ISR-Lisboa has confirmed funding for three major research projects, aiming to transform the future of healthcare through artificial intelligence, multimodal data integration, and precision medicine. These projects, in fields ranging from cardiovascular health, cancer care, and surgical optimisation, will deepen ISR-Lisboa’s role in Europe’s shift toward personalised, data-driven healthcare solutions.
“These three projects represent the future of medicine—multi data-centric, patient-specific, and powered by AI,” said PI Catarina Barata. “We are proud to lead initiatives that not only advance scientific knowledge but directly improve patient outcomes and healthcare system efficiency.”
A couple of projects have the ultimate goal of creating PathFinder – a data search platform for specific use cases in medicine:
Nextgen HOP-on: Personalising Cardiovascular Medicine Through AI
In partnership with a NextGen consortium composed of 22 partners, ISR-Lisboa is joining a Europe-wide mission to tackle the leading cause of death in the EU, cardiovascular disease. This Nextgen HOP-on project aims to personalise cardiovascular treatments by integrating a vast range of medical data: clinical, biosignals, imaging, and genomics.
Multimodal data holds the key to individualised care, but it comes with challenges, like data scarcity, missing modalities across institutions, and differing formats. ISR-Lisboa will contribute novel solutions using transformer-based AI models capable of handling incomplete datasets and building robust, unified representations of patient data. These tools will help clinicians make better-informed decisions tailored to each individual’s unique health profile.
Transforming Oncology: A Multi-Modal Framework for Personalized Cancer Care
Cancer treatment is undergoing a paradigm shift to highly personalised therapies. This project aims to accelerate this transformation with a multi-modal deep learning framework that brings together diagnostic imaging, pathology slides, clinical records, and molecular biomarkers into a single tumour representation.
This innovative project will address long-standing challenges in the field: incomplete datasets, lack of standardisation, and difficulty fusing different data types. By exploring redundancy and patient similarity, the framework will be designed to infer missing information, improving model robustness and real-world applicability.
Beyond the algorithmic breakthroughs, the project also aims to launch MMIST, a first-of-its-kind platform for curating and sharing standardized multi-modal cancer datasets. With this, the research community gains a much-needed foundation for benchmarking, collaboration, and reproducibility.
OptSurgAI: Optimizing Surgical Workflows and Outcomes with Artificial Intelligence
The third project, OptSurgAI, tackles surgery, one of the most critical and costly areas of healthcare. As hospitals increasingly adopt minimally invasive techniques, new challenges arise in predicting procedure duration, patient outcomes, and providing feedback to surgeons. OptSurgAI combines video data from laparoscopic procedures with radiology data images and patient metadata to deliver two powerful AI tools:
- A system to predict surgery duration and potential complications before and during the procedure.
- A model to forecast post-operative outcomes within a 30-day window.
The project goes a step further by using “affordance modeling” to assess surgeon skill, offering real-time feedback, and accelerating training. This feature could revolutionize surgical education, the training of surgical robots, and ensuring consistent quality of care. Led by a team of AI specialists at ISR-Lisboa and surgeons from Faculdade de Medicina da Universidade de Lisboa, together with Hospital Fernando Fonseca as collaboratorHospital Fernando Fonseca, OptSurgAI bridges technical innovation with clinical practice. All collected data will be released in an open-source format, encouraging continued research and investment in surgical AI.
