Bone marrow: A Dynamic & Diagnostic Tissue  

Bone marrow, one of the body’s most dynamic and complex large tissues, serves as the primary site of hematopoiesis – the continuous production of blood cells and platelets. Within this microenvironment, a rich population of pluripotent stem cells differentiate into distinct myeloid and lymphoid lineages. ¹   The constant turnover and maturation of these diverse cell types are what make the bone marrow an extremely dynamic tissue both from a biological and a diagnostic standpoint. 

Bone marrow examination (BME) remains an essential diagnostic tool for a wide range of hematologic and systemic diseases, including leukemia, myelodysplastic syndromes, and aplastic anemia. ^{2 , 3 , 4}   Despite being introduced in the 19th century, BME has proven challenging to automate and has largely remained a manual process, reliant on the expertise of the hematopathologist.  ^{2 .}

However, as digital pathology and artificial intelligence (AI) technologies advance and are commercialized, the bone marrow examination field is now on the cusp of a major transformation.

The Challenges of Manual Microscopic Examination

Traditional microscopic examination of bone marrow aspirate is a time-intensive process requiring skilled expertise. Although the technique has proven indispensable for diagnosis, it poses several well-recognized drawbacks

• Restricted visualization: Technology limitations allow for review of only cell snap-shots and not an entire slide, potentially missing diagnostically relevant areas ⁵ 
• Labor-intensive and time-consuming – Manual evaluation requires a sample size of 500 cells/differential for BMA that is challenging to meet within the busy laboratory workflow. ^{2 , 4} 
• High operator dependency – The quality and accuracy of results depend heavily on the experience and skill of the examiner, particularly in diagnostically complex cases^{. 2}
• Inconsistency and subjectivity: Diagnostic/classification inconsistencies that lack standardization resulting from specimens that can vary considerably 4 Bagg  
• Substantial inter and intra-rater variability – Even highly trained experts may             yield inconsistent results due to the subjective and qualitative nature of manual morphological assessment. ^{2 , 3 , 4}   

While these challenges are not unique to BMA smears-they also affect peripheral blood smear (PBS) analysis-but the inherent heterogeneity of bone marrow samples adds further complexity.

The Added Complexity of Bone Marrow Aspirate Evaluation

BMA smears are heterogeneous by nature, varying significantly in cellularity, distribution of bone marrow particles or the lack thereof, and smear preparation technique (e.g., wedge vs. squash smears). ³ Lewis

This heterogeneity complicates automated image analysis because:

• Multiple regions of interest (ROIs) must be evaluated and analyzed requiring advanced AI neural network models to identify such regions. ³ 
• BM contains important and unique cell types rarely seen in PBS including early erythroid precursors, megakaryocytes, and plasma cells. Any automated system must therefore be trained to identify and accurately classify these cell types. ³ 

These factors have historically slowed the transition from manual to digital hematopathology in BMA.

From Microscopy to Automation: Enter AI for Bone Marrow Examination

Recent years have seen a surge of research exploring the potential of AI and machine learning to replicate-and even enhance-traditional bone marrow cytomorphologic assessment 

Early efforts were focused on extracting hand-crafted single-cell features from digitized images and using them to classify the cell in question. However most previous studies of automated cytomorphologic classification have focused on physiological or PBS cell classification, leaving gaps in models able to identify the full spectrum of BM cytomorphology needed for the diagnosis of hematological malignancies. One particular obstacle has been the lack of large, high-quality, annotated datasets. ² 

 “This is particularly true in situations like the cytomorphologic examination of BM, where there is no underlying technical gold standard, and human examiners are needed to provide the ground truth labels for network training and evaluation.”  Matek C, et al

Studies have shown that convolutional neural networks (CNNs) were able to identify cellular and spicular areas on bone marrow aspirate smears, and were coupled with cell detection and classification models, enhancing digital BMA assessment 5 Pozdnyakova

Automated pipelines have also demonstrated the ability to analyze far more cells than a human examiner could feasibly review, thereby greatly decreasing the associated variance in differential cell counts. ³

These AI-based approaches have provided evidence that they can not only replicate the BMA manual workflow but also enhance it, creating the impetus for commercial AI-based platforms designed for automated BMA smear analysis in clinical use ⁵ 

Benefits of Digital Imaging for Bone Marrow Interpretation

Manual BMA smear review limits the number of cells and regions that can be examined within a reasonable timeframe. According to the International Council for Standardization in Hematology (ICSH) guidelines, a minimum of 500 nucleated cells must be counted for an adequate differential-representing only a minute fraction of the total cells available on a slide^{.3  4  6} 

AI-driven digital imaging offers many transformative benefits:

Comprehensive analysis – Automated systems can analyze thousands-or even hundreds of thousands-of cells across an entire slide, capturing a more representative cellular landscape. ^{3  4} 

Standardization and reproducibility – Algorithms apply consistent criteria to every case, reducing observer bias. ^{3  4} 

Improved turnaround times – The ICSH recommends reporting within 3–24 hours for urgent cases, yet not all laboratories can consistently achieve these targets. ⁶ LDigital workflows reduce reliance on manual procedures, lowering operational costs while decreasing turn-around time(TAT) and minimizing laboratory error, enhancing both efficiency and safety. ⁴    

Remote consultation and collaboration Digitized BMA slides enable secure remote hematopathologist evaluation, and real-time consultation with experts and colleagues, particularly on difficult cases, improving the efficiency and speed of diagnosis and overall quality of patient care. ^{3 , 4, 5}    

End to End Pipelines for BMA Analysis

Several research groups have investigated the feasibility of end-to-end BMA smear analysis from identifying regions of interest (ROIs) to generating comprehensive differential counts.

• Wang et al, trained ML models for detection of bone marrow particles and cellular elements ROI suitable for further analysis. ⁷ 
• Tayebi et al. implemented a “You Only Learn Once” (YOLO)-based model capable of detecting and assigning class probabilities to all cellular and non-cellular objects in BMA digital whole slide image (WSI). The collective cytological information was then summarized as a Histogram of Cell Types (HCT), which is a novel information summary of cell-type distributions. ⁸ 
• Lewis et al. created an automated pipeline capable of generating 11-component differential cell counts (DCCs) from whole-slide BMA. Their sequential process-identifying optimal regions, detecting individual cells, and classifying them into key hematopoietic categories-demonstrated a high statistical and diagnostic concordance with manual DCCs across a diverse set of BMA slides with varying pathologies and cellularity. ⁹ 

These innovations address the long-standing challenges of manual assessment-improving precision, reducing observer variability, to accommodate modern clinical workloads.

Commercial Advances: Scopio Labs and the First Clinical-Grade AI Platform

Building on this body of research, commercial platforms have now entered clinical practice. Scopio Labs,X100 and X100HT systems represent the first digital full-field imaging platforms for BMA. ⁴ 

Scopio’s Full-Field Bone Marrow Aspirate™ Application integrates Full Field imaging with advanced AI decision support and remote collaboration capabilities. The platform’s core components include: ⁴ 

• Full-Field digital review at 100x magnification, allowing high-resolution visualization of specimen morphology across the entire slide. Whole slide imaging generates a high resolution whole slide image suitable for initial general impression, without decision support.
• AI-powered Decision Support System (DSS) that detects and pre-classifies thousands of marrow cells, presenting a representative subset (typically 500 cells) for expert validation and automatically generating an ICSH standardized draft report for review.
• Secure remote access, enabling hematopathologists to review cases from anywhere over the hospital’s secure network with a standard computer and sign off cases remotely thus completing the full digital workflow. In addition, there is the added benefit of being able to annotate cells for discussion within the context of the slide and fully digitally and remotely  consult with morphology experts and colleagues.

In a multicenter clinical trial published in 2024, the Scopio platform demonstrated high concordance with traditional manual microscopy while achieving greater standardization, sensitivity, and workflow efficiency. 5 Pozdnyakova Importantly, the digital images provided sufficient resolution to reveal fine cytoplasmic and nuclear details-such as Auer rods, cup-like nuclear invaginations, and hypolobated megakaryocytes-all critical features in diagnosing hematologic malignancies. ^{4  5} 

Moreover, the platform enhanced the assessment of hematopoietic maturation patterns, rendering this traditionally subjective process more objective and reproducible . ⁵ 

The Future of AI-Driven Bone Marrow Evaluation

Current commercial and research efforts primarily focus on automating differential cell counts-a critical but foundational task. However, as technology matures, the next generation of systems will likely incorporate far more sophisticated analytical capabilities, including: ⁵ 

• Detection and quantification of rare cell types and dysplastic cell populations
• Automated identification of blast cells and leukemic subtypes
• Integration of BMA findings with trephine biopsy and flow cytometry data for multimodal diagnostics
• Predictive modeling to support treatment response monitoring and disease progression tracking

As datasets grow and annotation tools improve, AI models will increasingly learn from real-world clinical experience, leading to continuously improving accuracy and reliability.

Ultimately, digital hematopathology promises to transform bone marrow examination from a laborious manual process into a streamlined, standardized, and data-rich workflow-empowering hematopathologists to focus their expertise where it matters most: in interpreting results, guiding therapy, and improving patient outcomes.

**Disclaimer: Scopio’s Full-Field Bone Marrow Aspirate Application is CE-marked cleared for sales in additional regions. Not commercially available in the US for in vitro diagnostic procedures.

References

1. Green A. Chapter 2: The Normal Bone Marrow.pp14-25. In:  van der Walt J, Orazi A, Arber DA. Diagnostic Bone Marrow Haematopathology. Cambridge University Press; 2021.Available at : https://www.cambridge.org/core/books/abs/diagnostic-bone-marrow-haematopathology/normal-bone-marrow/B55B0892482A31E3A34D6C4236EF9002 Accessed 29 October 2025.
2. Matek C, Krappe S, Münzenmayer C, Haferlach T, Marr C. Highly accurate differentiation of bone marrow cell morphologies using deep neural networks on a large image data set. Blood. 2021 Nov 18;138(20):1917-1927. doi: 10.1182/blood.2020010568. https://pubmed.ncbi.nlm.nih.gov/34792573/ 
3. Lewis JE, Pozdnyakova O. Digital assessment of peripheral blood and bone marrow aspirate smears. Int J Lab Hematol. 2023 Jun;45 Suppl 2:50-58. doi: 10.1111/ijlh.14082. https://pubmed.ncbi.nlm.nih.gov/37211430/ 
4. Bagg A, Raess PW, Rund D, Bhattacharyya S, Wiszniewska J, Horowitz A, Jengehino D, Fan G, Huynh M, Sanogo A, Avivi I, Katz BZ. Performance Evaluation of a Novel Artificial Intelligence-Assisted Digital Microscopy System for the Routine Analysis of Bone Marrow Aspirates. Mod Pathol. 2024 Sep;37(9):100542. doi: 10.1016/j.modpat.2024.100542. https://pubmed.ncbi.nlm.nih.gov/38897451/ 
5. Pozdnyakova O. Hematopathology Practice in the Digital Era: What has Changed? Int J Lab Hematol. 2025 Aug 14. doi: 10.1111/ijlh.14515. https://pubmed.ncbi.nlm.nih.gov/40808632/ 
6. Lee SH, Erber WN, Porwit A, Tomonaga M, Peterson LC; International Council for Standardization In Hematology. ICSH guidelines for the standardization of bone marrow specimens and reports. Int J Lab Hematol. 2008 Oct;30(5):349-64. doi: 10.1111/j.1751-553X.2008.01100.x. https://pubmed.ncbi.nlm.nih.gov/18822060/ 
7. Wang C-W, Huang S-C, Lee Y-C, Shen Y-J, Meng S-I, Gaol JL. Deep
   learning for bone marrow cell detection and classification on whole slide images. Med Image Anal. 2022;75:102270
   https://pubmed.ncbi.nlm.nih.gov/34710655/ 
8. Tayebi RM, Mu Y, Dehkharghanian T, Ross C, Sur M, Foley R, Tizhoosh HR, Campbell CJV. Automated bone marrow cytology using deep learning to generate a histogram of cell types. Commun Med (Lond). 2022 Apr 20;2:45. doi: 10.1038/s43856-022-00107-6. https://pubmed.ncbi.nlm.nih.gov/35603269/ 
9. Lewis JE, Shebelut CW, Drumheller BR, Zhang X, Shanmugam N, Attieh M, Horwath MC, Khanna A, Smith GH, Gutman DA, Aljudi A, Cooper LAD, Jaye DL. An Automated Pipeline for Differential Cell Counts on Whole-Slide Bone Marrow Aspirate Smears. Mod Pathol. 2023 Feb;36(2):100003. doi: 10.1016/j.modpat.2022.100003. https://pubmed.ncbi.nlm.nih.gov/36853796/ 
10. Dagan A, Demircioglu S. Diagnostic importance of bone marrow aspiration evaluation: A single-center study. Pak J Med Sci. 2022;38(4):811-815. doi: https://doi.org/10.12669/pjms.38.4.4797 
11. Rindy LJ, Chambers AR. Bone Marrow Aspiration and Biopsy. StatPearls [Internet]. Treasure Island (FL): StatPearls Publishing; 2025 https://www.ncbi.nlm.nih.gov/books/NBK559232/    
12. van der Walt J, van der Walt J, Orazi A, Arber DA. The Bone Marrow Biopsy. In: Diagnostic Bone Marrow Haematopathology. Cambridge University Press; 2021:1-13. https://www.cambridge.org/core/books/abs/diagnostic-bone-marrow-haematopathology/bone-marrow-biopsy/8087DE0285508DE41A07FF190BABCC24