Smart Heart: GPU accelerated deep learning to find better solution for Coronary artery disease (CAD)
“The greatest truth is this: if the health of the individual is not cared for then all the big ideas die when the thinker becomes too ill or tired or burned-out to carry them out. The big thinkers are going to need health on their side to get it done.” Says Ann Garvin an author, speaker, and professor of health, stress management, research methods and media literacy.
Figure 1: HeartFlow (source)
Health is most important for all of us and in a recent report released by World Health Organization (WHO) it is stated that Coronary artery disease (CAD) is the world’s biggest killer, responsible for nearly 9 million deaths worldwide and diagnosed in 12 million to 13 million Americans each year. This is because it often goes undetected or it is misdiagnosed especially in women. Until the invention of angiogram which is an invasive and costly procedure, heart disease used to be difficult for doctors to diagnose.
We have Good News! Thanks to HeartFlow , a personalized medical technology company, dedicated to managing CAD. HeartFlow FFRCT analysis product is the non-invasive technology for CAD that offers insight and both the extent of coronary artery narrowing, and whether it is impacting blood flow.
The use of the HeartFlow FFRCT Analysis improves the diagnosis and care of patients with suspected CAD compared with usual care. The HeartFlow Analysis provides calculated Fractional Flow Reserve (FFR) information noninvasively using advanced analysis of cardiac CT angiograms (cCTA), which we denote as “FFRCT.” The workflow for the HeartFlow Analysis involves uploading cCTA DICOM image data to HeartFlow, data processing overseen by qualified HeartFlow Analysts using the FFRCT medical device software, and the return of the HeartFlow Analysis Results to the physician via a secure web interface. The HeartFlow Analysis provides FFRCT values throughout the coronary arteries, which physicians can easily integrate into diagnostic decision-making.
Figure 2: HearFlow Solution(Source)
Functional Understanding of FFRCT Process:
HeartFlow utilizes data from standard coronary CT angiography (cCTA) scans to create a personalized 3D model of the coronary arteries of the patient’s heart, then by applying complex fluid dynamics (CFD) and deep learning algorithms, to model the complexity of fluid flow.
The end result of heart analysis is a color coded map of the coronary arteries showing the extent to which any blockages impairing blood flow, the physician can use this information to develop a treatment plan that is right for that patient. To compute CFD HeartFlow making use of Nvidia GPU platform.
Figure: HearFlow Solution(source)
Accelerating CFD and deep leering with Nvidia GPU:
Building personalized models of the heart is very challenging and complex task. Besides creating a subvoxel-accurate model for each patient, HeartFlow’s system must simulate the flow of blood, vessel by vessel. Diagnosis is also time-sensitive in fast-paced emergency care departments. In order to compute complex fluid dynamics (CFD) and deep learning algorithms, Nvidia accelerated GPU platform is the ideal processor to achieve this accuracy.
As Nature recently noted, early progress in deep learning was “made possible by the advent of fast graphics processing units (GPUs) that were convenient to program and allowed researchers to train networks 10 or 20 times faster.”
Figure 4: Nvidia GPU cloud(source )
Other healthcare companies who are using GPU based deep-learning
- Deep Genomics is applying GPU-based deep learning to understand how genetic variations can lead to disease.
- Arterys uses GPU-powered deep learning to speed analysis of medical images. Its technology will be deployed in GE Healthcare MRI machines to help diagnose heart disease.
- Enlitic is using deep learning to analyze medical images to identify tumors, nearly invisible fractures, and other medical conditions.
To sum up, deep-learning breakthroughs have sparked the AI revolution. Machines powered by AI deep neural networks solve problems too complex for human coders. Adopting this in healthcare will definitely have an impact on society in a positive way, this helps physicians make definitive and personalized decisions more quickly, that means a better outcome for patients.
However, High-income countries have systems in place and they are the first one to get the most advanced and new breakthrough technologies. Many low- and middle-income countries do not have such systems, and they are lagging in availing the benefits of new breakthrough technologies, and hence governments and organization should work collaboratively to overcome such obstacles and make it available all over the world.