Fig. 7From: A weakly supervised deep learning model integrating noncontrasted computed tomography images and clinical factors facilitates haemorrhagic transformation prediction after intravenous thrombolysis in acute ischaemic stroke patientsIllustration of our WSDL framework. Multi-instance learning and attention mechanisms were adopted to construct the model. To increase the representation information of the input image, we use the multiwindow transfer module to integrate the image information with three window widths and window levels in the channel dimension. In addition, we proposed a novel loss, i.e., AS loss, which was used during model training to ensure the classification performanceBack to article page