A new ally for personalised transcranial electrical stimulation in Alzheimer’s disease
Neurological and neuropsychiatric disorders are among the leading global causes of disability and health loss, accounting for a substantial and growing share of Europe’s disease burden in terms of both human suffering and economic cost. According to current estimates, around 7.85 million people in the European Union were living with dementia in 2018, a figure projected to rise to 14.3 million by 2050. Alzheimer’s disease (AD), the most common form of dementia, accounts for 60-70% of all cases. Currently affects more than 10 million individuals across Europe, with prevalence expected to increase further as populations age.
Despite the scale and urgency of this challenge, therapeutic options remain largely inadequate. Existing treatments are often limited in efficacy, poorly personalized, and associated with highly variable outcomes. This highlights the need for more effective and individualized interventions capable of addressing the biological complexity and heterogeneity that characterize brain disorders such as AD.
Non-invasive brain stimulation as an emerging strategy
Against this backdrop, non-invasive brain stimulation has emerged as a promising therapeutic avenue. Encouraging results have reported in conditions including epilepsy, depression, and Alzheimer’s disease. In particular, transcranial electrical stimulation (tES), and more specifically transcranial alternating current stimulation (tACS), has attracted growing interest. This is due to its ability to modulate brain activity safely and painlessly through the application of low-intensity electrical currents via scalp electrodes.
However, clinical outcomes remain inconsistent and strongly dependent on individual brain characteristics, which are still only partially understood. A key limitation lies in the modelling approaches traditionally used to design stimulation protocols. These are largely based on static anatomical information and passive electric field simulations, which fail to capture the brain’s dynamic, multiscale organization and its active response to stimulation.
Addressing this gap is the central ambition of the Neurotwin project. This project aims to transform the way brain stimulation is designed and applied by developing personalized computational models. These models are capable of dynamically simulating the effects of interventions before they are delivered.
The promise and limits of non-invasive brain stimulation
tES has shown therapeutic potential in a range of conditions, including Alzheimer’s disease, epilepsy, and major depressive disorder. By influencing neuronal excitability and synchronizing brain oscillations, it offers a novel way to restore or enhance cognitive function and network connectivity. Importantly, it can be used both as a primary treatment and as an adjuvant therapy alongside other interventions, such as cognitive training.
However, clinical outcomes remain highly variable. While some patients benefit significantly from non-invasive brain stimulation, others experience little or no improvement. A key reason for this inconsistency lies in how stimulation protocols are traditionally designed. Today, brain stimulation strategies are largely based on anatomical information derived from MRI scans, which is used to estimate how electrical currents distribute across the cortex. Such models operate on the assumption that cortical field intensity and location directly predict clinical outcomes. In essence, they posit a linear relationship between field strength and therapeutic benefit.
Even when MRI data are used to personalize these models, they often overlook crucial aspects of brain function. Key aspects include the ongoing neural activity at the moment of stimulation and the functional connectivity between regions. A third critical element is the multiscale dynamics that govern signal propagation through brain networks. In doing so, the brain is effectively considered as a static and passive structure, rather than as a dynamic system that actively responds to external perturbations.
As a result, traditional models offer limited predictive power for optimizing treatment strategies. They also struggle to deliver consistent clinical benefits, particularly in complex neurological and neurodegenerative conditions. These limitations point to the need for a different modelling approach, one that can account for individual brain dynamics rather than relying solely on static anatomical information.
Digital twins for personalized neuromodulation
In recent years, digital twins have gained traction in biomedical research as advanced computational models capable of replicating the real-time behavior of organs or systems using individualized data. When applied to the brain, these models simulate how neural networks respond to various stimulation strategies. They also provide a safe and controlled environment in which to test interventions before applying them in clinical settings.
The Neurotwin project: scope and consortium
One of the most ambitious initiatives to bring this concept into clinical practice is Neurotwin, a European research project funded by the European Union’s Horizon 2020 programme with a grant of €4 million. Running from January 2021 to December 2024, the project was coordinated by Neuroelectrics Barcelona under the scientific leadership of Giulio Ruffini, co-founder of the company and a pioneer in the field of brain modelling and neurotechnology. It brought together a multidisciplinary consortium of eight partners across Europe and the United States. These included: Universitat Pompeu Fabra (Spain), Universidad Pablo de Olavide (Spain), Forschungsgesellschaft für Arbeitsphysiologie und Arbeitsschutz (IfADo) (Germany), Uppsala Universitet (Sweden), Massachusetts General Hospital (MGH) (USA), Beth Israel Deaconess Medical Center (USA), and Fondazione Santa Lucia IRCCS (Italy).
Building personalized digital twins of the human brain
The core objective of Neurotwin was to build personalized digital twins of the human brain. These computational models integrate individual anatomical (MRI), functional (EEG), and electrophysiological data. Unlike traditional models limited to static representations, these twins simulate the dynamic behavior of large-scale brain networks across time, both in health and disease. This allows for a more precise understanding of how stimulation protocols interact with ongoing brain dynamics.
To achieve this, the project followed a three-pronged modelling strategy. First, individualized finite-element simulations of the electric field were computed from MRI/CT data to identify how currents propagate through each person’s head. Second, whole-brain network models were adapted to reflect subject-specific connectivity and oscillatory activity. Finally, in silico therapy design allowed clinicians to virtually test a range of stimulation parameters. For example: including montage, intensity, frequency, and timing, to identify the most effective strategy before applying it in clinical practice.
Clinical insights from the Neurotwin pilot study
Within Neurotwin, the concept of personalized brain digital twins has moved decisively beyond theory and into clinical testing. A pilot study conducted at IRCCS Fondazione Santa Lucia in Rome enrolled 30 patients with mild to moderate Alzheimer’s disease. The study adopted a randomized, double-blind, sham-controlled, crossover design to evaluate a model-optimised transcranial alternating current stimulation protocol. In this design, each participant received both sham stimulation and an active, personalized intervention, allowing researchers to assess individual responses under tightly controlled conditions.
The study delivered stimulation daily at home over a period of eight weeks, for a total of forty sessions. It used a home-based neurostimulation system developed by Neuroelectrics and administered by a trained caregiver under remote clinical supervision. As such, the study represents one of the most structured attempts to evaluate long-term, home-based neuromodulation in a neurodegenerative population. It overcomes the practical constraints that have traditionally limited non-invasive brain stimulation to short, clinic-based protocols.
The first and most relevant outcome concerned, indeed, feasibility and safety. Repeated stimulation was successfully administered outside the clinical setting, even in a fragile patient group, with no critical safety signals emerging over the course of the intervention. Beyond feasibility, the trial provided robust neurophysiological evidence that the stimulation engaged its intended targets. Patients receiving stimulation guided by their digital twin models exhibited stronger modulation of individual brain oscillations, particularly in the gamma frequency range. They also showed greater functional connectivity in brain regions typically disrupted in Alzheimer’s Disease. These neurophysiological changes were paralleled by improvements in cognitive test scores, especially in memory-related tasks, compared to those who received sham stimulation.
Understanding why personalised stimulation works
Crucially, the trial also validates the predictive power of the Neurotwin models. These models are not “black boxes” that simply state which stimulation protocol to apply. Instead, they clarify why a given protocol is expected to work in a specific brain, based on its individual structure and dynamics. As Giulio Ruffini, coordinator of the project, explains, «For years, we have treated stimulation as a black box. With personalised Neurotwins, we can better diagnose and run safe, what-if tests before touching the patient, opening the door to precise, consistent patient personalisation and dosing of treatment.» This approach makes the models inherently comprehensible. They explicitly describe how electrical currents propagate through the brain and how large-scale neural networks generate and sustain activity.
Importantly, the digital twin is not a fixed model. It can be updated as new data are collected before and after stimulation, allowing predictions to be refined over time. Within the trial, the outputs generated by each digital twin are directly compared with the neurophysiological and clinical responses observed after treatment. This capability extends the role of digital twins beyond protocol optimisation, opening the possibility of testing and refining stimulation strategies in advance and reducing reliance on trial-and-error approaches in clinical practice.
References
1. Mencarelli, L. et al. A Randomized Sham-Control Trial for Home-Based Transcranial Alternating Current Stimulation (tACS) In Alzheimer’s Disease (2025).
2. Sánchez-Garrido Campos, G. et al. Preclinical insights into gamma-tACS: foundations for clinical translation in neurodegenerative diseases. Front. Neurosci. 19, (2025).
3. Vohryzek, J. et al. Dynamic sensitivity analysis: Defining personalised strategies to drive brain state transitions via whole brain modelling. Computational and Structural Biotechnology Journal (2023).
4. Vohryzek, J. et al. Design of effective personalised perturbation strategies for enhancing cognitive intervention in Alzheimer’s disease (2023).
