Spatio-temporal visual attention modelling of standard biometry plane-finding navigation.

•Modelling a sonographer’s visual attention transitions on fetal ultrasound (US) video frames using deep neural networks.•Used bidirectional Convolutional long-short term memory (biCLSTM) to model spatio-temporal information in US videos.•Use soft Dynamic Time Warping (sDTW) as a novel loss function to regularise temporal allocation of predicted visual attention.•Standard biometry planes detection informed by predicted visual attention maps.