Deep neural networks are used to automatically detect aortic valve events from cardiac signals from an epicardially placed accelerometer. PROJECT TITLE : Automatic Detection of Aortic Valve Events Using Deep Neural Networks on Cardiac Signals From Epicardially Placed Accelerometer ABSTRACT: The use of miniature accelerometers that are incorporated into pacing leads and then attached to the myocardium in order to monitor cardiac function is provided as background information. The acceleration signal needs to have its functional indices extracted in order to accomplish this goal. For this kind of extraction, it will be helpful to have a method that can automatically detect the time of aortic valve opening (also known as AVO) and aortic valve closure (also known as AVC). We tested whether or not Deep Learning could be used to detect these valve events from epicardially attached accelerometers. To establish ground truth for these valve events, we used high-fidelity pressure measurements. Method: A deep neural network consisting of a CNN, an RNN, and a multi-head attention module was trained and tested on 130 recordings from 19 canines and 159 recordings from 27 porcines covering a variety of interventions. These recordings were obtained through training and testing on 130 recordings from 19 canines and 27 porcines. Because there were insufficient data, nested cross-validation was utilized in order to determine how accurate the method was. When defining a correct detection as a prediction that is closer than 40 ms to the ground truth, the results showed that the correct detection rates for AVO and AVC in canines were 98.9% and 97.1% respectively, while the correct detection rates for porcines were 98.2% and 96.7%. The incorrect detection rates for AVO and AVC in canines were 0.7% and 2.3% respectively, while the rates for porcines were 1.1% and 2.3% respectively. In canines, the mean absolute error between correct detections and their respective ground truth was 8.4 milliseconds for AVO and AVC, while in porcines, the mean absolute error was 8.9 milliseconds for AVO and 10.1 milliseconds for AVC. For reliable and accurate detection of the opening and closing of the aortic valve, deep neural networks can be trained on the signals from accelerometers that are attached epicardially to the heart. Did you like this research project? To get this research project Guidelines, Training and Code... Click Here facebook twitter google+ linkedin stumble pinterest For crack detection, BARNet stands for Boundary Aware Refinement Network. Deep learning is used to automatically determine the severity of Pectus Excavatum from CT images.