Automatic bubble detection with neural networks on post decompression frames

Post decompression records in echocardiography are considered to detect micro bubbles and to survey unexplained decompression sickness which is commonly examined by standardized methods such as dive computers and tables. In this study, existent bubbles are detected on transthoracic echicardiografic frames recorded after recreational diving. Bubble detection is performed by Artificial Neural Networks which are trained using bubbles with different morphologies. We showed that bubbles would be detected on four cardiac chambers without image segmentation.