HET sistemų centras / HET Systems Centre
Transformer-based foundation learning for robust and data-efficient skin disease imagingItem type:Publication, research article[2026][S1][T007][31]; ;Alshaya, Hend ;Tehsin, SaraBouchelligua, WidedDiagnostics, 2026, vol. vol. 16, no. iss. 3, p. 1-31Background/Objectives: Accurate and reliable automated dermoscopic lesion classification remains challenging. This is due to pronounced dataset bias, limited expert-annotated data, and poor cross-dataset generalization of conventional supervised deep learning models. In clinical dermatology, these limitations restrict the deployment of data-driven diagnostic systems across diverse acquisition settings and patient populations. Methods: Motivated by these challenges, this study proposes a transformer-based, dermatology-specific foundation model. The model learns transferable visual representations from large collections of unlabeled dermoscopic images via self-supervised pretraining. It integrates large-scale dermatology-oriented self-supervised learning with a hierarchical vision transformer backbone. This enables effective capture of both fine-grained lesion textures and global morphological patterns. The evaluation is conducted across three publicly available dermoscopic datasets: ISIC 2018, HAM10000, and PH2. The study assesses in-dataset, cross-dataset, limited-label, ablation, and computational-efficiency settings. Results: The proposed approach achieves in-dataset classification accuracies of 94.87%, 97.32%, and 98.17% on ISIC 2018, HAM10000, and PH2, respectively. It outperforms strong transformer and hybrid baselines. Cross-dataset transfer experiments show consistent performance gains of 3.5-5.8% over supervised counterparts. This indicates improved robustness to domain shift. Furthermore, when fine-tuned with only 10% of the labeled training data, the model achieves performance comparable to fully supervised baselines. Conclusions: This highlights strong data efficiency. These results demonstrate that dermatology-specific foundation learning offers a principled and practical solution for robust dermoscopic lesion classification under realistic clinical constraints.
2Scopus© Citations 6 - research article[2026][S1][T007][33]
;Abdelbaki, Wiem ;Alshaya, Hend; ;Tehsin, Sara ;Said, SalwaBouchelligua, WidedBiomedicines., 2026, p. 1-33Background: Diagnosing CNS tumors through MRI is limited by significant variability in scanner hardware, acquisition protocols, and intensity characteristics at clinical centers, resulting in substantial domain shifts that lead to diminished reliability for automated models. Methods: We present a Domain-Adaptive MRI Learning Model (DA-MLM) consisting of an adversarially aligned hybrid 3D CNN-transformer encoder with contrastive regularization and covariance-based feature harmonization. Varying sequence MRI inputs (T1, T1ce, T2, and FLAIR) were inputted to multi-scale convolutional layers followed by global self-attention to effectively capture localized tumor structure and long-range spatial context, with domain adaptation that harmonizes feature distribution across datasets. Results: On the BraTS 2020 dataset, we found DA-MLM achieved 94.8% accuracy, 93.6% macro-F1, and 96.2% AUC, improving upon previously established benchmarks by 2-4%. DA-MLM also attained Dice score segmentation of 93.1% (WT), 91.4% (TC), and 89.5% (ET), improving upon 2-3.5% for CNN and transformer methods. On the REMBRANDT dataset, DA-MLM achieved 92.3% accuracy with segmentation improvements of 3-7% over existing U-Net and expert annotations. Robustness testing indicated 40-60% less degradation under noise, contrast shift, and motion artifacts, and synthetic shifts in scanner location showed negligible performance impairment (<0.06). Cross-domain evaluation also demonstrated 5-11% less degradation than existing methods. Conclusions: In summary, DA-MLM demonstrates improved accuracy, segmentation fidelity, and robustness to perturbations, as well as strong cross-domain generalization indicating the suitability for deployment in multicenter MRI applications where variation in imaging performance is unavoidable.
5 19Scopus© Citations 3 Addressing the Cybersecurity Skills Shortage in Lithuania: Policy Insights from the United KingdomItem type:Publication, research article[2026][S1b][T007][25]; ; ; ; Journal of Cybersecurity and Privacy: Building Community of Good Practice in Cybersecurity., 2026, p. 1-25Cybersecurity has become a critical challenge to policy as cyber threats continue to increase in frequency, sophistication, and societal impact, exposing the growing vulnerability of the critical infrastructure supporting vital societal functions. Globally, these risks are heightened by a persistent shortage of skilled cybersecurity professionals, which, in Europe, threatens the effective implementation of the Union’s Network and Information Security Directive 2 (NIS2) concerned with the enhancement and harmonization of the cybersecurity level across Member States, notably in terms of their critical infrastructure and involved entities. This article examines the cybersecurity skills landscape across the European Union (EU), with a specific focus on Lithuania, using the United Kingdom (UK) as a strategic benchmark subject. Adopting a comparative case study approach, the study explores and discusses governance arrangements, education and training pathways, labour-market dynamics, and quality-assurance mechanisms shaping cybersecurity workforce development. Technical, organisational, and transversal skills required to prepare an effective cybersecurity workforce in a rapidly evolving labour landscape are also discussed. Findings reveal that Lithuania faces an acute shortage of advanced practitioners and limited alignment between education provision, labour-market needs, and regulatory requirements. In response, the article proposes policy-informed strategies adapted from the UK’s structured and professionalised cybersecurity skills model, explicitly mapped to NIS2 workforce and capability requirements. Identified strategies emphasise the need of coordinated action across schools, higher education institutions, government, industry, and the wider community. Potential enablers and constraints for the operationalization of the identified strategies are further analysed and discussed. The study aims to contribute to ongoing policy debates by demonstrating how a strategic context-sensitive selection and adaptation of key components in established skills frameworks can support the development of a sustainable national cybersecurity skills ecosystem and enhance long-term digital resilience, not only in Lithuania but also in other Member States across the EU.
9 114Scopus© Citations 1 Cyber Security at Universities: Comparison of Australia, Lithuania and UkraineItem type:Publication, research article[2026][P1a2][S001][7] ;Warren, Matthew; The Proceedings of the 25th European Conference on Cyber Warfare & Security (ECCWS 2026)., p. 831-837Universities are common across all countries and perform a key function in terms of research and knowledge development as well as the development of the next generation of professionals. In this study the focus is the university sector of Australia, Lithuania and Ukraine. Each country manages their university sector differently from a security perspective. Australian universities operate under critical infrastructure regulations and foreign interference guidelines. Lithuanian institutions benefit from EU NIS2 directives and national collective defense systems. Ukrainian universities, facing intense cyber attacks since 2014 and have developed exceptional resilience through distributed infrastructure and international cooperation. This paper examines the distinct cyber security postures of the Australia, Lithuania and Ukrainian university sector.
11 55 Hybrid‑Threat Intelligence: A Critical Review of Semantic Integration Challenges and the Role of the HIPSTer Ontological FrameworkItem type:Publication, research article[2026][S4][T007,S001][34]; ; ; ; Journal of Intelligent Communication., 2026, p. 140-173Contemporary hybrid threats employ coordinated campaigns across information, cyber, and physical domains, maintaining plausible deniability while exploiting institutional vulnerabilities. This review conducts a scoping analysis following the PRISMA ScR framework (Preferred Reporting Items for Systematic Reviews and Meta-Analyses—Extension for Scoping Reviews) to evaluate Open Source Intelligence (OSINT), Social Media Intelligence (SOCMINT), and Natural Language Processing (NLP) capabilities relevant to hybrid threat detection. We systematically assess these technologies against a 12-point operational requirements framework derived from documented Russian and Chinese military OSINT methodologies and influence operation tradecraft. The analysis incorporates high-Technology Readiness Level (TRL) European initiatives to ground capability assessments in operational experience. While individual analytical disciplines are technically advanced, current defensive systems remain siloed and lack the cross-domain reasoning necessary to correlate technical cyber indicators with coordinated narrative manipulation. Requirements for cross-platform correlation and adversarial adaptation show only prototype stage coverage. Our findings reveal a persistent “semantic gap”: defensive systems collect extensive data but lack integrated semantic reasoning across domains and languages. To address this, we examine ontology-based approaches as architectural solutions, positioning the ‘Hybrid Information Psychological Societal Threats handling system’ (HIPSTer) framework as an illustrative case. HIPSTer specifically targets the multilingual nature of hybrid threats—particularly in Russian and Chinese contexts—achieving TRL-4 validation through high-efficiency semantic vectors and formal reasoning across diverse language benchmarks. Finally, the review analyzes how European regulations—including the General Data Protection Regulation (GDPR), AI Act, and Network and Information Systems Directive 2 (NIS2)—shape operational architectures through compliance by design imperatives. We conclude by outlining a prioritized research agenda to advance European hybrid threat detection toward operational maturity.
14 92