Digital heart image

Be still, my beeping heart: biomedical digital twins

Image credit: Dreamstime

Digital simulations of human anatomy and physiology are becoming increasingly sophisticated, with these models set to enhance teaching, clinical trials and medical research. How far can they go?

We have been recreating human anatomy for tens of thousands of years: from Palaeolithic carvings of exaggeratedly fertile women to the disturbingly sensual ‘Anatomical Venus’ sculpted from wax in the 18th century. Today, scientists are building increasingly complex models of the human body, moving from wax and marble to working in the digital realm.

Frank Sculli is CEO and co-founder of BioDigital, a New York-based company that provides detailed, cloud-based 3D anatomical simulations that are interactive, come in male and female form, and can mimic the symptoms of all common diseases.

“There were huge deficiencies in the way people understood health information,” said Sculli. “We recognised the profound impact that 3D imaging was having on industries like video games and movies, and we knew that if we were able to utilise the same technology to digitalise what’s happening inside the human body, we could have a similar impact on the way people understood their health.”

BioDigital and others are mapping the human body such that it can be explored in 3D by everyone from patients needing help visualising their conditions to experienced surgeons keeping abreast of advances in their fields. For instance, an interactive 3D model of the human head and neck built by scientists at the Glasgow School of Simulation and Visualisation using anatomical dissections and expert input is being used to train dental students to apply local anaesthetics.

These models promise a more immersive and accessible learning experience than most physical models (real cadavers are, after all, hard to come by), although perhaps the most intricate digital humans are designed not for teaching but for research.

In 1993, the International Union of Physiological Sciences Council meeting in Glasgow discussed the possibility of creating quantitative models to describe human physiology. This meeting set forth the ‘Physiome’ project and later the Virtual Physiological Human Institute (VPHi), which supports research and development into digital physiological simulations.

“[We aim] to come up with a suite of models that will allow us to take a holistic approach to prevention, diagnosis and treatment of diseases by means of computer modelling and simulation,” says Professor Liesbet Geris, executive director of the VPHi. Developing these models is no straightforward task, with researchers integrating the latest physiological knowledge with a range of computational techniques: from computational flow dynamics to machine learning, depending on scale, body parts and functions. This allows VPHi-affiliated groups to construct simulations of human physiology so detailed and reliable that they are already being used for safety testing of medical devices.

A VPHi-supported company called Medtronic recently used a digital trial to demonstrate that an alteration to its pacemaker design was safe when users underwent MRI scans. The US Food and Drug Administration (FDA) accepted that its modelling reliably reflected reality and approved the changes, negating the requirement for a two-year clinical trial.

While serious and legitimate concerns have been raised about the inadequacy of regulatory testing for some medical devices following a recent Guardian investigation, digital trials such as these could – under appropriate circumstances – reduce invasive clinical testing as well as offering an additional means of testing for safety. This is especially significant where there are practical barriers that restrict clinical trials (such as for paediatrics); integrating existing clinical information with digital evidence leads to more reliable conclusions than literature alone.

Geris says VPHi is collaborating with the FDA to develop rules and guidelines that will allow regulators to accept computer models and simulation as part of the evidence submitted for regulatory approval. “I think we have shown that computer models and simulations can be really useful in getting [medical devices] to the patient more quickly with less invasiveness,” she notes,“and the same trajectory is now starting for pharma.”

Geris hopes researchers will soon have ‘digital patients’ that integrate patient-specific information such as blood tests, allowing in silico trials to determine which patients groups benefit most from proposed treatments, and resulting in better targeted clinical trials.

Elsewhere, scientists are shifting from generic biological digital twins towards simulations that use individuals’ biometric data for more personalised models. A group of researchers at Imperial College London have been using deep-learning algorithms to analyse high-resolution imaging data and genetic information from thousands of volunteers in order to generate intricate models of the human heart. These models reflect the complexity and heterogeneity of cardiac anatomy and physiology far more accurately than the limited measurements cardiovascular patients receive in hospital.

“The end point isn’t just to create biophysical models of the heart,” says Dr Declan O’Regan, lead clinician for imaging research at Imperial College Healthcare NHS Trust. “We want to use those in clinical applications; we want to use them to understand the genetics of heart disease, and to use motion as well to make accurate predictions about when heart failure might develop and who is most at risk.”

The Imperial team’s heart models have already helped them to discover a common genetic defect that is related to a cause of heart failure. This has allowed clinicians to identify patients at higher risk of developing heart failure from their genetic information.

O’Regan is particularly excited about the group’s collaboration with the UK Biobank study, which could provide it with the cardiac MRIs and genetic information of 100,000 people; with this, the group could extend their techniques to populations of unprecedented size, applying machine learning to build more sophisticated and personalised models.

“If we have enough training data, we can train the algorithms not just on imaging but also on genetics and on other health data – blood tests, ECGs, other measurements,” he explains. With more measure­ments, it becomes possible to make predictions more patient-specific. “I think that’s something that patients really value,” he comments.

The techniques used to create these digital hearts are already being repurposed for other organs, including a bowel. A future possibility is to build systems of interlinked digital organs that can ‘talk’ to one another, allowing researchers to explore, for example, the relationship between cardiovascular health and dementia.

Construction of complete digital humans at this level of physiological detail may be out of reach now, but the possibilities raised even by relatively ‘basic’ digital models are beginning to transform how medical teaching, trials and research are done.

Sport

Digital athletes

Researchers at CSIRO’s Data61 – Australia’s leading data group – are building and using computational models of the human body, largely using fundamental physics to represent a range of activities inside and outside the body.

“Physical experiments have a range of limitations that do not constrain computer simulations,” explains Dr Simon Harrison, who leads Data61’s Digital Human Initiative. “It is difficult – or often impossible – to measure movements and strains of internal body parts, internal forces or stresses, and chemical changes inside the body without very invasive, expensive and undesirable procedures [but] we have shown that these measures can be calculated using computational simulation.”

These digital humans are opening up new possibilities in, among other sectors, the world of elite athletics, allowing sportspeople to test different techniques without putting themselves at risk of injury and overexertion. Using the digital models – which are often generated to include relevant surroundings, such as swimming pools – the researchers can run thousands of tightly-controlled but slightly varied simulations, and observe the impact on the digital athlete’s health and performance. For instance, one simulation developed by the team allowed coaches to experiment with new diving techniques that could be used by athletes competing in the Rio 2016 Olympic Games.

Sign up to the E&T News e-mail to get great stories like this delivered to your inbox every day.

Recent articles