Conference Agenda

Intracranial aneurysms are strongly influenced by local hemodynamics, motivat-ing the combined use of computational fluid dynamics (CFD) and experimental validation techniques such as particle image velocimetry (PIV). However, PIV velocity measurements in aneurysm models are often affected by a wide dynamic range and reduced accuracy close to the boundaries, limiting direct comparison with numerical predictions. In this paper, a new post-processing methodology is proposed to enhance the reliability of PIV-derived velocity fields and enable con-sistent validation of CFD results in idealized single-vessel aneurysm geometry. A rigid transparent PDMS phantom of a giant saccular aneurysm was realized using a lost core casting technique and tested in a steady-flow loop with refractive-index-matched working fluid. PIV acquisitions were performed using multiple inter-frame time intervals to capture both high-velocity regions in the parent ves-sel and low-velocity recirculation inside the sac. A novel iterative filtering strate-gy based on skewness reduction and percentile trimming was developed to re-move outliers, followed by a spatial stitching algorithm that com-bines multiple time interval acquisition datasets into a single composite velocity magnitude map. The resulting experimental velocity fields showed improved consistency across the full velocity range and enabled meaningful comparison with RANS CFD simulations, supporting the proposed framework as a robust benchmark method-ology for aneurysm hemodynamic validation.

Medical image generation can support the development of robust deep learning models by providing controlled synthetic data when real datasets are limited, imbalanced, or difficult to share. This study presents a preliminary pipeline for conditional chest X-ray generation using a class-conditioned Latent Diffusion Model. The model was trained on chest radiographs labelled as No Finding, meaning that no pathological finding is reported in the dataset metadata, and conditioned on radiographic view position, namely PA and AP projection. Images were encoded into a latent space using a frozen Stable Diffusion VAE, while a MONAI Diffusion Model UNet was trained to denoise latent representations conditioned on the target view class. Several configurations were evaluated by varying training epochs, conditioning tokens, guidance scale, inference steps, and AP oversampling. Generated images were assessed using FID for realism, MS-SSIM for diversity, and auxiliary ResNet18 and DenseNet121 classifiers for conditioning fidelity. The best results showed a substantial improvement in FID, especially for AP images, while classifier evaluation confirmed that generated samples were consistently recognized as intended PA or AP class. The results suggest that class-conditioned latent diffusion is a promising approach for controlled chest radiograph synthesis, although future work should include chest-specific autoencoders and anatomy-aware evaluation.

Defect detection on newly produced textiles is traditionally performed through human Visual Inspection, a process inherently prone to fatigue and subjective bias. This paper presents an Assisted Visual Inspection (AVI) prototype designed to seamlessly integrate an intelligent computer vision module into existing industrial fabric inspection lines.

A down-scaled operational mockup was engineered to provide a robust physical platform for mounting the module's components and performing real-time tests. The core of the vision module utilises Meta's DINOv3 model, implementing a self-supervised, few-shot learning framework that exploits the spatial periodicity of uniform textile patterns. This approach dynamically creates a "gold standard" reference of defect-free fabric, enabling the detection of unseen anomalies without the need for extensive, pre-labelled training datasets.

Preliminary results demonstrate that the module effectively flags anomalies in real-time, halting the line to project an Augmented Reality (AR) defect mask directly onto the fabric, thereby significantly boosting the reliability of human-in-the-loop inspection.

Lattice structures produced by additive manufacturing require reliable di-mensional verification, because their small features and internal porosity lim-it the use of conventional metrological instruments. This study evaluates an optical approach for estimating pore dimensions in stainless steel 316L lattice structures manufactured by binder jetting. Simple cubic lattices with a nomi-nal cell size of 500 µm and three strut thicknesses (150, 200, and 250 µm), corresponding to nominal pore sizes of 350, 300, and 250 µm, were ana-lyzed. Images were acquired on three mutually orthogonal planes using an Olympus DSX 1000 microscope and an OGP SmartScope 300 system. A us-er-defined MATLAB algorithm was developed to segment each pore, esti-mate its projected area, and calculate the corresponding pore size. The results were compared statistically to assess the influence of projection plane, nomi-nal pore size, and acquisition system. Both systems detected deviations from the nominal pore size, generally ranging from -20% to +5%, with higher ac-curacy on the YZ plane. The comparison indicates that optical acquisition combined with image analysis can support rapid and low-cost preliminary dimensional characterization of lattice structures, although further validation against a reference method, such as X-ray computed tomography, is re-quired.

Model-Based Definition (MBD) supports the transition from drawing-based documentation to model-centric product definition by associating Product and Manufacturing Information (PMI) with the 3D CAD model. Although MBD is increasingly considered an enabler of digital product definition and Model-Based Enterprise (MBE), its implementation remains uneven across standards, software environments, and industrial practices. This paper maps the current state of the art in MBD, focusing on how the concept is defined, standardized, and applied in the literature. The analysis shows that most con-tributions address downstream applications, specially manufacturing and inspection, while the design phase, where functional requirements are translated into geometric specifications and PMI, remains less supported. This gap highlights the need for shared terminology, interoperable data formats, and reference frameworks capable of supporting a continuous and traced information flow from early design stages to verification.