Kenji Sunagawa: Thank you very much for inviting me to this exciting meeting. My name is Kenji Sunagawa. I am a cardiologist. I’ve been taking care of patients for over 40 years. From my experience, I witnessed major advances in medicine, but at the same time, I noticed serious unmet needs. So I’d like to focus on these issues.

Despite major advances in medicine, many diseases remain incurable, uncontrollable, or unpreventable. Advances are resource intensive, and excessive social burdens create socioeconomic disparity, as shown in the right panel.

So to overcome this problem, we must know what our enemies are. This is illustrated in the top ten causes of death in the states. The number one is ischemic heart disease. Number four is stroke. Number eight is hypertensive heart disease. They are cardiovascular diseases. So if we combine them, it looks like this. There is no question that cardiovascular disease is by far the most frequent cause of death in the U.S. So we need to overcome this disease for sure.

However, why do so many people die of cardiovascular disease? It is also the same question: What is the function of the cardiovascular system? So this illustrates the function of the cardiovascular system. Nature’s design goal for the cardiovascular system is to perfuse organs and the peripheral tissues to meet their metabolic demands.

The lack of perfusion immediately threatens life. Powerful control systems are essential to keep up with dynamic changes in their metabolic demand. It can reach easily ten times, but in an athlete, it can reach maybe 20 times. The cardiovascular system is very reactive, and a disease state is something like a condition in which the cardiovascular system fails to perform its intended function.

What disease threatens our lives? There are many heart diseases: mechanical disease, myocardial disease, electrical disease, and congenital anomaly, all with the heart. The cause is heart exhaustion. The two dominant causes of cardiovascular disease are ischemic heart disease and hypertension. These diseases attack the heart, and the heart becomes exhausted. That end-stage picture of cardiovascular disease is called heart failure.

You may anticipate that the life expectancy or probability of survival is pretty low. It is indeed pretty low. [4:16] This illustrates the survival rate [of heart failure] as a function of the number of years after the diagnosis. In the 1950s, five-year survival was 30% in men. In 2000 it is still 40%. We have been improving this a bit better, but not quite. The same is true for women. So the improvement of the management of heart failure is an urgent need.

How can we achieve this goal? So this illustrates the fundamental structure of medicine. The patients here, sensing, we sense many variables from patients, and based on the sensing, we make a diagnosis. Diagnosis leads us to interventions, maybe a drug or device, and apply it to the patient. It’s a constant feedback loop, for sure. We have made major advances in each component, sensing and diagnosing the patient. However, I’m not going to walk you through this busy slide. I just want to point out there are common limitations, that is, the lack of understanding of the pathophysiology of cardiovascular disease.

If we don’t understand the disease, we don’t know what we should measure with what we should sense? We do not know how to diagnose. We do not know how to intervene. So this is critical: to understand pathophysiology.

How can we do that? That is a good question. That is a systematic problem in our science. The left panel is native science, and we have massive data in each layer, like genes up to the system. But even in the same layer, the interaction of molecular mechanisms is very difficult to characterize.

As a result, the connection in the same layer among different mechanisms is not very established. If it is a connection across the layer, between the layers, that is almost impossible to know. So if it goes up through the system, we don’t have a good way of integrating this. So if some mechanism allows us to integrate horizontally and vertically, it will be extremely informative. That is what we expect, as shown in the right panel: the bio digital twin.

So even if it is simple, a single layer, single box guideline, it will still be very useful. So can we do that thing with the cardiovascular system? Let me take a look at the heart. We have massive data on the heart: molecules, muscle, organs, and their mathematical representations are possible. Again, in the vascular system, we have massive data and some mathematical representation. But still, some critical information is missing. Let me explain each one of them.

About 50 years ago, Japanese scientists found that the heart is an elastic chamber. The chamber elasticity was demonstrated to be very soft in diastole and stiff in systole. The time course of changes in elastance is shown here. [8:57] It goes from diastole to systole and back to diastole. The left panel is in dogs, and the right is in humans. Indeed, the time course of elastance is almost universal across various animal species. So we can derive mathematical representation rather easily.

How about the vascular system? We know that the vascular system consists of these core elastic properties. But we don’t know for sure about their distribution. If we can develop a framework that does not require exact distribution yet establish the characteristics of the vascular system, that should be very useful as a tool to connect the vascular system and the heart. So we worked on that and developed a generalized distributed vascular system model. We validated this model by experiment, and it turns out it is simple but very useful. Now we have the heart and vascular models in terms of mathematical representation.

So we are now ready to develop a prototype cardiovascular bio-digital twin. The top panel is the heart, and the bottom is the vascular system. This illustrates the output time series with ventricular pressure, aortic pressure, and left atrial pressure. And this is pressure on the loop of the left heart. However, this is not good enough as a cardiovascular model of the bio-digital twin because our cardiovascular system is very reactive. Depending upon the perturbation to our body, our system reacts. That aspect is not included in this system.

What we need is some mechanism to control cardiovascular systems. One of the most powerful mechanisms known is the autonomic nervous system. It is indeed the case that the autonomic nervous system brings the bio-digital twin to life and makes it reactive. The brain monitors the vascular system and sends the command to actuate the heart and vascular system. They constitute the feedback system. So if this feedback structure is acceptable, do we have data to quantify this relation?

The answer was no. So we had to develop data indicating the reactive nature of the autonomic nervous system through the cardiovascular system. This illustrates baroreflex impacts cardiac function, heart rate, vascular resistance, and intravascular stressed blood volume. They expressed reactive responses. So once we know this in theory, we should be able to develop a reactive bio-digital twin.

So we did experiments and tested how good this representation is. Now the top is pressure perturbation to the cardiovascular system. We give this perturbation to the bio digital twin and the estimated reactive pressure response as shown here. [13:26] When there are no reactive characteristics, the pressure response should be zero, so flat. We imposed the same pressure perturbation on the real animal and observed the pressure response as shown in the red line. Look at these two lines. They are indistinguishable. We feel comfortable that our bio-digital twin is now reactive, just like the native system.

So if we had a bio-digital twin, what can we do? First of all, I think a virtual diagnosis is easy to materialize. In virtual therapy, we can try many different therapies in the bio-digital twin. So it is possible. But what is most exciting, at least for us, is that autonomous therapy, such as autopilot systems in the airplane, self-driving cars, and Space X. A sophisticated bio-digital twin is a game changer that leads us to open up a new medicine. I’m going to show you where we are in our bio-digital twin project.  

We created a disease state in an animal model of acute myocardial infarction and monitored it by the bio-digital twin through the system identification. So we know what’s wrong with the cardiovascular system, using the background knowledge to help drug selection and usage. They generated command signals to the controller. The controller controlled actuators, where drugs were filled in each infusion pump, then activated the autonomous treatment. So this is a feedback mechanism. In the next slide, I will show you the performance of the system.

This column indicates hemodynamic response, blood pressure, cardiac output, and left atrial pressure. The combination of low blood pressure, low cardiac output, and high left atrial pressure is called cardiogenic shock and is life-threatening. As soon as we activate a bio-digital twin, it starts an infusion of multiple drugs. If you looked at cardiovascular parameters, the cardiac function returned normal, resistance returned normal, and stressed blood volume returned to normal.

At the same time, blood pressure returned to normal, the cardiac output returned to normal, and left atrial pressure normalized. Therefore the cardiovascular bio-digital twin can quickly restore normal hemodynamics. Well-trained cardiologists conducted this study at the National Cardiovascular Center in Japan. They were one of the best-trained people in the country, but they admitted they couldn’t do this. I mean, the bio-digital twin outperformed these trained physicians.

So why did it happen? It is clear that even if they are well-trained, they can’t control multiple drugs simultaneously, but the bio-digital twin can.

But the story doesn’t end here. We redesigned the model to include metabolic characteristics of the native system in the bio-digital twin. The upgraded model taught us that lowering heart rate reduced cardiac oxygen consumption. Reducing oxygen consumption is essential to improve the prognosis of the heart, but reducing heart rate crushes hemodynamics. Now, this showed reducing heart rate. These red lines represent hemodynamics. Hemodynamics stayed constant. But if you look at drugs, those black lines, they change dynamically. Thanks to these dynamic changes, hemodynamics remained constant. As a result, we could reduce heart rate by 50 BPM and oxygen consumption by 25%. This is a significant reduction in oxygen consumption, and it may positively impact reducing myocardial infarction size.

Let me wrap up. Bio digital twin deepens our understanding of the cardiovascular system and enables virtual diagnosis and treatment. Bio digital twin-driven autonomous therapeutic intervention will contribute to standardizing therapy, improve treatment quality, and enable medical services to be available beyond time-space constraints.

Bio digital twin will help improve the efficiency of health care and establish a sustainable society. Innovative improvement of the bio-digital twin is essential to the realization of autonomous therapeutic intervention.

In conclusion, focused research on unmet needs is essential for establishing a sustainable future in medicine. Thank you very much.

Moderator: Thank you very much, Dr.Sunagawa. We have about five minutes for the Q&A session. So if you have any questions from the audience, please raise your hand.

Attendee: Yeah. Great talk. I am David Gracious from Johns Hopkins. So I’ve heard several talks in the last few days on this exciting software program. And mostly, I’ve heard that you all are positioning it to do this autonomous therapy. But another opportunity is to find new drugs or ways to deal with heart disease. I’m curious if in your modeling you’ve discovered any, like maybe this is a therapy you should use that’s not currently used by cardiologists because that could also save an enormous number of lives. If you discover a new chemical or a new therapy that may not be used in like autonomous, but you could tell the doctors right now, all over the hospitals, you know, use this approach when someone comes in with a disease or an emergency. So the question is, have you discovered anything new that is not known, you know, to cardiologists today?

Kenji Sunagawa: Well. Thank you very much for asking an excellent question. The short answer is no. We have not discovered any new drugs. But we know what we need; when we start an autonomous intervention, it becomes very clear that if we give drugs that change many things simultaneously, they complicate the control. If a particular drug changes a particular property of the cardiovascular system, then adding different drugs is not so complicated. We can control each critical variable one by one with a corresponding drug. But I don’t see any drug that can do that. As a result, when we use one drug, we often need a second drug to cancel the undesired effect of the first drug. Therefore, if possible, we’d like to have a drug that works on an isolated property of the cardiovascular system. But I’m not sure it is possible. So, as I said, the short answer is no, and in terms of the easy usage of drugs to promote autonomous treatment, the uni-functional drug is preferable. That is my answer.