In the study and management of aortic aneurysm, an Italian startup has developed a digital twin, significantly aiding in the treatment of this condition through the use of artificial intelligence

Harnessing the potential of digital twins in medicine can be immensely beneficial in the study of aneurysm, one of the most widespread cardiovascular diseases in Europe. It affects approximately 700,000 individuals and sees about 220,000 new cases diagnosed each year. In Italy alone, there are around 6,000 annual fatalities from this rapidly fatal pathology.

This opportunity gave rise to the European Horizon 2020 MeDiTATe project (Medical Digital Twin for Aneurysm Prevention and Treatment), resulting in the creation of LivGemini, a startup spin-off from the University of Rome “Tor Vergata”. The CEO and co-founder of this venture is 27-year-old biomedical engineer Leonardo Geronzi. Recently, this emerging medtech entity won the National Innovation Award in the Life Sciences-Medtech category.

---

TAKEAWAYS

---

Digital Twin in Medicine: The MeDiTATe Project and LivGemini Startup

Leonardo Geronzi, the CEO and co-founder of LivGemini, a spin-off startup from the University of Rome ‘Tor Vergata’, is a biomedical engineer with a deep passion for numerical simulations, or the computerized reproduction using mathematical algorithms of hemodynamic and biomechanical phenomena in the cardiovascular field. Geronzi’s initial goal at LivGemini was to provide doctors with tools to save as many lives as possible through accurate diagnoses. After graduating from the University of Pisa in 2019, he joined the emerging European MeDiTATe project the following year. Funded by the EU with €3.7 million, the project involves 24 partners, including universities, research centers, and industrial companies (among them Philips and General Electric’s healthcare division) throughout Europe. Geronzi is among the 14 active researchers, engaged in doctoral research on individual projects aimed at providing a complete framework of simulation technologies and imaging for industrial and clinical translation. The goal is to accelerate the process of personalized cardiovascular medical procedures.

The Italian scientist, along with Marco Evangelos Biancolini, a professor of Machine Design at the University of Rome “Tor Vergata” and the principal investigator of the four-year MeDiTATe project (focused on developing digital twins in medicine), has concentrated on the ascending aortic aneurysm. This condition is a pathological and permanent dilation of the first segment of the main artery of the human body.

“In the three years of research in which we developed specific simulation methods, we discovered that doctors handle the issue with a certain standard approach, evaluating solely the diameter of the vessel,” the CEO of LivGemini explains. “If the dilation exceeds 50 millimeters, surgery is performed; otherwise, the patient is managed with pharmaceutical care. From discussions with doctors, we understood that aortic ruptures can occur even below the 50 mm threshold, necessitating a more targeted and effective method to understand the phenomenon. Thus, we decided to tackle this clinical-medical problem by employing artificial intelligence methods.“

The Role of Artificial Intelligence

To develop a digital twin in medicine that is as precise and effective as possible, AI lends a hand:

“Thanks to artificial intelligence, it’s possible to capture properties otherwise not understandable at both the engineering and medical level. For this reason, we began to implement AI algorithms not just based on image properties, but also on the results of the numerical simulation performed on the aorta, extracting data from hemodynamic and biomechanical modeling.”

In the work conducted by LivGemini, various artificial intelligence techniques have been employed. Deep learning, particularly U-Net, a specific convolutional neural network for segmenting biomedical images, has been used to extract a real-time three-dimensional anatomical model, significantly faster than the manual methods still employed today.

“Furthermore, to assess the risks associated with aneurysm, we have adopted machine learning techniques starting from anatomical features, thereby delivering a personalized risk threshold for each patient.”

From the cross-referenced work on as accurate data parameters as possible, we have reached a predictive model capable of ensuring up to 94% accuracy. Geronzi explains the ongoing optimization project for the digital twin:

“Generally, digital twins can be divided into passive, semi-active, and active. To date, we have reached a semi-active model that allows the generation of a virtual anatomical replica. We have trained additional machine learning models, consisting of compressed data from numerical simulation results, and now we can conduct ‘what-if’ analyses, adopting different sets of values in one or more formulas to explore various scenarios and outcomes. For example, by varying the arterial pressure parameter on the digital model, it is possible to understand what might happen to the wall subject to dilation“.

The Future of Medical Diagnostics: Real-Time and AI-Driven

For the future application model of digital twin in medicine, specifically for the analysis of aortic aneurysm, LivGemini aims to build an active digital twin, as complete as possible and capable of ensuring an exact correspondence between the virtual and real models.

What are we still missing? The ability to acquire real-time patient data through wearable devices, thus monitoring the patient at every moment and in real-time, and, if necessary, being ready to intervene promptly at the onset of a problem.

“At the foundation, there is always the need to perform a 3D anatomical scan, extracting all possible useful information to outline the patient’s clinical picture. Once obtained, the model is trained to provide answers about the current state of the patient. From there, data can be extracted from wearable devices allowing the model to evolve, enabling it to follow the patient in everyday life“.

In the future digital twin application model in medicine, to be precise for aortic aneurysm analysis, LivGemini aims to build an active digital twin that is as complete as possible and can ensure an exact match between the virtual and real model.

To construct all these parameters, LivGemini has developed a specific AI-based software, Fusion V, set to be released in 2026. Work is currently underway on optimizing and scaling the tool commercially, now at TRL 5, to provide a very simple-to-use solution. It will be released in the near future after proper certification.

In the meantime, the spin-off is also working on augmented reality solutions, providing doctors with an even more accurate and immersive view, ensuring a safer context for a broader understanding of the problem.

Looking even further ahead, where do we aim to go?

“Throughout our journey with MeDiTATe, we have collaborated extensively with young doctors, more inclined towards technology and its contribution to medicine. In these discussions, a need emerged for alternative and technological solutions to address and solve medical problems. Therefore, I foresee that technology in the future can reduce diagnosis times as much as possible, providing a clearer picture and reducing uncertainty. But above all, capable of speeding up all diagnostic and prognostic processes, enabling the development of effective strategies to counter and overcome problems. For this reason, I anticipate an ever-increasing contribution of artificial intelligence in medicine, increasingly based on AI algorithms“.