A new case study from NVIDIA examines how the company’s Quadro GPUs are being used to power the first virtual 3D heart, a model so realistic it can be viewed from any angle in photorealistic detail.

The challenge started in 2006, when three clinicians at University College London Hospitals (UCLH) Heart Hospital, Drs. Sue Wright, Andrew Smith, and Bruce Martin, saw the need for a realistic model of the human heart — one that could be used to teach cardiac anatomy. The doctors reasoned that if a virtual model were based on an anatomical 3D data set, it could be used to create a simulated ultrasound image. They specifically sought a solution that would allow trainees to simulate a delicate examination technique called transesophageal echocardiography (TOE or TEE). In this procedure, an alternative method of performing an echocardiogram, the practicioner inserts a specialized probe down the patient’s esophagus to capture images of the beating heart.

In search of partners who could bring their design to life, the clinicians met with the creative team at the UK-based Glassworks, a visual effects company known for taking chances. They doctors laid out their plan for “an anatomically accurate, computer-generated heart model that could instantly produce an authentic ultrasound representation, be animated to beat in real-time to show changes in the heart’s shape during the cardiac cycle, and enable the doctor to view slices of the scene to gain valuable diagnostic insights.”

The Glassworks team signed on and project HeartWorks was born. The design artists used Autodesk Softimage 3D animation software running on workstations equipped with NVIDIA Quadro graphics processing units (GPUs) to create a realistic heart model with real-time animations.

Hector McLeod, Glassworks founder, explains that “immersive technology, where users can interact with 3D images on a screen, has been limited historically by the state of graphic technology, and by budgets, [but] NVIDIA’s … massively parallel GPU technology has changed that.”

McLeod elaborates: “What the NVIDIA GPU does within HeartWorks is load and display extremely complicated models, and render them at 30 frames a second, in a series of events that are themselves, complex. In any given frame we’re tapping the GPU to examine the model, slice it, annotate the views and display the two visualizations — the model and the ultrasound — in less than 1/30 of a second.”

HeartWorks has gone from conception to design and finally implementation. It is now being used as a teaching tool at Duke University in the Department of Anesthesiology, Division of Cardiothoracic Anesthesia and Critical Care Medicine. Since early 2009, the department has been using the interactive device to teach residents and fellows in their advanced Transesophagael Echocardiography program. Dr. Madhav Swaminathan of the Duke University School of Medicine’s Division of Cardiothoracic Anesthesia credits the simulation technology for effecting a “quantum leap” in their training program.