expert oversight, yielding faster, more consistent, and more sensitive diagnostics. As clinical expertise combines with AI, the result is faster, more accurate, and more accessible diagnostics, transforming patient care and laboratory medicine.
References:
(1) Barnes PW, McFadden SL, Machin SJ, et al. International Consensus Group for Hematology Review. Suggested criteria for action following automated CBC and WBC differential analysis. Lab Hematol. 2005;11(2):83–90.
(2) Cornet E, Perol JP, Troussard X. Performance evaluation and relevance of the CellaVision DM96 system in routine analysis and in patients with malignant hematological diseases. Int J Lab Hematol. 2008;30(6):536–542.
(3) Billard M, Lainey E, Armoogum P, et al. Evaluation of the CellaVision DM automated microscope in pediatrics. Int J Lab Hematol. 2010;32(5):530–538.
(4) Briggs C, Longair I, Slavik M, et al. Can automated blood film analysis replace the manual differential? An evaluation of the CellaVision DM96 automated image analysis system. Int J Lab Hematol. 2009;31:48–60.
(5) Katz BZ, Feldman MD, Tessema M, et al. Evaluation of Scopio Labs X100 full-field PBS: the first high-resolution full-field viewing of peripheral blood specimens combined with artificial intelligence-based morphological analysis. Int J Lab Hematol. 2021;43:1–9.
(6) Merino A, Puigví L, Boldú L, et al. Optimizing morphology through blood cell image analysis. Int J Lab Hematol. 2018;40(Suppl 1):54–61.
(7) Lee, M. K., & Kim, H. (2022). Red and white blood cell morphology characterization and hands-on time analysis by the digital cell imaging analyzer DI-60. PLoS One, 17(4), e0267638.
(8) Kratz A, Lee SH, Zini G, et al. International Council for Standardization in Hematology. Digital morphology analyzers in hematology: ICSH review and recommendations. Int J Lab Hematol. 2019;41:437–447
(9) De Iuliis V, Chiatamone Ranieri S, et al. Performance evaluation of the Scopio Labs X100HT digital morphology analyzer and abnormal cell detection in peripheral blood smears. Int J Lab Hematol. 2026;48(1):54–62.
(10)Lee, M. K., & Kim, H. (2022). Red and white blood cell morphology characterization and hands-on time analysis by the digital cell imaging analyzer DI-60. PLoS One, 17(4), e0267638.
(11) Jo, Y., Kim, S. H., Koh, K., Park, J., Shim, Y. B., Lim, J., Kim, Y., Park, Y. & Han, K. (2023). Artificial intelligence of digital morphology analyzers improves the efficiency of manual leukocyte differentiation of peripheral blood. BMC Medical Informatics and Decision Making. https://doi.org/10.1186/s12911-023-02153-z
(12) Koch, V., Wagner, S. J., Kazeminia, S., Sancar, E., Hehr, M., Schnabel, J., Peng, T. & Marr, C. (2024). DinoBloom: A Foundation Model for Generalizable Cell Embeddings in Hematology. arXiv preprint. https://doi.org/10.48550/arXiv.2404.05022
