Medical artificial intelligence (Medical AI) is expected to produce great results in modern medicine. The strongest point of modern AI is a feature extraction technology that exceeds the recognition capacity of human in image, speech, motion capture, and big data. Deep machine learning brought about the great development of current AI since it has four unique features (multi task, multi modal, supervised/unsupervised/semi-supervised study modes, automatic acquisition of hierarchical features). It is unknown how to utilize medical AI in circulatory control medicine. There are several theories about deep machine learning (supervised, unsupervised, and semi-supervised), and each of which has its unique feature. In order to develop AI effectively in medicine, it is necessary to understand the basic theories and its advantage/disadvantages of deep learning methods.
Embolic stroke, one of the most critical complications of a left atrial (LA) myxoma, may result in the postponement of surgical removal of LA myxoma. A 49-year-old man was diagnosed with a small subarachnoid hemorrhage after he collapsed. Magnetic resonance imaging revealed a large cerebral infarction; therefore, echocardiography was performed to detect intracardiac embolic sources and a large LA tumor suggestive of an LA myxoma was found. Emergency surgical removal was successfully performed under cardiopulmonary bypass within the first 24 h after the onset of the embolic stroke. An early surgical removal may reduce the incidence of further embolic events that could otherwise occur while waiting for surgery.