During the Women in Digital Health Event last Thursday, February 6th, 2020, Prof. Dr. Emanuela Keller provided some insights into the use of Artificial Intelligence (AI) algorithms in the neuro ICU of the University Hospital Zurich.
The Health-Trends team was on-site and summarized the most important impressions from this exciting presentation in a short report. Enjoy reading it.
A world-leading ICU department struggles with typical problems in the healthcare sector
The ICU department (Intensive Care Unit) of the University Hospital Zurich is a worldwide leading institution of its kind and can be compared to other similar departments in major leading hospitals around the world. Although the infrastructure of the ICU department is top-notch, Prof. Dr. Emanuela Keller made clear in her presentation that the ICU of the University Hospital Zurich is also struggling with some problems typical of today’s healthcare system:
- Multimorbidity of patients: Patients over 50 years of age suffer on average from two or more diseases. This multimorbidity increases the complexity of treatment in the ICU. It poses a particular challenge for physicians, as they often treat many patients (usually more than 10 per physician) and must set priorities accordingly. Keller compared this situation with a “House of Cards”: If a patient is treated incorrectly at a given time, massive complications immediately arise and there is too little treatment time left for the other patients. The house of cards collapses as a result.
- Complex patient situations: ICU patients are continuously monitored with many different technical devices due to their critical health conditions. These devices cause an average of up to 700 real-time alarms per day and per patient. However, many of these are false alarms triggered by, for example, movements of the patient, changed medication, etc. The above-mentioned multimorbidity and the individual health characteristics of the patient create complex patient situations, some of which require very individual treatment.
- Flood of medical data: In 2019, a total of 419 patients were treated in the ICU department of the University Hospital Zurich. Each of these ICU patients generates up to 40 GB of data from medical devices, biomedical sensors, etc. in and on the patient’s body in one day. These data must be continuously evaluated and interpreted in order to continue the treatment of the patient. The flood of medical data, combined with the already mentioned complex patient situations, leads to a massive challenge for the medical staff. Especially in an emergency situation, it is not possible to quickly integrate this flood of information into the decision-making process
- Poorly connected technical infrastructure: Not surprisingly, Keller also pointed out the poorly connected technical infrastructure and the lagging e-health discussion (especially the poor usability of electronic patient file systems). The above-mentioned technical equipment from different providers is unable to communicate with each other, or only to a limited extent. In addition, these devices generate data outputs that are not standardized. In combination with the enormous amount of data that every ICU patient produces daily, this leads to an almost unsolvable task for ICU staff.
- Academic information overload: Last but not least, Keller also pointed out that the publication of academic research results on the subject of precision medicine (as in other medical research fields) has increased exponentially in recent years. At the same time, the half-life of medical knowledge is constantly being reduced. This leads to a flood of information that can no longer be handled by today’s medical personnel.
As a summary of these challenges, there is a strong need for a data-based intelligent dashboard that consolidates and interprets the data in one place in an understandable way. In particular, it should be possible to take decisions quickly on the basis of current treatment situations (e.g. alarms due to a changing patient situation). Therefore, the ICU has created an “ICU Cockpit” with the help of Supercomputing Systems.
ICU Cockpit – Great Ambitions for Precision Medicine at Zurich University Hospital
The ICU-Cockpit is a joint IT research project of the University of Zurich, ETH Zurich, IBM Research and Supercomputing Systems with the goal of creating an integrated platform for patient monitoring and therapy support. Data from numerous medical devices are synchronized and algorithms for early warning systems and therapy support are developed, with the initial focus on multimodal neuromonitoring and the prevention of secondary brain injuries.
The aim of the project is to initiate a fundamental development in emergency and intensive care medicine and to achieve a significant improvement in the way diagnostics, treatment and risk management are handled in everyday clinical practice. Data from all patients will be collected at all times and continuously analyzed. In particular, the treatment should be continuously adapted on the basis of these data and predictions about the possible state of health and course of treatment should be made. Keller emphasized that the “ICU Cockpit” does not aim to replace ICU physicians but to provide them with the right tools to make the right decisions in complex patient situations in a timely manner.
Since each patient situation is unique and can only be compared with other patients to a very limited extent, a separate AI algorithm is trained for each patient. This personal algorithm evaluates the data and interprets it with regard to the current and future course of treatment. Keller even pointed out that the algorithms cannot be transferred to other patients. Such individual algorithms are particularly helpful in coping with the approximately 700 alarms per day and patient. In a first step, the alarms are classified (e.g. has there been a recent change in medication, has the patient moved specifically, etc.). If an alarm is identified as a real one, further analyses of the expected change in health status are carried out on the basis of this. For this purpose, logical analyses based on the various parameters (e.g. blood pressure, pulse, etc.) are executed and accordingly transferred to a color-supported correlation matrix. These colorings help the medical staff in making treatment decisions.
Prof. Dr. Emanuela Keller has pointed out that the ICU cockpit is a closed circle consisting of data collection, development of algorithms, validation, monitoring and adaptation of the treatment to the patient. Partners such as the University of Zurich, ETH, Supercomputing Systems or IBM have helped with the setup of the infrastructure or the development of algorithms. However, the confidentiality and security of patient data is a very high priority, which is why such a closed cycle was chosen. At the same time, Keller pointed out that in Switzerland, compared to other countries, there are fewer restrictions in terms of data law and that this is a great opportunity to further advance data-based research for ICU patient treatment.
Our conclusion of the evening
The lecture by Prof. Dr. Emanuela Keller took place in the context of a Women in Digital Health event at the headquarters of Supercomputing Systems in Zurich’s Technopark. It was a presentation in front of a small group of listeners, whereby the 20 or so attendees were seated comfortably at a large conference table. In our opinion, this fact contributed significantly to a good and open discussion at the end of the lecture and, together with Prof. Dr. Emanuela Keller, some details of this exciting topic could be highlighted once again.
From our point of view, the subjects of the lecture were also very exciting, as it described the topics of artificial intelligence and precision medicine in a very practice-oriented way and showed the possibilities of this technology that exist today in practical manner. Keller was able to give a very vivid overall picture of the challenges within the ICU of the University Hospital of Zurich and the associated possibilities of new technologies in healthcare. All in all, it was therefore a very successful and exciting evening. Many thanks at this point to the initiators of Women in Digital Health for the organization of this event.
About Women in Digital Health
Women in Digital Health is an initiative founded by Aisha Schnellmann, Daniela Gunz, Gabriela Hofmann and Sunjoy Mathieu, which is intended to provide a platform for exchange, especially for women with a connection to digital health. Regular events on various topics take place. The four initiators emphasize that men are also cordially invited at any time. The next event will take place on 4 March 2020 in the Zurich area (ETH Rocket Hub), where the ceremony for the founding of the association “Women in Digital Health” will also be held. You can find more information on this at https://womendigitalhealth.net/
Some impressions of the evening: