How Computer Trials can help Doctors Help us

When buying a car, choosing between two can be tricky. One might look cooler, but the other might be safer or use less petrol. To really know which one is better, you would have to drive both in different places, over many weeks, and in different weather conditions. That takes a lot of time, money, and effort. Now think about having a digital copy of each car that you can test from your couch. The computer can show how each car drives, how safe it is, and how it handles weather like rain or snow—all without ever stepping outside! That would make it much faster and safer to choose which car you prefer.

Doctors face a similar challenge when deciding between two medicines. They need to find out which one works best, and that usually means testing them on lots of people. This is called a clinical trial. These trials are important, but they can take years and involve real people who might get side effects. As in our car example, what if there was a way to test new treatments more safely and much faster than in a clinical trial? Imagine if, instead of testing treatments directly on real people, scientists could first test them on computer versions (called models) of human organs, like the heart. This is called an in silico trial. The term “in silico” comes from silicon chips in computers. Your heart is an important organ that keeps you alive by pumping blood around your body. Digital hearts behave just like real ones, allowing scientists to see how different treatments might help (or cause problems!) before trying them on real patients. By testing treatments in the digital world first with an in silico clinical trial, we can make healthcare smarter, faster, and safer!

What are Cardiac Digital Twins?

Before we can start an in silico clinical trial to test out the best ways to keep the heart healthy, we must create a digital model of a patient’s heart, sometimes called a digital twin. Think of creating a digital twin like designing a video game avatar. We want to customize your avatar so that it represents you, or in this case, your heart. We can match heart shape and size from special photographs that look inside the body. In this Frontiers for Young Minds article scientists explained how to create a 3D image of the heart by assembling a stack of 2D pictures of the chest. They first highlight (using a process called segmentation) the heart in the picture by carefully coloring the area of the image representing the heart. When they assemble this stack of heart pictures, they end up with a three-dimensional model that represents the heart shape of a specific patient. Whether someone is a boy or a girl, their height, and how much they play sports can all change the shape and size of patients’ hearts. Researchers have found that differences in the shape of the heart can have a big impact on how it works [1].

Heart shape and size are not the only things that change from person to person. For example, for every heartbeat, an electric wave flows over the heart, making the heart muscle squeeze. The speed of the electric wave and the strength of the heart muscle squeezing might be quite different if you compare your heart to your parents’ or your friends’ hearts. So, to finish customizing your heart avatar, we need to replicate your own heartbeat, using measurements from the hospital to give us clues. In the digital world, there are a lot of different settings that control various aspects of how your heart beats. Using physics and maths, combined with what we learnt from the hospital measurements, we can make the digital heart beat and pump blood. By adjusting the avatar settings, we can change how the digital heart beats. Finding the right combination of settings can require a lot of trial and error until we match your heartbeat as closely as possible.

From one Heart to Many: Building Models in a Flash

Doctors want a particular drug or surgery to work on everyone who receives the treatment. Therefore, they test it on as many patients as possible to make sure that it works for all different types of patients. In the past, clinical trials have used more men than women to test out their medicines. Since they did not test their treatments on enough women, sometimes they did not work as well in women as in men. It is important that a diverse group with equal numbers of men and women be included in clinical trials, as well as people of a variety of races and ages. So, in a traditional clinical trial, doctors usually gather large numbers of patients from many different backgrounds. Similarly, for in silico trials, we must create many models so that the trial represents as many patients as possible (Figure 1).

It is not easy to create heart models of a lot of people. First, segmenting the heart can take a lot of time. Artificial intelligence (AI) is a very useful tool for speeding up this process. Using AI, we can train our computer to be a speedy research assistant for this job by showing it different images and the models we made from those images. AI can then quickly process lots of images and identify the heart using patterns and clues it learnt during training. Instead of taking a few hours to create a 3D model of the heart, AI can create the model for us in minutes (See this article to learn more about how AI is used in medicine)!

After building the models, finding the right settings to customize different heartbeats can take a lot of trial and error, which is very time consuming, and each cardiac simulation requires a lot of computer power. Right now, you would need a supercomputer (more powerful than 90,000 PS5 consoles put together) to test out even one setting combination. If we want to customize a lot of different hearts, the amount of time and computer power needed rises, and it quickly becomes an impractical task. So, how can we generate the number of heart models needed for an in silico clinical trial? Again, AI comes to the rescue! Instead of using supercomputers to test all the possible setting combinations, we can just test out a few different combinations and then use this data to train a faster and simpler AI model. Then, we can explore all the possible setting combinations quickly, on a standard computer!

It still takes a lot of time and computer power to train an AI model. But scientists hope that by using all the information AI models have already learnt, customizing new hearts will become even easier in the future. After creating many heart models, we can then use our army of heart models for an in silico clinical trial.

From Models to Medicine

Scientists have started using their groups of models to try out different treatments. You learnt earlier that your heart has its own special kind of electricity that tells it when to squeeze and pump blood. But, sometimes, especially after someone has been really sick or had a heart attack, the electricity does not flow the right way. It can get mixed up or stuck, which can make the heart tired and weak.

To help fix this, doctors can put a small device (called a pacemaker) in the chest that sends tiny, gentle electric signals to help the heart squeeze the right way again. However, doctors are still trying to work out the best spots from which to send those signals. Using their groups of heart models, scientists can figure out which places in the heart would work best for those signals, and how to make them happen at exactly the right time [2]. That way, the heart can go back to squeezing as it is supposed to.

Another way scientists are using these models is to help children who are born with unusually shaped hearts. Most hearts have two pumps—one that sends blood to the lungs to get oxygen, and another that sends blood to the rest of the body. But some children are born with only one pump working properly. Doctors have learnt how to fix this with special surgeries, which help children live and grow, but many kids still get tired easily and cannot run around as much as their friends. By using heart models to try out different types of surgery, doctors and scientists can work together to predict which surgery will work best for a child, so they have the best chance to stay healthy and active [3].

Now that we can build models quickly and use AI to help us train them, digital heart models could become a really powerful tool in hospitals, helping doctors choose the best treatments and giving patients better, safer care.

Glossary

Clinical Trial: ↑ A research study where doctors test new drugs or treatments on people to see if they are safe and work well.

Model: ↑ A computer-made version of something real, like the heart, that helps scientists study how it works and test new ideas safely.

In Silico Trial: ↑ A virtual experiment that uses computer models instead of people to test how drugs or treatments might work.

Digital Twin: ↑ A computer-made “twin” that copies how something works in the real world. A cardiac digital twin is like a digital twin of someone’s heart.

Segmentation: ↑ A way to pick out part of a picture by coloring it and leaving out the rest. We use this to select the heart from images and ignore the rest of the body.

Artificial Intelligence: ↑ A computer technology that can learn from examples and solve problems almost like a human.

Supercomputers: ↑ Super-fast and powerful computers that help scientists figure things out much faster than regular computers.

Conflict of Interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Acknowledgments

TMGB acknowledges support from the Mechanics of Life doctoral scholarship programme. RKB is supported by the UK Engineering and Physical Sciences Research Council [EP/T517963/1]. CR is supported by the British Heart Foundation (RG/20/4/34803) and The Alan Turing Institute. SAN acknowledges support from the UK Engineering and Physical Sciences Research Council (EP/M012492/1, NS/A000049/1, and EP/P01268X/1), the British Heart Foundation (PG/15/91/31812, PG/13/37/30280, SP/18/6/33805), US National Institutes of Health (NIH R01-HL152256), European Research Council (ERC PREDICT-HF 864055), and the Alan Turing Institute. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

AI Tool Statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.