Answers are grounded in this paper and cite their sources.
Study implements morphometric analysis and machine learning for automated red blood cell screening, suggesting improved detection methods.
Accurate classification of red blood cells (RBCs) is essential for the early detection of hematological disorders. This study presents a dual-model approach that integrates morphometric analysis with machine learning to systematically evaluate the trade-offs between interpretability and predictive performance in RBC classification. Blood smear samples were collected from diverse sources, totaling 157 peripheral blood smears, to capture variability across clinical settings. For controlled analysis, a representative smear image was selected, from which 110 RBCs were segmented and annotated to demonstrate the proposed feature extraction and classification framework. Geometric descriptors, including area, perimeter, circularity, and eccentricity, were extracted and used to train a Support Vector Machine (SVM), achieving 98.19% accuracy, 97.47% precision, 100% recall, and a 98.72% F1-score. In parallel, a Convolutional Neural Network (CNN) was trained on augmented RBC image patches derived from the same dataset. CNN achieved near-perfect validation accuracy under the constrained experimental setup, which may be influenced by the limited dataset size and controlled imaging conditions. however, this result is influenced by the limited sample size and should be interpreted cautiously rather than as indicative of clinical-level performance. The results demonstrate that SVM provides interpretable, feature-based classification aligned with hematological descriptors, while CNN captures complex spatial patterns for improved predictive capability. Overall, this study highlights the feasibility of integrating morphometric descriptors with machine learning for automated RBC screening and emphasizes the trade-offs between interpretability, performance, and data requirements in resource-constrained environments.
No takes yet. Share an insight, caveat, or question.
Lazaro et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: