Brain
|
Multicenter study of deep learning-based prediction of MGMT gene promoter methylation status from digitized H&E stained whole-slide-image data in IDH wild-type glioblastoma
Funding line:
Else Kröner Memorial Fellowships
Using Grad-CAM algorithm, an activation map is displayed within a deep learning network. The labeled areas play a crucial role in predicting the molecular outcome from histological whole-slide-image data.
(© University Hospital Essen, Sied Kebir)
In recent years, we have learned that histological whole-slide-image data contain much more information than has previously been possible to determine using conventional methods. The Else Kröner Memorial Fellowship will support the implementation of a multicenter, pan-European project with the goal of developing a deep learning model that will use artificial intelligence to enable fully automated prediction of MGMT gene promoter methylation status from digitized H&E stained whole-slide-image data in glioblastoma patients. This tool can help to more rapidly and efficiently plan downstream therapy for glioblastoma patients that is relevantly dependent on MGMT gene promoter methylation status.
