Predicting the Trajectory of ICP in TBI Patients: Evaluation of a Foundation Model for Time Series
🌟🎙️ This November, we had the pleasure of hosting the talented Florian van Leeuwen for the thirteenth installment of our webinar series, Case Studies in Neurocritical Care AI!
Florian van Leeuwen is a PhD candidate in Statistics at the University of Utrecht. His research focuses on prediction models for tabular and time series data, particularly in cases with sparse information, missing data, or when there is a need to quantify uncertainty. He is part of the Missing Data (MICE) and Bayesian Statistics research groups. Previously, he worked at Leiden University Medical Center under the supervision of Ewout Steyerberg and Erik van Zwet. During this time, the focus was on projects related to Traumatic Brain Injury, which led him to visit Moberg Analytics as a visiting researcher this past summer. 🧠✨
About the talk:
‼️ Patients with traumatic brain injury (TBI) are at risk of dangerous spikes in intracranial pressure (ICP), leading to the life-threatening condition of intracranial hypertension (tIH). Detecting these changes early could be crucial for improving outcomes through timely clinical interventions. But how can we predict these sudden rises in ICP with limited patient data? In this talk, Florian explores whether pre-trained (foundation) models, leveraging the power of transfer learning, hold the key. He dives into a comparison of three advanced time series models, each predicting the next 30 minutes of ICP based on the previous 60 minutes. He further discusses general challenges with clinical prediction models and gives a short tutorial on how a pre-trained model can be used.
If you're interested in joining the webinar series, please register via this link: https://moberganalytics.us20.list-manage.com/track/click?u=7b9c221df909d527e19e04dd2&id=4c20d3100d&e=483b56d895
You can view recordings of past webinars here: https://www.youtube.com/playlist?list=PLkYOwrQDnbZEZ8EHlJ-bE6Do_EAWm8JDP
This webinar series is designed to help clinicians learn how to get more from their neurocritical care data. No prior experience is necessary.
Each session focuses on a real-world case study from us or from your colleagues. We walk participants through the steps to re-create the analytics that answers the neurocritical care question presented. We cover tools, data handling, feature engineering, cohort selection, model training, model tuning, visualization, and other topics. Sample code, Jupyter notebook, and data are provided to enable you to experiment and apply these methods to your own data sets and problems.
Please feel free to let us know if you have any questions about the concepts that were presented in this video or if you have case studies or suggestions for upcoming sessions.
Website: https://moberganalytics.com
LinkedIn: https://www.linkedin.com/company/moberg-analytics/
Twitter: https://x.com/moberganalytics?lang=en
Disclaimer:
The views and opinions expressed by the presenter and other third parties do not necessarily reflect those of Moberg Analytics, Inc. Moberg Analytics, Inc. makes no clinical claims regarding information described by the presenter and other third parties.
🌟🎙️ This November, we had the pleasure of hosting the talented Florian van Leeuwen for the thirteenth installment of our webinar series, Case Studies in Neurocritical Care AI!
Florian van Leeuwen is a PhD candidate in Statistics at the University of Utrecht. His research focuses on prediction models for tabular and time series data, particularly in cases with sparse information, missing data, or when there is a need to quantify uncertainty. He is part of the Missing Data (MICE) and Bayesian Statistics research groups. Previously, he worked at Leiden University Medical Center under the supervision of Ewout Steyerberg and Erik van Zwet. During this time, the focus was on projects related to Traumatic Brain Injury, which led him to visit Moberg Analytics as a visiting researcher this past summer. 🧠✨
About the talk:
‼️ Patients with traumatic brain injury (TBI) are at risk of dangerous spikes in intracranial pressure (ICP), leading to the life-threatening condition of intracranial hypertension (tIH). Detecting these changes early could be crucial for improving outcomes through timely clinical interventions. But how can we predict these sudden rises in ICP with limited patient data? In this talk, Florian explores whether pre-trained (foundation) models, leveraging the power of transfer learning, hold the key. He dives into a comparison of three advanced time series models, each predicting the next 30 minutes of ICP based on the previous 60 minutes. He further discusses general challenges with clinical prediction models and gives a short tutorial on how a pre-trained model can be used.
If you’re interested in joining the webinar series, please register via this link: https://moberganalytics.us20.list-manage.com/track/click?u=7b9c221df909d527e19e04dd2&id=4c20d3100d&e=483b56d895
You can view recordings of past webinars here: https://www.youtube.com/playlist?list=PLkYOwrQDnbZEZ8EHlJ-bE6Do_EAWm8JDP
This webinar series is designed to help clinicians learn how to get more from their neurocritical care data. No prior experience is necessary.
Each session focuses on a real-world case study from us or from your colleagues. We walk participants through the steps to re-create the analytics that answers the neurocritical care question presented. We cover tools, data handling, feature engineering, cohort selection, model training, model tuning, visualization, and other topics. Sample code, Jupyter notebook, and data are provided to enable you to experiment and apply these methods to your own data sets and problems.
Please feel free to let us know if you have any questions about the concepts that were presented in this video or if you have case studies or suggestions for upcoming sessions.
Website: https://moberganalytics.com
LinkedIn: https://www.linkedin.com/company/moberg-analytics/
Twitter: https://x.com/moberganalytics?lang=en
Disclaimer:
The views and opinions expressed by the presenter and other third parties do not necessarily reflect those of Moberg Analytics, Inc. Moberg Analytics, Inc. makes no clinical claims regarding information described by the presenter and other third parties.
){kind=link}
Comments