Next Generation Digital Morphology: Blast Preclassification in Bone Marrow Aspirates.

Area of Science:

• Hematology
• Medical Diagnostics
• Artificial Intelligence

Background:

• Morphologic evaluation of peripheral blood (PB) smears and bone marrow aspirates (BMA) is crucial for acute leukemia diagnosis, but it is time-consuming and prone to variability.
• Artificial intelligence (AI), especially deep learning, offers potential for automated leukocyte classification and blast detection, addressing limitations of manual methods.

Purpose of the Study:

• To conduct a narrative review of AI-based approaches for morphologic evaluation in hematology, specifically for blast recognition and leukemia screening.
• To analyze classical machine learning and deep learning methodologies applied to PB smears and BMA samples.
• To review commercial digital morphology platforms for their blast detection performance.

Main Methods:

• A narrative literature review was performed.
• Analysis included classical machine learning and deep learning techniques applied to peripheral blood and bone marrow samples.
• Commercial digital morphology platforms were assessed for blast detection capabilities.

Main Results:

• Deep learning models, particularly convolutional neural networks, demonstrate near-human accuracy (>90% sensitivity/specificity) for blast detection in PB smears.
• Classical machine learning shows moderate performance in blast recognition, limited by manual feature selection.
• AI performance in BMA analysis is more variable due to sample complexity; commercial platforms excel in mature cell classification but show modest accuracy for immature/neoplastic cells.

Conclusions:

• AI-based digital morphology systems are promising for hematology labs, enhancing efficiency and standardization.
• Current limitations in blast detection accuracy and generalizability prevent standalone diagnostic use; further validation and interpretability are needed.
• Widespread adoption requires rigorous validation, regulatory approval, and quality assurance, with caution advised for AI-driven BMA systems per ICSH guidelines.