The goal of the TwinCare-AF project is to develop innovative core methodologies for accurate and real-time calibration of cardiovascular electrophysiological models and to support medical decisions in the context of atrial fibrillation and catheter ablation therapy planning. The proposed approach will focus on the generation of digital twins of patient hearts, calibrated through robust and efficient machine learning techniques, and able to replicate measured clinical data, such as electrocardiogram and electrogram recordings. Specifically, physics-informed and/or deep-learning techniques will be extended and implemented within the context of anatomically-accurate and biophysically-detailed cardiac electrophysiology, to accelerate the solution of classical forward electrophysiological model, and to solve inverse problems for identifying patient-specific physical and tissue properties of the heart. Additionally, a robust methodology for verification, validation, and uncertainty quantification will be adopted to showcase the agreement between model predictions and empirical observations, and to provide reliable estimates of confidence in the model predictions. The developed approach will be used to predict atrial fibrillation progression and determine potential ablation sets for individual patients. The predictions of the developed model will undergo testing through in vivo intraoperative clinical measurements. To enhance easy flow, robust analysis, and interpretation of patient-specific data, the novel real-time mathematical workflow for atrial fibrillation simulations will be integrated into a clinically viable platform. These tasks will leverage leading-edge mathematical methodologies, improve the observation-to-diagnosis clinical process by efficiently handling patient-specific data, and support therapy planning, ultimately enabling a scalable translation to large population cohorts.

All information processing in nervous systems relies on spatial and temporal patterns of neural activity. While spatial patterns are dictated by neuroanatomy, the mechanisms that give rise to temporal activity patterns are diverse. They range from fast voltage dynamics of single neurons at one end of the spectrum to slow transcriptional and structural changes at the other, but the rules that shape signals at timescales in between milliseconds and minutes are poorly understood. The proposed research aims to uncover mechanisms of temporal information processing at these intermediate timescales, at which temporal patterns are thought to emerge from recurrently connected circuits. Detailed insight into the function of these circuits has been limited by the large number of circuit elements, by the lack of knowledge about their connectivity, and by the impracticability of recording from all circuit elements under naturalistic conditions. In Drosophila melanogaster, these limitations no longer apply. The comparatively low number of neurons, their well-mapped connectivity, and our ability to record and control their activities make mechanistic concepts testable. We will focus on three processes in the brain of Drosophila that unfold over three timescales ranging from milliseconds to minutes: 1) temporal filtering in the motion vision system, 2) sequential sampling of motion information in the lead-up to a perceptual judgement, and 3) temporal integration of distance during locomotion. Patch clamp experiments in the smallest of invertebrate neurons in vivo will allow us to record activity at the highest temporal resolution. We will combine this technique with behavioural, genetic, and imaging experiments to test the roles of individual neurons, their biophysical properties, and their synaptic connections in processing signals at intermediate timescales. The proposed experiments will further our understanding of motion vision, perceptual decision-making, and path integration.

PoCCardio is a large-scale project developing a point of care (PoC) device for cardiovascular disease (CVD) risk stratification in post-myocardial infraction (MI) patients. This project includes a large clinical trial with MI patients and collects clinical data, biomarker readings and outcome data. In order to fully analyze this dataset a holistic bioinformatics approach is necessary. PoCCardio-SB will employ a “systemic” approach, which considers a disease in the framework of a complex network emerging from the integration of information from various levels. This falls within the realm of Systems Bioinformatics (SB). SB harnesses the powerful methods of network science to integrate information across different levels/sources as well as extract useful features from complex biological networks. This is critical for advancing Predictive, Preventive and Personalized Medicine, shifting healthcare from delayed and untargeted treatments to precise, preventive strategies. The Bioinformatics Department at The Cyprus Institute of Neurology and Genetics will contribute by refining the diagnostic and prognostic signature for CVD using state of the art SB. Genomic and proteomic biomarker readings from PoCCardio will be utilized to develop integrated networks combining both sources. The networks will also incorporate prior knowledge, derived from public database. These patient-specific, “knowledge-based” networks will be utilized to train Machine Learning algorithms such as, Graph Convolutional Networks. Ultimately, a fine-tuned, decision-making, knowledge-based software will be implemented for predicting “extremely high risk” CVD patients, as well as their response to treatment. PoCCardio-SB will also develop a digital diagnostic protocol and demonstrate its usability at the PoC. This will include a user-friendly interface as well as the backend SB software. PoCCardio-SB will then integrate this diagnostic protocol with the POC device for improved diagnostic assessment.

Beside Bacteria and fungi, Archaea, the "third domain of life" have been described as a stable component of the human microbiome. Yet, only few archaea have so far been isolated from humans. The project aims to demonstrate that a fundamentally new product, derived from archaea, revolutionizing current standards of treatments for human health, can be developed from our initial breakthrough achievement, the cultivation of these archaea. The benefits of using archaea derived from human bodies, promise a sustainable, safe and beneficial treatment in the future.

Increasing demand for sophisticated clinical diagnostics makes current diagnostic capacities insufficient. A potential solution lies in semi-automatic systems speeding up the diagnosis process. Artificial intelligence (AI) and machine learning seem to be very promising approaches to the automation of diagnostic systems. However, most academic AI systems are opaque black boxes that cannot be easily understood, tested and certified. Also, academic AI solutions are often hard to reproduce, and their evaluation is insufficiently connected with clinical practice. This motivates MU and MMCI to team with two advanced partners (AP), MUG and TUB, and establish a BioMedAI infrastructure allowing close cooperation of computer science and clinical experts to develop explainable trustworthy AI solutions. Both AP possess rich experience with AI solutions for healthcare. Namely, processing large amounts of sensitive image and clinical data, interactive machine learning methods with a human-in-the-loop, and validating AI methods for healthcare. The main body of the BioMedAI project concentrates on training computer science researchers at MU and clinical experts at MMCI in the development of explainable AI methods based on high-quality medical data and validated in a clinical setting. Concretely, we propose organizing thematic workshops, virtual training with hands-on experience in developing explainable AI tools, and two summer schools. One will be oriented towards basic research in explainable AI methods for image and clinical data processing, and the other one towards the FAIR management of sensitive medical data. Furthermore, the BioMedAI project will also increase the visibility and presence of the explainable AI research in healthcare at MU and MMCI by training a PR manager responsible for presenting the research to various stakeholders, and by training the existing project management staff at MU and MMCI in writing grant applications for projects in EU and elsewhere.