Active
Multimodal Integration for Precision Medicine
Collaboration with Gustave Roussy Institute in the context of the new PRISM center for Precision Medicine.
Research Focus
My research lies at the intersection of mathematics, artificial intelligence, and oncology. I focus on developing novel approaches to address fundamental challenges in modalities integration and enhancing knowledge transfer to provide personalized treatment insights.
5
Publications
3
Active Projects
3
Collaborations
Main Project
Multimodal Integration for Precision Medicine
Active Project
Project Overview
Multimodal integration for precision medicine leverages deep learning to combine diverse data types, such as genomics, histopathology, and clinical records, to provide personalized treatment insights. By using explainable methods, these models enhance interpretability, allowing clinicians to understand the underlying decision-making process, thus increasing trust and facilitating more informed clinical decisions.
Key Challenges
- Data Heterogeneity: Integrating diverse data types, such as genomics, histopathology, and clinical records, which have different formats, scales, and noise levels.
- Missing Modalities: Handling incomplete data where not all modalities are available for every patient, without compromising model performance.
- Interpretability: Developing explainable models that maintain high accuracy while providing transparent and clinically relevant insights to support decision-making.
- Generalization and Robustness: Ensuring that models generalize well across different patient populations and clinical settings, maintaining robustness to variations in data quality.
Impact
Multimodal integration using deep learning explainable methods holds the potential to improve precision medicine by providing more personalized treatment insights and supporting better-informed clinical decisions.
Research Areas
Deep Learning Methods for Multimodal Integration
Developing methods for integrating multiple data modalities in precision medicine.
Histopathology Analysis
Using deep learning to extract meaningful patterns from tissue images, aiding in disease diagnosis and treatment planning.
Spatial Omics
Leveraging Spatial Omics technologies to map molecular profiles within tissue architecture, providing insights into cellular interactions and disease mechanisms.
Multimodal Integration in Precision Medicine
Our research focuses on developing innovative approaches to integrate multiple data modalities in precision medicine. Through the CustOmics and Multimodal CustOmics frameworks, we work on combining various data types such as genomics, transcriptomics, and medical imaging to better understand disease mechanisms and improve treatment strategies.
Key Aspects:
- Development of deep learning architectures for multi-omics integration
- Integration of imaging and molecular data
- Interpretable machine learning methods
- Applications in precision oncology
Graph-based Methods for Biological Data
We develop graph-based approaches to analyze complex biological data, particularly focusing on spatial transcriptomics and histopathology analysis. Our Novae framework provides a sophisticated solution for understanding spatial organization in biological tissues.
Key Aspects:
- Graph neural networks for spatial data analysis
- Zero-shot domain inference
- Batch effect correction
- Spatial domain analysis
Deep Learning in Healthcare
Our research in deep learning for healthcare focuses on developing interpretable and robust solutions for medical applications. Through projects like Hyper-AdaC, we work on creating adaptive solutions that can handle the complexity of medical data while providing interpretable results.
Key Aspects:
- Interpretable deep learning models
- Adaptive clustering methods
- Survival analysis
- Clinical validation
Research Laboratory
Laboratory of Mathematics & Computer Science
CentraleSupélec • Université Paris-Saclay
The MICS Laboratory (Mathematics and Informatics of Complex Systems) brings together researchers in applied mathematics and computer science. Our research focuses on modeling, analyzing, and controlling complex systems, with applications in various fields including healthcare, biology, and industry.
Academic Excellence
Located at CentraleSupélec, one of France's leading engineering schools
Healthcare Partnership
Close collaboration with Gustave Roussy Cancer Institute
Research Network
Part of Université Paris-Saclay's research ecosystem
Research Collaborations
University Partners
- Gustave Roussy Institute
- Dana-Farber Institute
- Karolinska Institute