GOAL: Improving the ML model for sepsis prediction in preterm newborns

During this challenge organized by FruitPunch AI for Health and UMC Utrecht you will be working on improving the existing XGBoost prediction model with more features, exploring advanced models and developing a new model by taking time-series data into account. 

UMC Utrecht’s Neonatology department is currently using an XGBoost model to predict sepsis in newborns. Only three variables have been taken into account in the current model: heart frequency, respiration, oxygen saturation. The current model is analyzed using SHAP to make it more explainable, and doctors receive risk scores of the babies every hour which are effectively the model’s confidence scores for classification.

The goal of this project is to diagnose sepsis earlier, such that doctors can get notified and take action, reducing mortality in preterm newborns.

Challenge Info Session  

Who are we looking for? 

Anyone with an interest in artificial intelligence or machine learning can apply. We expect some experience with programming languages and an interest in machine learning and AI. Some of the techniques you could be using and should spark your interest are, own experience and initiative with prediction is appreciated as well:

• Medical Analytics
• Time series
• Classification
• Explainable AI

You can join as a contributor (12 hours per week commitment for 2 months), coach (2-4 hours per week, only for experienced ML professionals) and teacher (to give one relevant ML / domain masterclass).

During the challenge we will organize masterclasses on relevant topics like the use of the cloud computation resources Microsoft, IBM & the HPC Lab have made available for us.

Did you know

• 1 in 10 infants born in the United States is born preterm
• Around 15% of preterm infants will develop a late onset sepsis
• The mortality rate in neonatal sepsis was found to be 20%