Towards an explainable automatic phenotypic characterization of hereditary anemias
This study proposes an explainable framework for automatic hereditary anemias (HAs) phenotypic characterization based on machine learning (ML) regression models and quantitative phenotype ranking scores, such as the weighted-ranking score (WRS) developed by the RBC-Iron Genetics Lab at the University Federico II of Naples. WRS, based on seven clinical features, quantifies disease severity. Higher WRS values indicate more severe phenotypic expressions of HA. To avoid manual WRS estimation, a custom Multi-Layer Perceptron (MLP) model is introduced for automated computation, enhancing phenotypic characterization. Additionally, explainable Artificial Intelligence (xAI) is applied to identify key predictive features, providing clinicians with interpretable insights for disease assessment and personalized treatment planning.
Authors
- Antonella Nostroso (ORCID: https://orcid.org/0009-0007-3419-1641)
- Muhammad Suffian
- Lisa Miccio (ORCID: https://orcid.org/0000-0001-9427-881X)
- Immacolata Andolfo (ORCID: https://orcid.org/0000-0003-0493-812X)
- Pietro Ferraro (ORCID: https://orcid.org/0000-0002-0158-3856)
- Cosimo Ieracitano (ORCID: https://orcid.org/0000-0001-7890-2897)
- Nadia Mammone (ORCID: https://orcid.org/0000-0003-4962-3500)
- Pasquale Memmolo (ORCID: https://orcid.org/0000-0002-9607-7728)
- Francesco Carlo Morabito (ORCID: https://orcid.org/0000-0003-0734-9136)
- Vittorio Bianco (ORCID: https://orcid.org/0000-0003-1956-4976)
- Roberta Russo (ORCID: https://orcid.org/0000-0002-3624-7721)
- Anthony Iscaro
Institutions
- Institute of Applied Science and Intelligent Systems (IT)
- CEINGE Biotecnologie Avanzate Franco Salvatore (Italy) (IT)
- University of Reggio Calabria (IT)
- University of Naples Federico II (IT)
Publication Details
- Published
- 2025-08-04
- DOI
- https://doi.org/10.1117/12.3066660
- Primary Topic
- Biomedical Text Mining and Ontologies
- Type
- article
- Field-Weighted Citation Impact
- 0.00