Abstract
This paper addresses the development of a novel zero-shot learning method for remote marine litter hyperspectral imaging data classification. The work consisted of using an airborne acquired marine litter hyperspectral imaging dataset that contains data about different plastic targets and other materials and assessing the viability of detecting and classifying plastic materials without knowing their exact spectral response in an unsupervised manner. The classification of the marine litter samples was divided into known and unknown classes, i.e., classes that were hidden from the dataset during the training phase. The obtained results show a marine litter automated detection for all the classes, including (in the worst case of an unknown class) a precision rate over 56% and an overall accuracy of 98.71%.

Abstract
Linked Data is used in various fields as a new way of structuring and connecting data. Cultural heritage institutions have been using linked data to improve archival descriptions and promote findability. The required detail in manual descriptions of cultural heritage objects can be taxing and time-consuming. Given this, in EPISA, a research project on this topic, we propose to use the contents of the digital representations associated with the objects to assist archivists in their description tasks. More specifically, to extract information from the digital representations useful for an initial ontology population that should be validated or edited by the archivist. We apply optical character recognition in an initial stage to convert the digital representation to a machine-readable format. We then use ontology-oriented programming to identify and instantiate ontology concepts using neural networks and contextual embeddings.

Abstract
Large European research consortia in the health sciences face challenges regarding the governance of personal data collected, generated and/or shared during their collective research. A controller in the sense of the GDPR is the entity which decides about purposes and means of the data processing. Case law of the Court of Justice of the European Union (CJEU) and Guidelines of the European Data Protection Board (EDPB) indicate that all partners in the consortium would be joint controllers. This paper summarises the case law, the Guidelines and literature on joint controllership, gives a brief account of a webinar organised on the issue by Lygature and the MLC Foundation. Participants at the webinar agreed in large majority that it would be extreme if all partners in the consortium would become joint controllers. There was less agreement how to disentangle partners who are controllers of a study from those who are not. In order to disentangle responsibilities, we propose a funnel model with consecutive steps acting as sieves in the funnel. It differentiates between two types of partners: all partners who are involved in shaping the project as a whole versus those specific partners who are more closely involved in a sub-study following from the DoA or the use of the data Platform. If the role of the partner would be comparable to that of an outside advisor, that partner would not be a data controller even though the partner is part of the consortium. We propose further nuances for the disentanglement which takes place in various steps. Uncertainty about formal controllership under the GDPR can stifle collaboration in consortia due to concerns over (shared) responsibility and liability. Data subjects’ ability to exercise their right can also be affected by this. The funnel model proposes a way out of this conundrum.

Abstract
Dementia is a broad term for a large number of conditions, and it is often associated with Alzheimer's disease. A reliable diagnosis of this disease, especially in the early stages, may prevent further complications. As such, machine learning algorithms can be applied in order to validate and correctly classify cases of dementia or non dementia in adults, assisting physicians in the diagnosis and management of this clinical condition. In this study, a dataset containing magnetic resonance imaging comparisons of demented/non demented adults was used to conduct a Data Mining process, following the Cross Industry Standard Process for Data Mining methodology, with the main goal of classifying instances of dementia. Different machine learning algorithms were applied during this process, more specifically Support Vector Machines, Decision Trees, Logistic Regression, Neural Networks, Naive Bayes and Random Forest. The maximum accuracy of 95.41% was achieved with the Naive Bayes algorithm using Split Validation.

Abstract
In the recent decade, marine origin products have been growingly studied as building blocks complying with the constant demand of the biomedical sector regarding the development of new devices for Tissue Engineering and Regenerative Medicine (TERM). In this work, several combinations of marine collagen-chitosan-fucoidan hydrogel were formed using a newly developed eco-friendly compressive and absorption methodology to produce hydrogels (CAMPH), which consists of compacting the biopolymers solution while removing the excess of water. The hydrogel formulations were prepared by blending solutions of 5% collagen from jellyfish and/or 3% collagen from blue shark skin, with solutions of 3% chitosan from squid pens and solutions of 10% fucoidan from brown algae, at different ratios. The biopolymer physico-chemical characterization comprised Amino Acid analysis, ATR-FTIR, CD, SDS-PAGE, ICP, XRD, and the results suggested the shark/jellyfish collagen(s) conserved the triple helical structure and had similarities with type I and type II collagen, respectively. The studied collagens also contain a denaturation temperature of around 30-32 degrees C and a molecular weight between 120 and 125 kDa. Additionally, the hydrogel properties were determined by rheology, water uptake ability, degradation rate, and SEM, and the results showed that all formulations had interesting mechanical (strong viscoelastic character) and structural stability properties, with a significant positive highlight in the formulation of H-3 (blending all biopolymers, i.e., 5% collagen from jellyfish, 3% collagen from skin shark, 3% chitosan and 10% of fucoidan) in the degradation test, that shows a mass loss around 18% over the 30 days, while the H-1 and H-2, present a mass loss of around 35% and 44%, respectively. Additionally, the in vitro cellular assessments using chondrocyte cells (ATDC5) in encapsulated state revealed, for all hydrogel formulations, a non-cytotoxic behavior. Furthermore, Live/Dead assay and Phalloidin/DAPI staining, to assess the cytoskeletal organization, proved that the hydrogels can provide a suitable microenvironment for cell adhesion, viability, and proliferation, after being encapsulated. Overall, the results show that all marine collagen (jellyfish/shark)-chitosan-fucoidan hydrogel formulations provide a good structural architecture and microenvironment, highlighting the H-3 biomaterial due to containing more polymers in their composition, making it suitable for biomedical articular cartilage therapies.

Abstract