RIIC - Representational Intelligence for Intensive Care

Research at RIIC

One of the most exciting opportunities unlocked by deep learning is the ability to compress complex, high-dimensional information into digestible concepts that humans can engage with directly. This is especially valuable in intensive care and anesthesia, where clinicians must make rapid, high-stakes decisions while navigating a constant stream of vital signs, laboratory trends, imaging, medications, and procedural context.

At RIIC, we aim to build and evaluate representational embeddings that distill this richness into clinically meaningful structures. By closely examining the latent spaces of deep learning models, we can detect subtle shifts in data distributions, retrieve similar cases from the past, and train models that adapt to evolving clinical environments. In this fashion, we can build applications that are truly safe, useful, and impactful in clinical practice.

Latest News & Updates

First RIIC Lab outing!

Last week, we had the first outing of the RIIC lab. We went to the Mi Barrio in Vienna for some delicious Venezuelan food and enjoyed the live music.

AI Auditing for Trustworthy Healthcare

In the IEEE EMBS Frontiers of Biomedical Imaging and Analysis Webinar Series co-hosted by Camila, Dr. Su-In Lee (University of Washington) will present her lab's work on auditing AI systems to uncover hidden biases and failure modes, advancing transparency, reliability, and trust in clinical AI.

Patient-friendly AI-generated radiologyreports

In our new JACR paper, we study how the presentation of AI-generated patient-friendly radiology reports influences patient understanding, highlighting the importance of interface design for effective patient communication.

Ensemble Monitoring Models

Our manuscript on monitoring black-box ICH detection models with deep ensembles, led by Zhongnan Fang at the AIDE lab, Stanford, was accepted to npj Digital Medicine.

Keynote @ IEEE ICDM

I will be giving a keynote at the OWAD: Open World Anomaly Detection in Dynamic and Evolving Environments workshop, held alongside the IEEE International Conference on Data Mining in Washington, D.C. on November 12.

Nathan's summer project

Nathan recently completed his Stanford Radiology internship project. He gave a great presentation at the RSL meeting and won two fellowships to continue his education at UCLA. Congratulations, Nathan!

CL in Medical Imaging Survey

Our survey on Continual Learning in Medical Imaging, led by Mohammad Qazi from Mohammad Yaqub's group, has been accepted to ACM Computing Surveys.

Synthetic data generation

Our video on diffusion models for Radiology is finally out :D

Bias Mitigation through Continual Learning

Our BM-CL preprint is out! Lucas Mansilla, from Enzo Ferrante's lab, proposes a novel approach to bias mitigation through continual learning that improves model performance for disadvantaged groups without any leveling-down.

ICML 2025

The very talented Yash Shah, who I had the pleasure to supervise alongside Ehsan Adeli, recently presented our paper "Confounder-Free Continual Learning via Recursive Feature Normalization" at ICML. Good job, Yash!

Deep Trajectory Clustering Highlight

Our work on Deep Trajectory Clustering for detecting alcohol drinking pathways in the latent space, presented at the Research Society on Alcohol, was selected as one of 10 highlighted works.

Lab members

Dr. Camila González

Camila González is a tenure-track Assistant Professor at the Medical University of Vienna and the Principal Investigator of the Representational Intelligence for Intensive Care (RIIC) lab, which develops machine learning applications for intensive care and perioperative medicine. She completed her PhD at the Technical University of Darmstadt and worked as a Postdoctoral Researcher at Stanford, where she focused on dynamic learning and monitoring for clinical settings with ongoing data collection. Her research has received multiple distinctions, including the MICCAI Young Scientist Award, the François Erbsmann Award at the IPMI conference, the BVM Award from the German Conference on Medical Image Computing, and the Freunde der TU Darmstadt Award for Outstanding Scientific Achievements. Her work has been featured in outlets such as Computer Vision News and the AI-Ready Healthcare podcast. Beyond her research, Camila has served as president of the MICCAI Student Board for two years and will take on the role of Career Development & Student Chair for MICCAI 2026. She is also a board member of the ContinualAI research society and webinar chair for IEEE EMBC.

View Profile

Dr. Amir Vahdani

Amir Vahdani is an MD from Tehran University of Medical Sciences. His background is in medicine, but he's spent the last several years working at the intersection of clinical research and machine learning — mostly on medical image analysis, with a focus on segmentation and uncertainty quantification. Before coming to Vienna, he was based at Tehran's Image-Guided Surgery Lab, where he worked on deep learning models for intraoperative ultrasound and chest CT images. Amir has also had collaborations with the Hospital for Special Surgery in New York on musculoskeletal imaging, and with Iran's Health Insurance organization on EHR data mining, among others.

Georgina Petropoulou

Georgina Petropoulou is a PhD Researcher at the Medical University of Vienna, specializing in Representational Intelligence in Intensive Care Medicine. Her career spans several institutions, most recently including CERN, where she engineered Monte Carlo simulations and Kafka-based data pipelines for high-energy physics with the strategic goal of translating these models into nuclear medicine, radiopharmaceuticals, and radiotherapy planning. At King’s College London, Georgina conducted research into machine learning-enabled assessments of neonatal cortical brain development, utilizing MRI to identify biomarkers for novel neuroprotective treatments and automated disease diagnosis, following her MSc in Medical Engineering and Physics with Distinction. Her experience in generative and diffusion models was further refined at Imperial College London, focusing on complex medical image segmentation across MRI, CT, and X-ray modalities. With additional research foundations from University College London (UCL) and years of industrial innovation—paired with collectively winning the 2023 Technological Innovation Edison Award for developing AI products from scratch—Georgina is dedicated to bridging the gap between AI, medical engineering, and life-saving clinical applications.

Helena Hazeu

Helena Hazeu is a PhD candidate in the Representational Intelligence for Intensive Care (RIIC) lab where she contributes to improved critical care with AI. She holds a BSc in mechanical engineering from the Technical University of Vienna and a MSc in Computational Cognitive Science from the Rijksuniversiteit Groningen, complemented by over 2 years of industry experience developing software and AI solutions for energy and healthcare.

Join the lab

We are seeking motivated Master's students from the Medical University of Vienna or TU Wien who are interested in working at the intersection of representational intelligence and healthcare. If you are interested, please feel free to reach out to us directly.