Artificial intelligence could protect the elderly and the sick from falls – this new hungarian invention could save lives

Dániel János Daróczi, a physics student at the Faculty of Science of Eötvös Loránd University, is developing a wearable technology within the Anchor Dynamics project that uses artificial intelligence and sensors to detect the signs of movement instability — and prevent falls before they happen.

What motivated you to embark on developing such an extraordinary invention?

My mother has been living with Parkinson’s disease for twelve years, which means I have had to face this illness since early childhood. That alone would be a powerful impulse for anyone. Through my mother, I came into contact with the patient community, where I met people like her in hospitals, during treatments, and within tightly knit patient groups.

These encounters reinforced my impression that the conventional problems most people associate with the disease — such as tremors — can now be treated to some extent, thank God. Surgical interventions are available in several neurological conditions, such as deep brain stimulation, which can reduce tremors in patients.

At the same time, these experiences made me realize that the greatest burden of the disease is the fear of falling.

That realization left a deep mark on me.

This aspect of the illness is rarely discussed.

Symptoms can be alleviated to a certain degree, but no medication can prevent a patient from losing their balance and falling.

Is loss of balance related to tremors?

It’s an important question, but there is no direct connection between the two symptoms. On the one hand, medications themselves can worsen balance; on the other, the disease also affects it. And then there are elderly people who have not been diagnosed with any illness, yet still suffer from frequent falls.

If we think about it, hip fractures among elderly relatives occur in almost every family. This is a serious issue because it can be demonstrated quantitatively that in people over the age of 65 hospitalized with such injuries the mortality rate exceeds 25 percent. Patients do not die directly from the fall or the fracture, but from the consequences: they are bedridden for montsh, weaken physically, lose their appetite and their health indicators decline. Beyond this, the mental impact is also significant: the vulnerable state they find themselves in, the loss of autonomy, self-confidence and self-esteem.

What can be done in such cases?

The options are fairly limited. We can place the patient in a wheelchair, but then they lose even the minimal muscle activity and coordination they would otherwise retain through movement. This is why patients themselves dislike this solution — they are reluctant to give up their freedom of movement.

At the same time, moving freely can be dangerous. My mother falls frequently, hits her head, and suffers abrasions. This is extremely difficult to handle both for the patient and for those who care for them.

In one of your lectures, you said that in an average Hungarian family there are two people living with neurodegenerative diseases (such as sclerosis multiplex, Alzheimer’s disease or Huntington’s disease), as well as one person who has suffered a stroke.

Worldwide, nine million people live with Parkinson’s disease, and tens of millions suffer from other conditions where falling is a potential consequence. There are also hundreds of millions of elderly people who do not suffer from any specific illness but are still affected due to impaired motor coordination.

I have long been thinking about how this problem could be addressed technologically, because I believe that many of the world’s challenges can be solved through technology.

Our project was launched within the framework of the Hungarian Startup University Program, which supports university students. The Faculty of Science at ELTE also has a focal unit that helps turn ideas into reality. I submitted an application that was accepted, allowing me to form a team and participate in a competition where we were able to secure financial support to launch the project.

Who makes up the team?

An engineer, a programmer, and myself as a physicist.

What is each team member responsible for?

The engineer oversees the hardware implementation, primarily the development of the sensor system. The programmer is responsible for the software, ensuring that—with the help of artificial intelligence—we can train the system on human movement patterns and thus predict loss of balance and falls. I am responsible for mathematical modeling and communication.

How does the invention of Anchor Dynamics detect signs of movement instability?

The sensing software attempts to make mathematical predictions based on movement patterns.

The idea is somewhat reminiscent of the sci-fi film Minority Report, where precogs predict crimes before they happen, allowing police to intervene in advance. What prevents the patient from actually falling?

The system is built on three pillars. The first is sensing. Human movement and balance are coordinated through the cooperation of visual input, the balance organs in the inner ear and the brain. This is what we need to model and replace technologically.

We have built a sensor system that includes multiple sensors placed on different parts of the body. Each sensor contains an accelerometer, a gyroscope and a magnetometer. When placed on the limbs, this system can “tell” the user’s current physical state.

So the invention has to be worn?

Exactly. With sensors placed on the legs, we can determine how the leg is moving at any given moment. This constitutes the sensing component. The next step is predicting what will happen based on the collected data.

This is where machine learning and AI come into play. Purely physical modeling is insufficient to determine future outcomes, but with AI, the system learns how the patient typically moves, specifically learning the individual’s unique movement patterns.

When an anomaly is detected — when a fall becomes likely — the system intervenes.

How can the system recognize the precursors of a fall if it has never “seen” one before?

We feed various movement forms into a program that simulates real-world behavior and multiplies the data. It’s as if we were moving a virtual mannequin according to the patient’s movements, then replicating that mannequin countless times.

This means we don’t need to observe a person for tens of thousands of hours, the artificial intelligence learns from simulated data. The algorithm learns from a massive amount of incoming information. The software itself already exists; it just wasn’t developed for this purpose originally.

How long does it take for the AI to adapt to a specific person?

A few hours. From the moment the product is purchased, one day is enough for the system to learn the patient’s movement patterns.

Incredible. What exactly is this wearable robotic device, which you have named an exoskeleton?

You can imagine it as an external skeleton worn on the body — hence the name — that can improve the wearer’s movement.

We understand that intervention is possible before a fall occurs, but who or what actually prevents the fall?

The exoskeleton applies counterforce at the joints. When the patient can no longer correct their movement on their own and would otherwise fall, the system intervenes and holds them back.

Electric motors mounted on the exoskeleton generate the necessary forces at the joints.

Could the system fail to intervene, or intervene insufficiently?

Of course, unforeseen situations can always occur, for example, if the patient trips over an object. But we have considered this as well. We are already planning to integrate an airbag system similar to those used in cars. This technology already exists and is used by motorcyclists. We want to incorporate it so that our system provides comprehensive protection.

When could the device become reality and enter a testing phase?

We have focused primarily on software development. The invention can become a real product once the software is perfected—once it can make accurate predictions—and once we succeed in integrating our technology into an external partner’s product.

Exoskeletons already exist today, for example in the military, the tourism sector, and in rehabilitation, and the market is growing rapidly. Integrating our invention into an existing product would make the implementation faster and more cost-effective.

How much would it cost to produce the final product?

Tens of millions of forints would be required before full realization. First, we need to create a pilot project that can demonstrate the product, after which it could become widely accessible.

We have already established at the beginning of the conversation that there is a clear need for such a solution.

As populations age, the problem will grow exponentially. Yet this invention could save enormous costs: from hospital beds and overburdened healthcare staff to medication and caregiver fees, and the financial, physical, and mental exhaustion of families.

The English translation was produced with large language model and reviewed by a human editor.