Brain-Body factoRs mediating altEred Bodily representations in mUltiple pathological conditionS

Project title: Brain-Body factoRs mediating altEred Bodily representations in mUltiple pathological conditionS

Agreement number: NeuronC2/IV/61/BB-REBUS/2025

Program name: Era-Net Neuron

Project implementation period: 01.05.2025 – 30.04.2028

Project value: 1 222 453,13 PLN

Funds granted for Lublin University of Technology: 1 222 453,13 PLN

Principal Investigator: dr hab. inż. Kamil Jonak, profesor uczelni

Abstract: The BB-REBUS project focuses on computational modelling of disrupted brain–body communication and uses advanced data-analysis algorithms to explain body perception disturbances in patients after stroke, with spinal cord injury and with multiple sclerosis. The key contribution of the team from Lublin University of Technology is to design a complete EEG signal-processing pipeline – from acquisition algorithms, through filtering, segmentation and artefact removal, to automatic extraction of clinically relevant features. To this end, dedicated programming procedures will be developed in the Matlab environment, including modules that use the fast Fourier transform to move from the time domain to the frequency domain and to compute power spectra in the delta, theta, alpha, beta and gamma bands. On this basis, algorithms will be created to calculate interhemispheric asymmetry indices and intrahemispheric fronto-parietal activity profiles, which will make it possible to computationally characterise the diverse neuronal states of the studied patients. In parallel, the team will develop software for assessing functional brain connectivity using the phase lag index (PLI) and the Hilbert transform, enabling precise estimation of instantaneous phase differences between signals from multiple electrodes. The results of these computations will feed into another algorithmic module, in which, using graph-theoretical methods and the minimum spanning tree (MST) algorithm, the structure of biological neuronal networks will be reconstructed for each patient. The entire computational pipeline will be implemented and tested on a newly established EEG workstation and then transferred – in the form of ready-to-use procedures and scripts – to clinical centres in Switzerland, Germany and Italy. Data recorded by the international partners will be standardised and automatically processed by the developed algorithms to ensure comparability of results and high quality of digital records used for further modelling. In the second main stage of the project, the team from Lublin University of Technology will build a centralised database to which encoded EEG results, images of brain lesions, clinical and demographic data as well as outcomes from questionnaires and experimental tasks will be uploaded. Based on this repository, a set of scripts will be prepared for preprocessing, handling missing data, balancing classes and converting many variables into a binary form describing the presence or absence of body perception (BP) disturbances and related deficits. Next, dimensionality reduction algorithms such as principal component analysis (PCA) and t-SNE will be applied, allowing the key feature combinations to be extracted computationally and the cluster structure in the multidimensional data space to be visualised. On this foundation, the team will develop and implement various machine learning models, including support vector machines (SVM), decision trees and boosting methods, used to classify patients with and without BP disturbances. These models will first be trained on data from post-stroke patients and then – in close collaboration with the other partners – transferred and validated on datasets concerning multiple sclerosis and spinal cord injuries. As part of the computational work, systematic cross-validation and sensitivity analyses of the models are planned, involving the gradual inclusion and exclusion of demographic, clinical, behavioural variables and EEG biomarkers to assess their impact on classification accuracy. The results of these computations will be used to identify algorithmic patterns of brain–body relationships that best predict the development of body perception disturbances, the course of rehabilitation and the potential for recovery. An important technological output of the project will be a set of open-source BB-REBUS algorithms, including modules for EEG signal processing, neuronal network analysis, dimensionality reduction and machine learning. This software will be made available to the international research community, enabling replication of results, further development of the tools and their adaptation to other clinical populations with sensorimotor disorders. All the designed IT solutions will be tightly integrated with the needs of clinical practice so that physicians and therapists can receive clear reports automatically generated on the basis of complex algorithmic analyses. The project also envisages the preparation of training materials and courses based on the developed digital tools, which will facilitate the implementation of algorithmic methods for assessing and monitoring BP disturbances in routine patient care. As a result, BB-REBUS will become a comprehensive, highly computerised research–clinical–educational platform centred on novel algorithms for EEG time-series processing and machine learning that support personalised neurorehabilitation of patients with body perception disturbances.