Background Head CTs with perfusion imaging have become crucial in the selection of patients for treatment for mechanical thrombectomy. In recent years machine learning has rapidly evolved and found applications in a wide variety of health care tasks. We report our experience with training a neural network to predict the presence and sidedness of a perfusion deficit in patients with acute ischemic stroke.

Methods Dynamic multiphase 4-dimensional contrasted head CTs of patients with suspicion for acute ischemic stroke were obtained. An LRCN network was constructed consisting of a convolutional neural network stacked on top of an LSTM layer.

Results 139 (35.1%) patients had a right sided perfusion deficit, while 199 (50.3%) had a left sided deficit, and 58 (14.6%) had no evidence of a deficit. The best model was able to achieve a loss of 0.88 and an accuracy of 85% on validation data. ROC curves were generated for right sided perfusion deficit, left sided perfusion deficit, and no perfusion deficit and an AUC was calculated for each class. For a right sided deficit, the AUC was 0.92, for left sided deficit 0.94, and for no deficit the AUC was 0.95.

Conclusion The field of machine learning, powered by convolutional neural networks for the task of image recognition and processing, has quickly developed in recent years. We have constructed an artificial neural network that can identify and classify the presence of a perfusion deficit and the sidedness on CT perfusion imaging.

Disclosures J. Vargas: 1; C; NVIDIA. 4; C; NVIDIA. A. Spiotta: None. A. Chatterjee: 2; C; Viz.