Detecting noise and artifact in CW ultrasound signal processing using machine learning and cloud-based tools

Continuous wave (CW) ultrasound systems are extremely sensitive to movement, noise, and artifacts of reflective tissues within the body that return doppler ultrasound signals to the receiver. In the application of CW ultrasound to clinical applications, classifying and handling noise/artifacts is essential for broad clinical adoption. A machine learning (ML) algorithm is commonly used for pattern recognition of large sets of data, such as physiological signals, and it has been used recently for biomedical applications. Moreover, cloud computing allows the execution of large and complex calculations without the need for expensive or dedicated hardware. The aim of this project is to develop cloud-based computing tools in the Python programming language that will annotate, classify, and label physiological signals. These annotations and labeled signals will then inform and feed digital and ML signal processing methods. Furthermore, process automation for new incoming data and quality assurance checks for the remotely acquired signals will also be developed as cloud-based tool.

Faculty Supervisor:

Jeremy Brown

Student:

Alejandro Ivan Villalba Euan

Partner:

Flosonics Medical

Discipline:

Engineering - computer / electrical

Sector:

Manufacturing

University:

Dalhousie University