Computer Vision News - May 2020

2 Summary RISP Vision Projects 10 Currently about 30 Million Americans are diagnosed with Cardiovascular diseases , which are also responsible for up to a third of all deaths in the US. Cardiac ultrasound, or echocardiography , plays a crucial role in the diagnosis and management of cardiovascular diseases. Non-invasive, safe, low cost, and providing real-time video sequences of the functioning heart, echocardiography is used throughout all stages of patient care: from diagnosis and periodic check-ups in cardiology clinics and hospital wards, to indispensable aids used during surgery and interventional procedures . Two-dimensional trans thoracic echocardiography (TTE), together with Doppler analysis, account for the vast majority of echocardiography studies performed. By placing the ultrasound probe at precise and specific locations and orientations on the chest, various 2D cross-sections (“views”) of the heart can be obtained through “acoustic windows” which allow the ultrasound waves to pass without being obstructed by bone. By measuring the Doppler frequency shift of the back-reflected ultrasound waves, the blood flow velocity at various points in the heart can also be analyzed. A typical Echocardiography study can produce tens of ultrasound video clips of different views and with different types of Doppler analysis, providing a wealth of information to aid diagnosis and monitoring of many cardiovascular conditions. However, obtaining precise high-quality ultrasound clips can be challenging, and requires taking into account height, body mass index, lung health, body position, and more. Even if the required view is successfully recorded, these factors can affect the quality of the resulting video clip. Considering also variations in equipment and/or operator experience, there can be very large differences in quality between one study and another, and even within the same study. Furthermore, obtaining the Echocardiography study is only one half of the task. The study then needs to be assessed by skilled medical professionals to identify and classify pathologies and determine treatment paths. This assessment phase can be time consuming, prone to error, and produce subjective results. Thus, it is not surprising that much R&D effort is being invested in developing Using Deep Learning to Aid Echocardiography