Computer Science Department

Master in Computer Science · 2025

Read Computer Science Department by Master in Computer Science (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-13846.txt.

Introduction

Antonio Strippoli designs a friendlier interface for a medical image analysis tool, in a master's thesis from 2025. The tool, VoxLogicA, checks images against spatial logic rules, with a focus on brain scans. Its power was hard for doctors and researchers to use, since the controls were complex.

The thesis sets three goals: design a modern web interface that is easy to use, build it with current tools such as Svelte and Niivue, and test it. User studies and real case analyses checked how well the design worked in practice. The aim is to make spatial model checking a practical aid in real clinical settings, not just a research demo.

Catalog record

CREATORMaster in Computer Science
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE102 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-13846

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Computer Science Department Master in Computer Science Abstract ”Life is short, craft long, opportunity fleeting, experimentations perilous, and judgment difficult.” Hippocrates 1 Introduction The role of spatial logic in medical image analysis has emerged as a promising approach through the VoxLogicA project, offering ways to describe and analyze anatomical structures in ways that align with domain-specific medical reasoning [ 5 , 13 ] . Recent advancements, such as the application of spatial model checking for nevus segmentation, demonstrate its feasibility in achieving robust results comparable to expert annotations while maintaining computational efficiency and explainability [ 3 ] . This thesis work emerges from three complementary research initiatives: • THE (Tuscany Health Ecosystem) [ 52 ] focuses on developing digital health ecosystems for a plethora of applications – among which personalized healthcare and brain health. THE aims at preparation for technological transfer, and has a specific action devoted to advancement of the technological nature of VoxLogicA; • STENDHAL (Spatio-Temporal Enhancement of Neural nets for Deeply Hierarchical Automatised Logic), a PRIN project [ 30 ] , is a foundational and multi-disciplinary project devoted to enriching spatio-temporal analysis via neural-symbolic integration, combining neural networks with formal spatial model checking to enable expert-driven verification of medical protocols and safety properties in longitudinal healthcare studies; • FM4HD (Formal Methods For the Healthcare Domain based on spatial information) [ 51 ] focussed less on the inter-disciplinary issues of STENDHAL and more specifically on the challenges offered by the enrichment of the expressiveness of the spatio-temporal logics and its implementation. Our work bridges a critical gap common to all three projects. Spatial model checking has shown theoretical promise through tools like VoxLogicA, and has demonstrated potential practical impact via a number of successful case studies. However, clinical adoption requires addressing fundamental human-computer interaction challenges. Thus, spatial model checking’s clinical value can only be fully realized through simultaneous advances in formal methods and interaction design. This creates a unique research tension: how can we preserve the formal guarantees of spatial logic while creating interfaces that align with both computer science researchers’ and medical professionals’ cognitive workflows? We want to demonstrate that, in the field of medical imaging, formal methods: • Can be operationalized through careful interface design; • Enable new human-computer collaboration patterns in diagnostic workflows; • Require usability studies to complete the research-to-practice cycle. 1.1 Research Objectives This thesis addresses the need for a user-friendly interface to VoxLogicA, a spatial model checking tool for image analysis. The primary objectives include: • Conducting a comprehensive design study for a modern web interface for VoxLogicA, aimed at enhancing accessibility for medical professionals and researchers; • Implementing the interface using modern web development technologies and best practices; • Conducting empirical validation through structured user studies and real-world case analyses to assess interface efficacy. 1.2 Thesis Structure This thesis is organized as follows: Chapter 2 provides the multidisciplinary context of the research through four integrated pillars: spatial model checking’s mathematical framework, neuroimaging analysis challenges, UI design principles for medical interfaces, and modern web architectures. The synthesis operationalizes theoretical insights through comparative technology assessments, justifying subsequent implementation choices. Chapter 3 presents VoxLogicA UI as a modern web-based interface for spatial model checking in neuroimaging, designed to bridge the gap between VoxLogicA’s advanced analytical capabilities and clinical/research usability requirements. Motivated by a critical analysis of legacy neuroimaging tools - revealing persistent usability challenges - the solution is designed and implemented using the latest web development technologies (e.g. Svelte and SkeletonUI) and best practices (e.g. MVVM architecture, RESTful API design, Docker deployment). The project architecture follows a layered pattern, with each layer having specific responsibilities and communicating only with adjacent layers. Chapter 4 details the concrete realization of the VoxLogicA UI system through an integrated examination of implementation strategies across the full technology stack. The development environment lays the groundwork for the project, featuring docker containerization and continuous integration. The backend exposes a hierarchical security-conscious REST API structured around core medical imaging concepts (datasets, cases, layers), with a particular highlight being a novel workspace cloning mechanism addressing clinical collaboration needs through immutable snapshots that preserve data sovereignty while enabling derivative analyses. The frontend exposition progresses through a vertical slice of the visualization pipeline, from NiiVue WebGL integration to reactive layer management components, culminating in an analysis panel to write and run spatial logic queries. Finally, adaptive distribution implementations are discussed, accommodating heterogeneous clinical and research infrastructures through flexible distribution pathways spanning local containerized installations and cloud-native deployments. Chapter 5 transitions from technical implementation to empirical validation via a multi-layered validation strategy combining rigorous usability testing and real-world exposure initiatives. The analyses were conducted through controlled user studies employing mixed-methods research design, combining quantitative metrics like System Usability Scale (SUS) scores with qualitative observational data. Two parallel studies - targeting computer science researchers and medical physics students - reveal divergent usability patterns that reflect the system’s dual role as both research environment and clinical analysis tool. Public exposure of VoxLogicA UI is then described, ranging from engagement with the NiiVue team for future technical collaborations to clinical engagement efforts at Bari Hospital and Bari University with a radiologist, a general practitioner and some students. Chapter 6 synthesizes the project’s contributions through a critical reflection on achieved outcomes and remaining challenges. 2 Background Knowledge 2.1 Formal Methods In the context of modern software systems, where complexity continues to grow and reliability becomes increasingly critical, formal methods emerge as a fundamental approach to ensure system correctness [ 43 , 29 ] . At their core, formal methods represent the application of mathematics to the modelling and analysis of Information and Communication Technology (ICT) systems, providing a rigorous framework for verification and validation . Formal methods enable developers and researchers to verify that a model of a system design or implementation satisfies properties derived from its specification. When a model fails to meet any of these properties, a defect is identified; conversely, a model is deemed correct when it fulfils all the properties of interest. Formal methods enable the early integration of verification in the design process, providing systematic verification techniques, and significantly reducing the time required for system validation. Their importance is particularly evident in safety-critical systems, where they are considered a highly recommended verification approach. Notable examples from industry underscore their value: formal verification techniques could have prevented well-known failures in systems such as the Ariane-5 rocket , Mars Pathfinder , and Intel’s Pentium II processor . Figure 2.1 : Ariane-5 rocket’s catastrophic failure during its maiden flight in 1996, costing over $370 million. The disaster was caused by a software error in the Inertial Reference System (SRI), which led to an operand error and subsequent system shutdown. Formal methods could have prevented this failure by identifying and addressing the software design flaws and exception handling issues before launch. [ 16 ] 2.1.1 Model Checking: an overview Model-based verification techniques employ mathematical models to precisely and unambiguously describe potential system behaviors. This rigorous approach often uncovers incompleteness, ambiguities, and inconsistencies inherent in informal system specifications. Among these techniques, model checking stands out as a formal verification method that systematically inspects all possible states of a system model to ascertain compliance with a given set of properties

Read the complete text