Jarle Urdal

Automatic prediction of Therapeutic Activities During Newborn Resuscitation Combining Video and Signal Data


To increase our knowledge on resuscitation of newborns, it is crucial to objectively quantify what is currently being done in terms of therapeutic activities, such as ventilation and stimulation, and how they affect resuscitation outcomes. In the current study, activities during 1535 newborn resuscitations between 2013 and 2018 at Haydom Hospital are automatically quantified by estimating a timeline describing the start and stop of the activities.

The proposed approach is combining methods using both video and time series signal data recorded by NeoBeat during resuscitation. From video, the activity recognition is done by a 3D CNN method. For the signal data, feature extraction is performed on ECG and accelerometer signals and thereafter machine learning is done to perform detection of “stimulation” activity. The best results are achieved with all signals and video available, for the activity “stimulation” we get an AUC of 0.86, sensitivity of 82.32%, specificity of 82.23%, and precision of 57.59%.