Researchers at the Center for Scalable Data Analytics and Artificial Intelligence (ScaDS.AI) are developing an AI model for stroke care intended to help assess which patients are suitable candidates for thrombectomy on a more individualized basis, according to a presentation by doctoral candidate Marie-Sophie von Braun at the “Brains on Silicon” AI conference in Dresden. The project was developed with the Department of Neurology at Leipzig University Hospital. When deciding on thrombectomy, CT perfusion scans are evaluated using specific threshold values to indicate which brain tissue has suffered irreversible damage and which may still be salvageable. The model processes multiple CT scans plus clinical and demographic information using a Convolutional Neural Network (CNN) designed to recognize patterns based on existing patient data. In an independent test group of 101 patients, the median Dice score for the previous method was 0.27, while the CNN achieved a median of 0.51; a value of 1 indicates a perfect match. Von Braun is now addressing how reliable predictions are for individual patients, and the model should later indicate uncertainty, with similarity to training data a key factor. “The AI cannot make the decision,” she emphasized.