Adaptive Digital Twin

The »Adaptive Digital Twin« project aims at embodying the central nervous system of healthy subjects and patients in the digital space. We study and develop co-adaptive models that enable simulation, statistical analysis and visual inspection of the brain using individually acquired functional and structural data. The combination of these different data sources and applications enables the assessment and modelling of anatomical changes, for example induced by tumor growth or surgical intervention, and the prediction of these effects on the structural anatomy, network and functionality of the individual central nervous system. The different data sources for these models are based on MRI data, neuropsychological testing and non-invasive brain stimulation. Based on these various data acquisitions, we develop subject-specific analytical methods for preoperative planning and assessment, which are tested and refined by the clinicians at the Department of Neurosurgery at Charité – Universitätsmedizin Berlin.

Brain Roads emerged from the Adaptive Digital Twin. This subproject is funded by the German-French DAAD. It connects students of graphic design and software engineering. Together, neuroscientists, psychologists and humanities scholars develop new possibilities of data visualization and interactive tools in relation to neuroscientific research.

The subproject »Mattering Uncertainties« grapples with the question of the interdependency between algorithmic uncertainty and human assertiveness. All structures and processes represented in the digital model are subject to statistical uncertainty. At the same time, the surgeons' planning tasks, such as deciding on the exact location of sectioning during tumor surgery, requires weighing the risks and benefits of different alternatives. Performing the ›right‹ cut requires much more than information processing. Rather it is a process of understanding a situation together with an intelligent device, highlighting the gaps of the infrastructure actually helps both sides, the human and the computer system. We aim to produce interactive data visualizations to help surgeons better relate to relevant information from the brain model, including its uncertainties, but at the same time, considering the individual perspective and need for explanation. For this, we are conducting participatory design studies regarding uncertainty as information to explore the modalities of perception during the decision-making process. Through interface design research, we will evaluate different visualization approaches to convey the types of uncertainty that we have identified as relevant in the surgeon's practice. In close cooperation with medical experts and interaction designers, this subproject aims to redefine the sensemaking process that takes place between the human and the machine. Our goal is to challenge the limitations that are commonly assumed about the capability of a machine to express its skills and share its doubts with the users and vice versa.

Prior to any surgical action in the brain, it is critical to map the functions in the neural material in order to preserve cognitive abilities as best we can. In the subproject »Active Symbolic Brain Materials«, we explore the structural features of the active neural matter in human subjects. We investigate the role played by those structures in cognitive activities such as language, symbols or thought. To this end, ›brain-inspired neural models‹ will be developed in order to mimic essential language processing functions in specific subjects, and on different patient populations. Simultaneously collaborating with the project Symbolic Material, we will simulate and compare the evolutionary developments in neural structures in humans and non-human primates to better understand their role in symbolic reasoning.