AIDA
Explainable multimodal deep learning for personalized oncology, with a focus on robust fusion, missing modalities, and clinical validation in non-small cell lung cancer.
Clinical Application
AIDA
Explainable multimodal deep learning for personalized oncology, with a focus on robust fusion, missing modalities, and clinical validation in non-small cell lung cancer.
Overview
AIDA develops explainable multimodal deep learning for healthcare scenarios where single-modality pipelines are not enough. The project investigates when and how to fuse heterogeneous evidence, how to learn stronger shared representations, and how to maintain robustness when data are incomplete or some modalities are missing.
The translational application is personalized oncology in non-small cell lung cancer, where radiomic, pathomic, and electronic health record data are combined to predict response, relapse, progression-free survival, and overall survival. The project also studies clinical trust through explainability mechanisms designed for physicians and domain experts.
Research Directions
- Multimodal representation learning across radiomics, pathomics, and electronic health records.
- Explainable AI methods for attention, counterfactual reasoning, and user-facing clinical transparency.
- Prospective validation in non-small cell lung cancer with comparison against conventional clinical markers.
Related Publications
- Handling Missing Modalities in Multimodal Survival Prediction for Non-Small Cell Lung Cancer2026multimodal learning clinical prediction medical imagingAIDA digilung
- Predicting lung cancer survival with attention-based CT slices combination2026
- Next-Gen Health: from Multimodal AI to Foundation Models2025foundation models generative AI multimodal learning medical imaging
- Hierarchical embedding attention for overall survival prediction in lung cancer from unstructured EHRs2025oncology foundation models generative AI
- Benchmarking GAN-Based vs Classical Data Augmentation on Biomedical Images2025
- A systematic review of intermediate fusion in multimodal deep learning for biomedical applications2025multimodal learning
- Transformer-based AI approach to unravel long-term, time-dependent prognostic complexity in patients with advanced NSCLC and PD-L1 ≥50%: insights from the pembrolizumab 5-year global registry2025clinical prediction oncology explainability
- MARIA: A multimodal transformer model for incomplete healthcare data2025multimodal learning foundation models clinical prediction
- Can Foundation Models Really Segment Tumors? A Benchmarking Odyssey in Lung CT Imaging2025foundation models medical imaging
- Multi-stage intermediate fusion for multimodal learning to classify non-small cell lung cancer subtypes from CT and PET2025radiomics oncology multimodal learning
- Exploring Negated Entites for Named Entity Recognition in Italian Lung Cancer Clinical Reports2024
- A deep learning approach for overall survival prediction in lung cancer with missing values2024clinical prediction oncology
- Toward a Multimodal Deep Learning Approach for Histological Subtype Classification in NSCLC2024oncology multimodal learning medical imaging
- Named Entity Recognition in Italian Lung Cancer Clinical Reports using Transformers2023oncology foundation models medical imaging