Project

Back to overview

ICU-Cockpit: IT platform for multimodal patient monitoring and therapy support in intensive care and emergency medicine

English title ICU-Cockpit: IT platform for multimodal patient monitoring and therapy support in intensive care and emergency medicine
Applicant Keller Emanuela
Number 167195
Funding scheme NRP 75 Big Data
Research institution Neurochirurgische Intensivstation Neurochirurgische Klinik Universitätsspital Zürich
Institution of higher education University of Zurich - ZH
Main discipline Biomedical Engineering
Start/End 01.02.2017 - 30.04.2019
Approved amount 573'853.00
Show all

All Disciplines (5)

Discipline
Biomedical Engineering
Neurology, Psychiatry
Internal Medicine
Information Technology
Electrical Engineering

Keywords (5)

Data streaming; Intensive care medicine; Machine learning; Multimodal monitoring; Data mining

Lay Summary (German)

Lead
Das Projekt möchte eine grundlegende Entwicklung in der Notfall- und Intensivmedizin anstossen - und die Arbeitsweise im Klinikalltag bezüglich Diagnostik, Therapie und Risikomanagement wesentlich verbessern.
Lay summary

Auf der neurochirurgischen Intensivstation wird seit 2014 in Kollaboration mit der ETH Zürich, IBM Research Rüschlikon und dem Industriepartner Supercomputing Systems das Projekt "ICU-Cockpit" entwickelt. Dabei erfasst modernste Informationstechnologie die Daten zahlreicher medizintechnischer Geräte in Echtzeit und einer Auflösung von bis zu 200 Hertz. Die Daten werden zeitsynchronisiert und verschlüsselt gespeichert. Nun geht es darum, dass ICU-Cockpit Artefakte in den Biosignalen erkennen und eliminieren kann. Ausserdem entwickeln wir Algorithmen zur Früherkennung epileptischer Anfälle und sekundärer Hirndurchblutungsstörungen.

Die Halbwertszeit medizinischen Wissens beträgt wenige Jahre. Der Wissenszuwachs ist für Ärzte nicht mehr zu bewältigen. Zusätzlich nimmt in der "personalisierten Medizin" die Menge an vorhandenen Daten pro Patient exponentiell zu. In der Intensiv- und Notfallmedizin kommen noch Signale in Echtzeit von multiplen Sensoren im und am Körper hinzu. Diese Informationsflut kann – insbesondere in Notfallsituationen – nicht mehr zur raschen Entscheidungsfindung integriert werden.

Basierend auf "Data-Mining", maschinellem Lernen und künstlicher Intelligenz werden die enormen Mengen an gespeicherten Daten dazu verwendet, komplexe pathophysiologische Zusammenhänge zu modellieren, um Algorithmen für Frühalarmsysteme und Therapieempfehlungen zu entwickeln.

Schon die Reduktion von Signal-Artefakten und Fehlalarmen erhöht die Patientensicherheit auf der Intensivstation. Das Erkennen von Risikokonstellationen und die Vorhersage kritischer Komplikationen erlauben es, früher therapeutisch zu intervenieren. Therapieentscheide, heute oft empirisch gefällt, werden durch Datenanalysen und aktuellstes medizinisches Wissen untermauert. Aus der Datenstromanalyse in Echtzeit entstehen neue pathophysiologische Erkenntnisse – und neues Wissen durch selbstlernende Systeme.

Direct link to Lay Summary Last update: 26.07.2017

Lay Summary (French)

Lead
Le projet vise un développement fondamental dans le domaine de la médecine intensive et d’urgence ainsi qu’une amélioration de la manière de travailler au quotidien dans les hôpitaux en matière de diagnostics, de traitements et de gestion des risques.Le projet "ICU-Cockpit" est développé depuis 2014 aux soins intensifs de neurochirurgie de l’Hôpital universitaire de Zurich, en collaboration avec l’ETH Zurich, IBM Research à Rüschlikon et le partenaire industriel Supercomputing Systems. Une technologie informatique de pointe est utilisée pour traiter les données des nombreux appareils médicaux en temps réel avec une résolution allant jusqu’à 200 Hz. Les données sont synchronisées et enregistrées sous forme cryptée. Il s’agit maintenant de permettre à ICU-Cockpit de reconnaître et d’éliminer des artéfacts dans les biosignaux. Nous développons par ailleurs des algorithmes pour la détection précoce des crises d’épilepsie et des troubles vasculaires cérébraux secondaires.
Lay summary

Le savoir médical évolue très rapidement. La croissance des connaissances est telle que les médecins n’arrivent plus à la gérer. En "médecine personnalisée", le volume des données existantes par patient augmente par ailleurs de façon exponentielle. En médecine intensive et d’urgence s’ajoutent à cela des signaux en temps réel émis par de multiples capteurs placés sur et dans l’organisme. Ce flot d’informations, notamment dans des situations d’urgence, ne peut plus être intégré dans une prise de décision rapide.

Sur la base de l’exploration de données, de l’apprentissage automatique et de l’intelligence artificielle, les énormes quantités de données enregistrées sont utilisées afin de modéliser des mécanismes physiopathologiques complexes et de développer des algorithmes pour des systèmes d’alarme précoce et des recommandations thérapeutiques.

La réduction des fausses alarmes et des artéfacts dans les signaux augmente la sécurité des patients aux soins intensifs. La détection des constellations de risques et la prévision des complications critiques permettent d’intervenir plus tôt au niveau thérapeutique. Les décisions thérapeutiques, souvent prises aujourd’hui de manière empirique, sont étayées grâce des analyses de données et aux connaissances médicales les plus actuelles. De l’analyse des flux de données en temps réel naissent de nouvelles connaissances en physiopathologie ainsi qu’un nouveau savoir grâce à des systèmes d’apprentissage automatique.


Direct link to Lay Summary Last update: 26.07.2017

Lay Summary (English)

Lead
The project is aimed at initiating a fundamental development in emergency and intensive medicine - and bringing about a substantial improvement in the way diagnostics, treatment and risk management are handled in everyday clinical practice.
Lay summary

The “ICU cockpit” project has been under development in the neurosurgery intensive care unit since 2014 in partnership with ETH Zurich, IBM Research Rüschlikon and industry partner Supercomputing Systems. State-of-the-art information technology captures data from numerous pieces of medical equipment in real time and at a resolution up to 200 Hertz. The data are time-synchronised and encoded prior to storage. The task now is to enable the ICU cockpit to identify and eliminate artefacts in the biosignals. We are also developing algorithms for the early detection of epileptic seizures and secondary impairment of cerebral perfusion.

Medical knowledge has a half-life of just a few years. Doctors struggle to keep up with the explosion of information. In addition, the volume of data per patient is increasing exponentially in the field of personalised healthcare. In intensive and emergency medicine, the situation is compounded by real-time signals from multiple sensors in and on the body. In an emergency situation, in particular, it is not possible to integrate this flood of information rapidly into the decision-making process.

In an approach combining data mining, machine learning and artificial intelligence, the enormous volumes of stored data will be used to model complex pathophysiological situations for the purpose of developing algorithms for early warning systems and therapeutic recommendations.

The reduction of signal artefacts and false alarms increases patient safety in the intensive care unit. Identifying risk constellations and predicting critical complications permits earlier therapeutic intervention. Therapeutic decisions, which are often made on an empirical basis, are supported by data analysis and state-of-the-art medical knowledge. Real-time data flow analysis generates new pathophysiological information – and new knowledge thanks to self-learning systems.


Direct link to Lay Summary Last update: 26.07.2017

Responsible applicant and co-applicants

Employees

Publications

Publication
Decision tree analysis in subarachnoid hemorrhage - Prediction of outcome parameters during the course of aneurysmal subarachnoid hemorrhage using decision tree analysis.
Hostettler Isabel, Richter Johannes, SchmidJoseph, NeidertMarian, SeuleMartin, BossOliver, PangaluAthina, GermansMenno, MuroiCarl, KellerEmanuela (2018), Decision tree analysis in subarachnoid hemorrhage - Prediction of outcome parameters during the course of aneurysmal subarachnoid hemorrhage using decision tree analysis., in J of Neurosurgery, 129(6), 1499-1510.
Not to cry wolf: distantly supervised multitask learning in critical care.
SchwabPatrick, KellerEmanuela, MuroiCarl, Mack David, SträssleChristian, KarlenWalter (2018), Not to cry wolf: distantly supervised multitask learning in critical care., in Proc 35th Int Conf Mach Learn (ICML), Microtome Publishing, Stockholm, Sweden.
Temperature variability in the day-night cycle is associated with further intracranial pressure during therapeutic hypothermia
Nogueira Adriano Berreto, Keller Emanuela (2017), Temperature variability in the day-night cycle is associated with further intracranial pressure during therapeutic hypothermia, in J Transl Med., 170-170.
Automated false alarm reduction in a real-life intensive care setting using motion detection.
MuroiCarl, MeierSilvio, De LucaValerai, MackDavid, SträssleChristian, SchwabPatrick, KarlenWalter, KellerEmanuela, Automated false alarm reduction in a real-life intensive care setting using motion detection., in Neurocritical Care.

Scientific events

Active participation

Title Type of contribution Title of article or contribution Date Place Persons involved
Jahrestagung der Deutschen und Österreichischen Gesellschaften für Epileptologie und der Schweizerischen Epilepsie-Liga Talk given at a conference Personalisierte Überwachung von Epilepsiepatienten in Videoaufzeichnungen. 08.05.2019 Basel, Switzerland Gabrani Maria;
Jahrestagung der Deutschen und Österreichischen Gesellschaften für Epileptologie und der Schweizerischen Epilepsie-Liga Poster Automatische Detektion des Burst-suppression-Musters bei Patienten auf der Neuro-Intensivstation 08.05.2019 Basel, Switzerland Narula Gagan;
18th Day of Clinical Research Poster Automatic detection of burst-suppression-pattern in neurocritical care patients. 11.04.2019 Zurich, Switzerland Narula Gagan;
Deephealth-Data Analytics, Machine Learning & Medical Decision Support Talk given at a conference 175. ICU Cockpit – gathering high resolution raw data for predictive medicine 25.10.2018 Zurich, Switzerland Keller Emanuela;
16th Annual Meeting Neurocritical Care Society Poster Motion detection based automated alarm classification in a real-life intensive care setting 24.09.2018 Boca Raton, Florida, United States of America Muroi Carl Izumi;
Annual Meeting Zentrum für Neurowissenschaften Zürich (ZNZ) Poster IT platform for multimodal patient monitoring and therapy support in intensive care and emergency medicine. 29.03.2017 Zurich, Switzerland Keller Emanuela; Muroi Carl Izumi;
13th Symposium of the Zurich Center of Integrative Human Physiology (ZIHP) Poster A simple visualization for inter-correlations in multi-modal signal environments. 29.03.2017 Zurich, Switzerland Keller Emanuela; Muroi Carl Izumi;


Knowledge transfer events

Active participation

Title Type of contribution Date Place Persons involved
Stiftung der Deutschen Wirtschaft Talk 23.03.2019 Constance, Germany Keller Emanuela;
Swiss eHealth Forum Talk 07.03.2019 Berne, Switzerland Keller Emanuela;
21. Novartisgesundheits-Forum Talk 14.11.2018 Berne, Switzerland Keller Emanuela;
Scientific Board Meeting Philips Health Systems Talk 10.07.2018 Böblingen, Germany Keller Emanuela;
Seminar Series Contemporary Debates in Bioethics Institute for Biomedical Ethics, University of Basel, Talk 29.03.2017 Basel, Switzerland Keller Emanuela;
- ICU Cockpit – gathering high resolution raw data for predictive medicine. Deephealth-Data Analytics, Machine Learning & Medical Decision Support, University Hospital Balgrist Talk 29.03.2017 Zurich, Switzerland Keller Emanuela;


Communication with the public

Communication Title Media Place Year
Talks/events/exhibitions Workshop ICU Cockpit at University Hospital Zurich International 2019
Talks/events/exhibitions Workshop ICU Cockpit at IBM Rüschlikon German-speaking Switzerland International 2018

Associated projects

Number Title Start Funding scheme
200568 Clinical Decision Support System for Neurocritical Care Tapping into High and Low Speed Data Lakes 01.06.2021 Project funding (Div. I-III)
187331 From Tools to Teammates: Human-AI Teaming Success Factors in High-risk Industries 01.09.2020 NRP 77 Digital Transformation

Abstract

In neurocritical care, patient monitoring is based not only on standard intensive care monitoring, but also on numerous data obtained from very complex pathophysiological changes in the human brain. The medical staff often cannot integrate the huge amount of clinical data continuously generated by different devices. The amount of available data per patient is enormous (petabyte), and the data are multimodal. Examples are recordings from electrocardiography, artificial ventilation, electroencephalography, hemodynamics, metabolism and video monitoring. The lack of data integration and usability is one of the major reasons why only a small part of the knowledge that physicians use in this field is evidence based. The growth of knowledge in medicine, as well as personalized medicine including genotype data is no longer manageable for physicians alone in daily clinical practice.Machine learning and artificial intelligence, which enable for semi-automatic analysis of these complex and big data, are fundamental tools. Started in 2014, ICU-Cockpit is a joint research project between the University, ETH and IBM Research Zurich, in order to create an integrated platform for patient monitoring and therapy support. A framework based on data analysis, machine learning and modeling will allow for therapy support resulting in a better outcome for patients. ICU Cockpit can trigger a fundamental change in safety culture and standard operating procedures in daily emergency and intensive care medicine as well as telemedicine, and opens up enormous potential for clinical studies and scientific evidence.
-