• Tanvi Sheth

Preventing Dementia Through A Virtual Twin Brain: Notes from Dr. Steven Levine’s webinar

At BrainSightAI we are fascinated with Dassault Systemes’ work in 3D modelling and decided to list our learnings from a seminar from their team. In the seminar Preventing Dementia Through A Virtual Twin Brain, Dr. Steven Levine, Senior Director of Virtual Human Modeling, talks about The Living Heart project and the Virtual Twin Brain project. You may view the webinar here.



3D Modelling and simulations


Virtual systems and 3D simulations can be used to improve the odds of success of multiple products and inventions, including aeroplanes, automobiles, and even medical imaging and reconstruction. Dassault Systemes have been involved in collaborating with multiple entities, like research institutes, hospitals, and governing bodies to work on creating virtual copies of two critical organs - the human heart and the human brain.


When it comes to the human body, in order to accurately understand and encompass its complexities, the first step is reverse engineering. The best way to do this is to start by reconstructing 2D medical scans into 3D representations. It has become possible to extend such work into orthopaedic, cardiovascular, and neurological fields, by segmenting images into fine details and assigning realistic behaviour, to understand what goes on underneath.


The Living Heart Project


To address the challenge of promoting medical understanding of physiological processes virtually, the Living Heart Project was undertaken. The driving factor was that being able to peer inside the heart via a simulation could help clinicians in solving the world’s most prominent cardiovascular problem - heart disease. How the team built a digital twin of a human heart is outlined by the points given below.


  • The heart is multiscale: The heart has to be captured from its function right down to its molecular level. Therefore, they built the phenomenological model of the heart from detailed tissue models. It could therefore serve as a foundation to incorporate other scales further down the link, which is the cellular biophysical chain and biochemical pathways, ultimately narrowing down to molecular level behaviour. Then using AI and other techniques, such patient-specific models could be aggregated to population models.

  • Building the living heart: In order to create a fully functional heart model based on real human heart mechanisms, the geometric representation of the real heart tissue from radiological images was generated. Detailed fibre orientations also had to be accounted for as they control electrical and structural response. The precisely specified electrical system was coupled with the physical model to activate the typical pumping motion. Valves, fluid pressure and cavities, and other biophysical properties must be included in the model, which is then connected to a system model, which represents circulation (haemodynamic model). The result of all of these combinations was a complete beating heart model (Living Heart Model 1.0)

  • Leveraging potential: With the help of the 3D model of the human heart, the entire functioning of the heart can be considered, which allows exploration of structural as well as electrical problems such as valve disease and arrhythmias respectively. Apart from this, medical device designing and testing becomes possible completely virtually, without animal testing or clinical trials.

  • Case studies: For patients whose heart valves are not functioning efficiently, and who are unable to go through a surgical procedure, a minimally invasive alternative of TAVR is conducted. The Living Brain model has made it possible for surgeons to carry out a virtual procedure of implanting a Trans-Aortic Valve Replacement (TAVR) device and testing its effectiveness against different problems related to the aortic valves. Virtual surgeries are a breakthrough, as they give the surgeon an eye inside the procedure, before it is actually carried out, thus minimising associated risks.

In another case, collaborators and researchers address cardiotoxicity, specifically drug-induced arrhythmias. They are able to map the molecular level change (a blocking potential) caused by the drug onto a full 3D model, predict the effects of conduction through the living heart, which then produces clinical measures such as ECG or ejection fraction. Drug dosage safety analysis is another area covered by the Living Heart project.


The Living Brain Project


Following the success of the Living Heart Project, Dr. Levine began to explore the development of a virtual human brain model. While this project is still underway, its scope and potential can be outlines by the following applications -


  • Traumatic Brain Injury (TBI): For TBI, researchers generated a head-and-neck model instead of just the brain model. In order to do this, clinical image data was transformed into a 3D geometrical structure, and then segmented into the necessary functional elements. For such a model, the properties only represent the physical characteristics of the head, but interpreting the psychological impact of brain function as a result of TBI would require a more detailed model. However, this model is still extremely sophisticated, with 30 million degrees of freedom, three and a half million elements to describe the details of the human head and brain, and 33 distinctly identified anatomical structures for independent analysis. Personalized 3D simulations could significantly aid decompressed craniotomies carried out for TBI.

  • Neuromodulation: When it comes to non-invasive and invasive neuromodulation, like TMS, ECT, and DBS, there are a number of challenges and risks to overcome. For Deep Brain Modulation (DBS), Using the methodologies to reconstruct a virtual brain, a map of areas of critical function were created. While being a well-established therapy, there are still aspects of DBS that can be improved with such technology, including appropriate targeting, accurate placement of electrodes, and the precise delivery of electrical energy impulses.

  • Neurodegenerative disorders: A prominent hypothesis for neurodegenerative diseases is that certain toxic proteins self-propagate and create lesions, which lead to cell death and brain tissue atrophy. While the disorder may be experienced differently by different people, due to their unique dysregulation of protein, the underlying mechanism (as explained above) is assumed to be similar. Such propagation mechanisms can be investigated in the form of longitudinal clinical studies with the help of the Living Brain model.

The brain is reconstructed digitally using protocols similar to the Living Heart Project. To understand disease progression, the researchers compute biomarker abnormality and create a temporal map of the toxic protein. Using a propagation model, the time evolution and path of the proteins can be predicted and analysed. Along with this, tissue atrophy along the grey and white matter areas can be mapped, and in post-processing, tissue shrinking is also computed.


In the future, BrainSightAIwould also like to develop a connectome-based model for dementia and other neurodegenerative diseases. This can be of particular importance in understanding the fidelity of patient representation and interpretation of what actually may be happening to the patient.