08/10/2021 Vilnius, Lithuania
The synergy between data science and medical knowledge changes the traditional approach towards patients who need mechanical ventilation. Information technology (IT) specialists using the latest data processing technologies provide medical personnel with new possibilities for decisions making and patients’ monitoring. A Lithuanian based company Acrux Cyber Service with the help of critical care specialists from the Republic Vilnius University Hospital has developed a novel system where all the data from sensors of the ventilator are collected and analyzed using various algorithms in order to adapt ventilation to individual needs as much as possible and ensure maximum safety and quality of mechanical lung ventilation.
The new generation ventilators provide and display an enormous amount of real-time data for each breathing cycle. For a long time, all this information was used to a limited extent because the data collection, storage, and analysis was a complex and resource-intensive process. Devices for centralized management are quite expensive and have their own shortcomings in the data processing. Luckily, rapidly evolving data science using artificial intelligence technologies has found a way to process information and create an innovative system, where medical personnel are able to monitor the settings and parameters of each ventilated patient. This system uses machine learning algorithms to automatically detect errors, patient-ventilator asynchronies, and other discrepancies between the patient and the device and warns the hospital staff. Moreover, the system employs various predictive analytics methods to offer the most suitable lung ventilation technique for each patient.
We talked about this novel system with one of the innovators Dr. Saulius Vosylius, a professor of medicine at Vilnius University and Head of the Center of Anesthesiology and Critical Care Medicine in the Republic of Vilnius University Hospital. Dr. Saulius Vosylius has over 30 years of experience in intensive care with his special interests in mechanical ventilation.
“At the very beginning, the medical team decided which type and method of lung ventilation is the best for each patient. The initial parameters – volume or pressure, flow, frequency, oxygen fraction (FiO2), etc. are set by the physician. After setting up the device, a lot of data is generated by the ventilator itself. The problem is with data storage and processing because an enormous amount of data is provided during each inspiration and exhalation which is hard to process for medical staff. Of course, devices for centralized management of patient data exist but with a number of limitations.”
Here, a solution was proposed by IT specialists. They came up with an idea to collect the data directly from the devices and send it to the server via a safe wireless connection. There the data is filtered and analyzed using various data science methods, including machine learning algorithms. Either real-time or summarized data could be accessed from the connected personal device – tablet, mobile phone, or laptop. It is important to note, that the data is completely depersonalized. The application does not store any personal information, yet the attending physician is able to fill in data on the dynamics of the patient’s health.
Using this platform, the medical personnel can remotely monitor the parameters of ventilation of each patient in real-time or take a look at the summary report for certain period of time. The generated report shows how long the patient has been in the therapeutic range of ventilation, how often, and what type of asynchronies have occurred. Moreover, the system itself is able to detect discrepancies and trends in certain parameters which could be related to changing patient’s conditions. These features not only facilitate the work of medical specialists but also ensures maximum safety for patients.
As Dr. Saulius Vosylius said, the biggest challenge during the ventilation process is ventilation-induced lung injury (VILI) so the risk for the development of VILI should be minimized.
„If the airflow pressure is too high, the air is blown too quickly or there are discrepancies between patient respiratory forces and ventilator (so-called patient-ventilator asynchrony), the risk for VILI increases. The main goal of our project was to create a data analysis system which could detect early changes in ventilation parameters related to lung injury, find out the causes and suggest solutions to prevent VILI“, said the professor.
The created system uses machine learning algorithms to recognize all irregular or incorrect respiratory cycles (e.g., the patient tries to inhale and the machine does not respond). In addition, these are not only recognized but the probable causes are provided when known (e.g. leak or excessive activity of the patient). During volume and pressure control ventilation it is also possible to calculate mechanical power, related to the risk of VILI.
„An innovative artificial intelligence-based system, which is capable to process and analyze hundreds and thousands of different parameters of ventilated patients and provide the medical staff with the warnings and recommendations, is indispensable in intensive care units. Our system can warn staff about the changes in the ventilation process and make suggestions for ventilation methodology trying to adapt settings for individual needs. The early detection of deteriorating respiratory function enables us to make changes of ventilation settings for the best quality and the optimal treatment results“, said Dr. Saulius Vosylius.
The professor believes that the current system could also be used to reduce human errors when irregular or incorrect respiratory cycles remain unnoticed. This would also be very useful for the less experienced medical workers or junior staff because ventilation parameters can be monitored remotely by the expert and useful recommendations are provided by the system itself. The software does not take on the decision-making function, but it provides the data which helps to make the best decision.
„Mechanical lung ventilation is often seen as a method, but we should keep in mind that lung ventilation is a process that requires not only medical knowledge but also continuous monitoring and analysis of the ventilation parameters. We must assess the quality of the ventilation process in order to reduce hospital’s financial burden and save our human resources“, the professor noted.
To date, the current system is being tested with invasive mechanical lung ventilation, but it is expected to also adapt algorithms for non-invasive lung ventilation methods to optimize non-invasive lung ventilation capabilities in the very near future.
Dr. Jokūbas Drazdas, co-founder of Acrux Cyber Service is glad that the cooperation of data specialists and physicians has managed to create a unique and innovative system to facilitate the work of medical staff as well as maximize patients‘ safety.
The new system is being tested in the Republic Vilnius University Hospital, where data from lung ventilation devices are analyzed and processed continuously. The developers emphasize the importance to involve more medical centers for wider recognition of the software. Acrux Cyber Service has started the process of certification of an innovative system in the EU, which is expected to expand the possibilities of using the system in Europe.