How can we tell at a relatively early stage whether a COVID 19 patient is likely to develop a severe course? With this knowledge, clinics could better respond to severe waves of disease. Data scientists tried to find an answer to this question during the HIDA International Virtual COVID-Data Challenge.
Overcrowded emergency rooms, overworked hospital staff, inadequate intensive care - many people still clearly remember the news that reached us during the first Corona wave from hospitals in northern Italy, for example. Knowing who among the admitted patients most urgently needs the limited treatment resources would greatly facilitate the work of physicians in such a stressful situation. But how would it be possible to tell at a relatively early stage after hospital admission whether a Covid-19 patient might be developing a severe course?
The International Virtual COVID-Data Challenge organized by HIDA posed this question to an international audience of young data scientists: 63 junior scientists from 15 countries competed to improve an algorithm that uses X-ray images to predict the likely course of disease for Covid patients. A challenging team effort awaited the participants, who collaborated on solutions across five time zones from April 28-29, 2021.
Mild or severe disease progression?
The tasks were based on data collected at six different hospitals in northern Italy during the first outbreak of COVID-19 in March 2020. The disease course of the patients from whom the data were collected has been retrospectively classified as either severe (when the patient required mechanical ventilation or died) or mild (all other courses).
At the time when the fate of the admitted patients was still open, namely during triage, a number of clinical tests were performed on the patients in addition to radiographs of the lungs, which generated several parameters. Of these, 16 were considered relevant to subsequent disease progression and were included in the challenge data set.
For the data scientists tackling this cutting-edge research question, the challenge was to solve two tasks, which were independently judged by the panel. While the first step was to complete missing clinical data from these test series using statistical methods, the second task was to classify the subjects into "mild" and "severe" courses according to the severity of their disease based on this data. - This is an important finding that could help hospitals in the future to treat patients appropriately and with consideration for existing hospital capacities in such special overload situations as the one in Italy in spring 2020.
International collaboration - both before and behind the computers
Two winning teams, kitmetrics and friends (as winners of the first sub-task) and nem-team (as winners of the second sub-task), were able to convince the jury and thus contribute to the improved applicability of the algorithm in clinical research. Which way the two teams chose for their solutions can be seen again here. We congratulate the winning teams and all participants who faced this challenging task!
Get out of pandemic isolation and get to the computers: that was the motto for the participants of the H^3 Helmholtz Autumn Hackathon in September 2021, who worked in direct exchange with each other again for the first time.
The state of Baden-Württemberg is a major funder of HIDSS4Health and is providing over 1,2 million euros of additional funding for the school for 2021 to 2024. But what was it about the HIDSS4Health concept that impressed policymakers? We spoke with Dr. Simone Schwanitz, Senior Ministry Official at the Baden-Württemberg Ministry of Science, to learn more.
An excursion into permafrost research: For three months, Ngai Ham Chan from GFZ visited the PermaRisk research group at AWI with the HIDA Trainee Network. What did he learn there about climate change and data science methods?