Publications

The exponential growth of healthcare data, driven by advancements in medical research and digital health technologies, has underscored the critical need for interoperability and standardization. However, the heterogeneous nature of real-world clinical data poses significant challenges to ensuring seamless data exchange and secondary use for research purposes. These challenges include syntactic inconsistencies (e.g., variable use of terminologies like ICD-10 vs SNOMED CT), semantic mismatches (e.g., differing conceptualizations of disease staging across institutions), and structural fragmentation (e.g., laboratory results encoded in free text rather than structured fields). Fast Healthcare Interoperability Resources (FHIR) has emerged as a leading standard for structuring and harmonizing healthcare data, enabling integration across diverse systems. This work presents a FHIR-based transformation pipeline that leverages Resource Description Framework (RDF) to convert raw, conceptually heterogeneous oncology data into research-ready, semantically enriched datasets. By representing FHIR resources as RDF graphs, our approach enables semantic interoperability, enhances data linkage across heterogeneous sources, and supports automated reasoning through ontology-based queries and inference mechanisms. The pipeline employs a templated conversion strategy, allowing for the declarative definition of mappings that enable domain experts to focus on the data model. In Cancer Virtual Lab, we applied this methodology to a real-world oncology dataset comprising 36,335 anonymized patient records, successfully converting 1,093,705 clinical records into 1,151,559 distinct RDF-based FHIR resource types. The process incorporated syntactic and semantic validation, along with expert review, to ensure technical correctness and clinical relevance. Our results demonstrate the feasibility of semantically integrating oncology data using FHIR and RDF, fostering machine-readable, interoperable knowledge representation. This enriched representation supports data quality monitoring and improvement, data harmonization, longitudinal analysis, advanced analytics, and AI-driven decision support, promoting large-scale secondary use.

Brain tumor segmentation is a crucial task in medical imaging that involves the integrated modeling of four distinct imaging modalities to identify tumor regions accurately. Unfortunately, in real-life scenarios, the full availability of such four modalities is often violated due to scanning cost, time, and patient condition. Consequently, several deep learning models have been developed to address the challenge of brain tumor segmentation under conditions of missing imaging modalities. However, the majority of these models have been evaluated using the 2018 version of the BraTS dataset, which comprises only $285$ volumes. In this study, we reproduce and extensively analyze the most relevant models using BraTS2023, which includes 1,250 volumes, thereby providing a more comprehensive and reliable comparison of their performance. Furthermore, we propose and evaluate the adoption of Mamba as an alternative fusion mechanism for brain tumor segmentation in the presence of missing modalities. Experimental results demonstrate that transformer-based architectures achieve leading performance on BraTS2023, outperforming purely convolutional models that were instead superior in BraTS2018. Meanwhile, the proposed Mamba-based architecture exhibits promising performance in comparison to state-of-the-art models, competing and even outperforming transformers. The source code of the proposed approach is publicly released alongside the benchmark developed for the evaluation: https://github.com/AImageLab-zip/IM-Fuse.

The evaluation of machine-generated image captions is a complex and evolving challenge. With the advent of Multimodal Large Language Models (MLLMs), image captioning has become a core task, increasing the need for robust and reliable evaluation metrics. This survey provides a comprehensive overview of advancements in image captioning evaluation, analyzing the evolution, strengths, and limitations of existing metrics. We assess these metrics across multiple dimensions, including correlation with human judgment, ranking accuracy, and sensitivity to hallucinations. Additionally, we explore the challenges posed by the longer and more detailed captions generated by MLLMs and examine the adaptability of current metrics to these stylistic variations. Our analysis highlights some limitations of standard evaluation approaches and suggest promising directions for future research in image captioning assessment.

Federated Learning (FL) enables collaborative model training across distributed clients while preserving data privacy. While existing approaches for aggregating client-specific classification heads and adapted backbone parameters require architectural modifications or loss function changes, our method uniquely leverages intrinsic training signals already available during standard optimization. We present LIVAR (Layer Importance and VARiance-based merging), which introduces: i) a variance-weighted classifier aggregation scheme using naturally emergent feature statistics, and ii) an explainability-driven LoRA merging technique based on SHAP analysis of existing update parameter patterns. Without any architectural overhead, LIVAR achieves state-of-the-art performance on multiple benchmarks while maintaining seamless integration with existing FL methods. This work demonstrates that effective model merging can be achieved solely through existing training signals, establishing a new paradigm for efficient federated model aggregation. The code is available at https://github.com/aimagelab/fed-mammoth