Qvik was involved in scientific research that created machine learning based algorithms to predict mortality during intensive care after traumatic brain injury. Though the project is still a proof-of-concept, the algorithms can predict the probability of the patient dying within 30 days with an accuracy of 80–85%.
Traumatic brain injury (TBI) is a common cause of mortality and morbidity around the world. Even with careful treatment in an intensive care unit (ICU), one in three patients with severe TBI dies.
Patients that suffer from severe TBI are unconscious, and they are monitored carefully. Only one variable, such as intracranial pressure, may yield hundreds of thousands of data points per day, so the vast amount of data is impossible for the human brain to analyze. This is why predicting patients’ medical condition with the help of machine learning algorithms comes handy.
“Understanding the data required a lot of domain specific knowledge, and during the project we got to work with a solid team of medical professionals,” says our software engineer Olli Paakkunainen. “On a more personal note, this was an interesting project since this was the first time I got to do work that was actually related to real life-and-death situations.”
Gathering the data and creating the algorithms is a collaborative project between three Finnish university hospitals: Helsinki University Hospital, Kuopio University Hospital and Turku University Hospital. The study was published in Scientific Reports. Qvik’s software engineer Olli Paakkunainen and data scientist Mikko Kemppainen helped conduct the statistical analyses. The analyze was done using Google Cloud Platform’s tools such as BigQuery, Cloud Storage and Datalab.
Helsinki University: Artificial intelligence-based algorithm for intensive care of severe traumatic brain injury
Scientific Report on Nature: Machine learning-based dynamic mortality prediction after traumatic brain injury
STT info: Suomalaistutkijat kehittivät tekoälyalgoritmin aivovammapotilaiden tehohoitoon