ICAI Stroke Lab

From 112 To Rehabilitation

In this lab, we investigate how data-driven AI-based methods can support healthcare professionals to improve the outcome of stroke patients by optimizing the whole stroke patient journey. We will focus on AI-based patient stratification, AI-based functional outcome prediction after endovascular therapy, better therapeutic outcomes through AI-based image guidance in endovascular interventions, and data-driven individualized stroke recovery policies. In addition, this lab will study the acceptance of such AI-models by clinical decision makers in daily clinical routine.

News

Visiting the ICAI Lab day
ICAI Stroke Lab’s participation in the ICAI Lab Day
The ICAI Stroke Lab officially opens
Speeches, cake, and group pictures...

The Team

Ellen

In the lab, Ellen will generally focus on the user acceptance and implementation of AI tools in stroke care. She is especially eager to delve deeper into the ethical and sociotechnical dilemmas through the implementation journey. She has a background in Health Sciences, where she received her bachelor’s degree at the University of Twente and completed her research master's degree at the Erasmus University. Apart from the academic setting, she enjoys working out at the gym, sketching portraits and recently she tried to learn to play the guitar.

hu@eshpm.eur.nl

Personal PageORCID
Peter

Peter focuses on utilizing AI methods to enhance stroke care in the crucial prehospital phase. Understanding the importance of time in stroke treatment, Peter strives to develop models that can contribute to shorter response and treatment times, ultimately leading to improved functional outcomes for stroke patients. His specific area of interest lies in the period between the initial emergency call and the patient’s arrival at the hospital.

p.vanhulst@erasmusmc.nl

Personal PageORCID
Xi

Xi holds a background in Clinical Medicine and completed her master degree in Clinical Research at Erasmus MC, where she focused on the study of calcification in cerebral arteries. With a profound interest in neurology and exploring AI’s role in healthcare, her current research focuses on in-hospital clinical decision support and outcome predictions for stroke patients. Given the heterogeneity in the causes and outcomes of stroke patients, she believes AI has the potential for identifying optimal treatment for each individual patient.

x.li.1@erasmusmc.nl

Personal PageORCID
Govert

Studied Human Movement Sciences and Biomedical Engineering. Like a true Movement Scientist, he is fascinated by human motion and often prefers to be in motion himself, particularly when it involves running. Govert's research is centered around stroke rehabilitation, in which he focuses on the development and validation of prognostic models that predict functional outcome.

g.vandergun@erasmusmc.nl

Personal PageORCID
Frank

Frank is a computer scientist and medical doctor working on the implementation of state-of-the-art computer vision methods within the context of (neuro-)interventional radiology. The goal of his PhD project is to assist the operator during the procedure, for instance by using deep learning to visualize occluded blood vessels. Previously, Frank has worked on functional outcome prediction using imaging. In his free time, Frank thoroughly enjoys making music.

f.tenijenhuis@erasmusmc.nl

Personal PageORCID

About the Lab

ICAI Stroke Lab develops Machine Learning and AI models to enhance stroke patient outcomes across their journey and evaluates their clinical applicability. The lab is a collaboration between Erasmus MC and the Erasmus University of Rotterdam, who bring in clinical domain and academic expertise, and Philips interested in the development of products for clinical decision support. At the Stroke Lab we are dedicated to improve outcomes for stroke patients, a condition that affects approximately 1 million Europeans each year, with a 25% lifetime risk [1]. Stroke has a 30% overall mortality rate and stands as a leading cause of death in developed countries [2], with up to 50% of patients facing permanent disability [3]. Presently, over 6 million Europeans live with the aftermath of stroke [4], incurring significant treatment and rehabilitation costs totaling €38 billion annually in Europe.

Aim

Our goal is to enhance stroke patient outcomes and reduce associated socioeconomic burdens. We will achieve this by providing data-driven modeling tools to assist healthcare professionals in decision-making, thus optimizing the entire stroke patient journey from the initial emergency call to rehabilitation. Specifically, our aims are to improve: patient stratification: quickly and accurately sort patients into groups for better treatment; functional outcome: predict how well patients will recover; the use of images to guide treatment and improve its effectiveness; recovery plan based on patients’ current ability and constantly updating forecasts for stroke recovery. These plans will help deciding if a patient should go home or get care elsewhere, as well as selecting the right rehabilitation.  acceptance of such AI-models by clinical decision makers .

Plans

The lab consists of five different research lines. The first four PhD students will focus on the different parts of a stroke patient's care pathway, with the aim to develop tools useful in that specific stage. The stages are:

  1. prehospital stage
  1. patient stratification, clinical decision support and therapy preparation
  1. image guided intervention
  1. rehabilitation

The fifth PhD student will investigate, across all parts of the stroke patient journey, aspects related to the acceptance of machine learning tools in daily clinical practice.

The overall approach of the lab will be to

  • Implement a FAIR data infrastructure and collect and store data (clinical data, imaging data, interventional data, rehabilitation data, patient-reported data) along the full patient journey;
  • Utilize these data to develop, optimize and validate machine learning tools for optimal decision making across the care chain; we will here focus on diagnostic and prognostic models. We will investigate accuracy, reproducibility of the models, and will investigate how to communicate algorithm output to the end user;
  • Implement and validate decision support tools (dashboards, prediction models) directly accessible to clinicians;  
  • Ensure the acceptance and integration of these models into daily clinical routine.  

[1] Meairs S, Wahlgren N, Dirnagl U, Lindvall O, Rothwell P, Baron JC, Hossmann K, Engelhardt B, Ferro J, McCulloch J, Kaste M, Endres M, Koistinaho J, Planas A, Vivien D, Dijkhuizen R, Czlonkowska A, Hagen A, Evans A, De Libero G, Nagy Z, Rastenyte D, Reess J, Davalos A, Lenzi GL, Amarenco P, Hennerici M. Stroke research priorities for the next decade–A representative view of the European scientific community. Cerebrovasc Dis. 2006;22(2-3):75-82. doi: 10.1159/000093098. PMID: 16790993.

[2] Béjot Y, Bailly H, Durier J, Giroud M. Epidemiology of stroke in Europe and trends for the 21st century. Presse Med. 2016 Dec;45(12 Pt 2):e391-e398. doi: 10.1016/j.lpm.2016.10.003. Epub 2016 Nov 2. PMID: 27816343.

[3] Hankey GJ. Long-term outcome after ischaemic stroke/transient ischaemic attack. Cerebrovasc Dis. 2003;16 Suppl 1:14-9. doi: 10.1159/000069936. PMID: 12698014.

[4] Truelsen T, Ekman M, Boysen G. Cost of stroke in Europe. Eur J Neurol. 2005 Jun;12 Suppl 1:78-84. doi: 10.1111/j.1468-1331.2005.01199.x. PMID: 15877785.