MDS Unpacked: Diverse Hematology With Dysplasia At The Core

Myelodysplastic syndrome (MDS) represents a complex group of clonal hematopoietic disorders ¹ that continue to challenge clinicians and pathologists alike due to their heterogeneity, overlapping features, and variable risk of progression. 1  Characterized by ineffective hematopoiesis and morphologic dysplasia, MDS primarily  affects older adults and may evolve into secondary acute myeloid leukemia (sAML). Bone marrow examination remains the diagnostic cornerstone ^{1, 2} – an intricate process that demands both expertise and precision. Yet, as digital pathology and artificial intelligence (AI) advance, new solutions are emerging to enhance diagnostic accuracy, reproducibility, and workflow efficiency.

The diverse spectrum of MDS – Epidemiology and Clinical Context

Once considered a rare condition, recent data suggest that MDS may be more prevalent than once believed. An estimated 60,000 individuals in the United States are currently living with MDS, with more than 10,000 new diagnoses each year. ³ 

The median age at diagnosis exceeds 70 years, with an overall incidence of 3–5 cases per 100,000 person-years—rising to more than 50 per 100,000 among those aged 75 and older. MDS can also occur in younger adults, where hereditary predisposition should be considered, particularly in patients younger than 40. Such syndromes may affect multiple organ systems and heighten the risk of treatment-related complications and secondary malignancies. ¹

Key demographic patterns include:

• Age-related risk: Incidence increases dramatically with age, particularly after 75 years ¹
• Gender distribution: Modest male predominance, except for the isolated del(5q) subtype, which occurs more frequently in women ¹
• Therapy-related cases: Approximately 10% of cases follow exposure to chemotherapy or radiation ¹

MDS is clinically diverse and is associated with other comorbid chronic conditions, which can complicate diagnosis and management. Accurate identification and quantification of abnormal bone marrow cell populations are crucial, as delayed or incorrect interpretation of aspirate findings can lead to misdiagnosis. Individualized risk assessment—considering both disease biology and patient-specific factors—remains central to guiding optimal treatment strategies. ^{1, 4}

Bone marrow examination for MDS

Morphologic evaluation of the bone marrow (BM) remains the gold standard for diagnosing MDS. ² Morphological diagnosis of MDS is based on the presence of dysplasia affecting ≥10% of cells in at least one hematopoietic lineage, together with bone marrow blast levels ranging from 5% to 19%. ¹

To improve diagnostic accuracy, an international working group has established minimal diagnostic criteria and related co-criteria designed to guide clinicians in recognizing features suggestive of MDS. These criteria focus on signs of abnormal bone marrow function, including: ¹

• Increased stem cell proliferation
• Clonal expansion of myeloid cells
• Atypical immunophenotypic patterns
• Distinct gene expression profiles

Complementing these standards, the European LeukemiaNet recommends several mandatory investigations for a definitive diagnosis. These include detailed evaluation of peripheral blood smears, bone marrow biopsy, bone marrow aspirate, and cytogenetic analysis. ¹  

Recognizing Dysplasia in Bone Marrow Lineages

Cytopenia and dysplasia remain central to diagnosis; however, evaluating dysplastic changes is inherently subjective and may vary between observers. ¹

Although dysplasia is fundamental to the diagnosis of MDS, its considerable morphologic diversity and irregular presentation can make clinical interpretation challenging. ⁵ 

Distinct dysplastic features may appear across the three major hematopoietic lineages:

• Dyserythropoiesis may be characterized by nuclear budding, internuclear bridging, karyorrhexis, multinuclearity, megaloblastoid changes, ring sideroblasts, vacuolization, and periodic acid–Schiff (PAS) positivity ¹
• Dysgranulopoiesis can manifest as abnormal cell size, nuclear hypo- or hypersegmentation, reduced or absent granules, and the presence of Pseudo–Chédiak–Higashi granules, Döhle bodies, or Auer rods ¹
• Dysmegakaryopoiesis typically presents with micromegakaryocytes, nuclear hypolobation, or multinucleation ¹

Blast cells and prognostic classification

Blast percentage is a key determinant of disease classification and prognosis. The minimal diagnostic criteria for MDS require the following  specific BM alterations: ⁶

• Dysplasia affecting ≥10% of one or more major hematopoietic lineages
• ≥15% ring sideroblasts
• 5–19% myeloblasts in BM smears 

An elevated blast count is considered indicative of myelodysplasia, particularly useful in aiding the diagnosis of advanced MDS. However , early forms of MDS with subtle morphological abnormalities are diagnosed primarily by excluding other conditions.  ⁶

To determine BM blast percentage, a 500-cell differential count of all nucleated cells in a smear or trephine imprint is recommended. ⁷

Blast percentage also plays a critical role in prognosis and disease classification, aiding in the differentiation of  higher-risk MDS from acute leukemia. Notably, the distinction between high risk MDS and acute leukemia is increasingly blurred, as acute leukemia may now be defined with blast counts ranging from 10% to 30%, depending on genetic profiles. ²

AI in bone marrow evaluation and reporting for MDS

“AI in MDS could be utilized to improve the accuracy and speed of reading and quantification of dysplastic cells.” ⁸

Machine learning (ML) and artificial intelligence (AI) are showing growing potential in and transforming the evaluation of myeloid neoplasms, particularly MDS. Recent studies have demonstrated that supervised learning approaches can identify biomarkers and diagnostic indicators, supporting clinical decision-making. AI has been applied using multimodal data from flow cytometry, bone marrow cells, and peripheral blood cells to aid in MDS diagnosis. ⁹

In addition, AI is increasingly used to analyze bone marrow specimens directly, focusing on the detection of blasts and the classification of normal versus dysplastic hematopoietic cells. Early research has employed deep learning to automatically distinguish cell lineages and identify dysplasia in bone marrow aspirates from patients with MDS, highlighting the potential of these technologies to complement traditional morphological assessment. ⁸

Artificial Intelligence in Bone Marrow Evaluation

AI and machine learning (ML) are rapidly transforming the evaluation of myeloid neoplasms. Supervised learning models have shown promise in identifying diagnostic biomarkers, classifying dysplastic cells, and integrating multimodal data—including flow cytometry, peripheral blood, and genomic information. ⁹

AI-based systems can: ^{8, 9}

• Detect and quantify dysplastic and normal cells across hematopoietic lineages
• Enhance reproducibility in blast detection and lineage classification
• Support clinicians in risk stratification and prognosis prediction

Early deep learning applications have demonstrated the ability to automatically distinguish cell lineages and identify dysplasia in bone marrow aspirates—supporting the role of AI as a complement to expert morphological assessment. ⁸

Scopio Labs X100: Redefining Bone Marrow Evaluation in MDS

The Scopio Labs X100 Full Field Bone Marrow Aspiration (BMA) System is the first clinical-grade digital platform for comprehensive BMA evaluation, combining whole-slide imaging (WSI) at 100X oil-immersion resolution with an AI-enabled decision support system (DSS) for remote analysis and reporting. ¹⁰

Key Features and Capabilities

The system uses a computational photography approach to reconstruct high-resolution images from low-resolution full-field captures, scanning slides in less than five minutes per cm² without altering standard Romanowsky or Prussian blue preparation protocols.

AI-Based Decision Support System:

• Analyzes, pre-classifies, and quantifies cells across trilineage hematopoietic elements (myeloid, erythroid, megakaryocytic)
• Identifies lymphocytes and plasma cells
• Enables operator review and adjustment of both region-of-interest (ROI) designations and individual cell classifications
• Provides interactive, ICSH-standardized reporting

Workflow Integration:

• Browser-based application requires no software installation
• Accessible securely from any workstation within a medical facility network or via remote connection
• Supports touch imprint slides
• Integrates with Laboratory Information Systems via HL7 or DICOM

Validated Performance

Validation studies demonstrated high correlation with reference manual microscopy methods: ¹⁰

• Overall agreement for BMA assessment: 91.1%
• Efficiency: ~90%
• Sensitivity: >81%
• Specificity: >92% for both Romanowsky- and Prussian blue-stained samples
• Repeatability and reproducibility: Coefficients of variation below 20% for discrete measurements and standard deviations below 5% for differential counts

The platform enables rapid, accurate digital BMA analysis, facilitating remote expert review, research into normal and neoplastic hematopoiesis, and standardized reporting. ¹⁰

Conclusion
Myelodysplastic syndromes exemplify the complexity of modern hematopathology—demanding deep expertise, careful interpretation, and multidisciplinary collaboration. As AI-driven digital BMA examination continues to evolve, it offers powerful platforms to standardize and accelerate MDS diagnosis, reduce variability, and improve clinical decision-making. At the heart of MDS lies dysplasia—diverse in form but unified in its diagnostic importance. By combining expert insight with advanced imaging and computational intelligence, clinicians are now better equipped than ever to unpack the diversity within MDS and deliver more precise care.

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. Chanias I, Stojkov K, Stehle GT, Daskalakis M, Simeunovic H, Njue LM, et al, On Behalf Of The Swiss Mds Study Group. Myelodysplastic Syndromes in the Postgenomic Era and Future Perspectives for Precision Medicine. Cancers (Basel). 2021;13(13):3296. doi: 10.3390/cancers13133296 
2. Oster HS, Van de Loosdrecht AA, Mittelman M. Diagnosis of myelodysplastic syndromes: the classic and the novel. Haematologica. 2025;110(2):300-311. doi: 10.3324/haematol.2023.284937
3. Ridgeway JA, Tinsley S, Kurtin SE. Practical Guide to Bone Marrow Sampling for Suspected Myelodysplastic Syndromes. J Adv Pract Oncol. 2017;8(1):29-39. Epub 2017 Jan 1. PMID: 29900015; PMCID: PMC5995536.
4. Tayebi RM, Mu Y, Dehkharghanian T, Ross C, Sur M, Foley R, et al. Automated bone marrow cytology using deep learning to generate a histogram of cell types. Commun Med (Lond). 2022;2:45. doi: 10.1038/s43856-022-00107-6
5. Liang C, Li J, Cheng J, Chen S, Ye Z, Zhang F, Wang Z, et al. Characteristics of bone marrow cell dysplasia and its effectiveness in diagnosing myelodysplastic syndrome. Hematology. 2018;23(2):65-76. doi: 10.1080/10245332.2017.1347247
6. Invernizzi R, Quaglia F, Porta MG. Importance of classical morphology in the diagnosis of myelodysplastic syndrome. Mediterr J Hematol Infect Dis. 2015;7(1):e2015035. doi: 10.4084/MJHID.2015.035
7. Gupta G, Singh R, Kotasthane DS, Kotasthane VD. Myelodysplastic syndromes/neoplasms: recent classification system based on World Health Organization Classification of Tumors – International Agency for Research on Cancer for Hematopoietic and Lymphoid Tissues. J Blood Med. 2010;1:171-82. doi: 10.2147/JBM.S12257
8. Lee N, Jeong S, Park M-J, Song W. Deep learning application of the discrimination of bone marrow aspiration cells in patients with myelodysplastic syndromes. Sci Rep. 2022;12:18677. doi:10.1038/s41598-022-21887-w
9. Stagno F, Mirabile G, Rizzotti P, Bottaro A, Pagana A, Gangemi S, Allegra A. Using Artificial Intelligence to Enhance Myelodysplastic Syndrome Diagnosis, Prognosis, and Treatment. Biomedicines. 2025;13(4):835. doi: 10.3390/biomedicines13040835.
10. Bagg A, Raess PW, Rund D, Bhattacharyya S, Wiszniewska J, Horowitz A, et al. Performance Evaluation of a Novel Artificial Intelligence-Assisted Digital Microscopy System for the Routine Analysis of Bone Marrow Aspirates. Mod Pathol. 2024;37(9):100542. doi: 10.1016/j.modpat.2024.100542.
11. Memorial Sloane Kettering Cancer Center. Classification and Staging of Myelodysplastic Syndrome (MDS). [Internet] Available from https://www.mskcc.org/cancer-care/types/myelodysplastic-syndrome/diagnosis/classification-staging [Accessed 29 October 2025]