Deep learning models trained to read and analyze CT scans of the brain can help predict patient reactions to intravenous thrombolysis treatments, according to a study published in Academic Radiology on April 30.
Researchers at the Royal Adelaide Hospital in Australia created two neural network models. When used together to read CT brain scans, the models predicted whether thrombolytic therapy would be a safe and effective treatment with 71 to 74 percent accuracy.
Thrombolysis is typically used to dissolve blood clots in arteries to the heart, brain and lungs, which can cause heart attacks, ischemic strokes and acute pulmonary embolism. Despite its many uses, the therapy poses serious risks to patients, including the possibility of intracranial hemorrhage or gastrointestinal bleeding.
More articles about AI:
Machine learning used to detect tuberculosis strains' drug resistance
AI-powered kidney disease diagnostic tool receives FDA 'Breakthrough Device' designation
Novartis, Pfizer, Centene CEOs on AI's limitations in healthcare