TriHemo-MCV: A Tri-Class Clinical Dataset for Anemia and Compensated Microcytosis

TriHemo-MCV introduces a novel tri-class framework for hematological diagnosis, including the often-overlooked 'Compensated Microcytosis' phenotype. It contains 5,037 de-identified patient records from four tertiary hospitals in Bangladesh, each with Complete Blood Count parameters and lifestyle questionnaire variables. The dataset was curated by Susmoy Biswas and released on figshare in April 2026 under a CC-BY-4.0 license.

Use Cases

• Multi-class anemia classification based on CBC parameters and lifestyle features.
• Hemoglobinopathy screening using the compensated microcytosis class.
• Developing explainable AI models in hematology based on structured clinical features.
• Simulating federated learning scenarios using data from multiple hospitals.
• Training clinical decision support systems for laboratory diagnosis.

Strengths

• 5,037 patient records provide a substantial sample size for model development.
• Includes a novel third class (Compensated Microcytosis) with 1,641 samples, enabling more nuanced modeling.
• Data is sourced from four hospitals, which likely contributes to demographic and instrumental diversity.
• Labels were assigned based on predefined thresholds and verified through multi-level clinical review.

Limitations

• Data is from a specific geographic region (Bangladesh), which may limit generalizability to other populations.
• The dataset size of 5,037 records and 1.1 MB is relatively small for some modern deep learning applications.
• Description metadata is limited; actual data quality and feature distributions require manual inspection after download.

Provenance

Source

Retrospectively collected from four tertiary hospitals in Bangladesh.

Collection Method

Clinical records and structured lifestyle questionnaires.

Freshness

Last updated 2026-04-25 09:10:31

Geography

Bangladesh