Skip to main content
Fig. 4 | BioMedical Engineering OnLine

Fig. 4

From: 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 patients

Fig. 4

CAM of a correctly predicted HT case, as shown in C. The patient was admitted to our hospital with sudden inactivity of the left limb for 2 h, and haemorrhage and clear infarct lesions were not found on baseline NCCT (A). Then, the patient was given rt-PA 56 mg. The cranial NCCT was re-examined within 24 h, and there were HTs in the left thalamus and midbrain cerebral peduncle (D). Superimposing the heatmap on the native image (B) highlights the left thalamus and midbrain cerebral peduncle, which were the regions of HT that occurred after IVT (as shown by the arrow), thus proving that the model predicting upcoming HT was favourable

Back to article page