Medical imaging techniques for the analysis and measurement of composite materials

Text Complet
The purpose of this thesis is the research and development of methods, algorithms and tools for the analysis of composite materials. In order to improve their properties, one must employ specific techniques in order to gain insight of the structure of their internal reinforcement fibres. While these goals can be tackled in multiple ways, a common and established way is the employment of non-destructive means such as the micro-CT (micro-computed tomography). Due to its technical similarities with medical CT, first, we analyse the most relevant features across different state-of-the-art open-source medical imaging software using a novel hierarchical evaluation framework. From this evaluation it has been observed that many of the 3D-based visualization features can be used for the analysis of composites reinforcement fibres. However, the measurement of their properties requires specialized methods to reconstruct them in order to quantify their orientations, curvatures, waviness, or other aspects. For this reason, we propose applying the following two methods respectively: (a) an efficient and uniformly behaving streamline-based micro-CT fibre tracking algorithm using volume-wise structure tensor and signal processing techniques; and (b) a frequency-limited waviness and curvature measurement algorithm for noisy and irregularly sampled composite fibre trackings. The proposed methods have been implemented and integrated in the Starviewer platform. As these must remain understandable, explainable and controllable by an expert user, black-box approaches or the ones requiring training datasets have been discarded. Despite the differences between the volumetric acquisitions and polygonal fibre reconstructions, the proposed methods consider the dataset as a valuable signal to preserve. It is progressively processed and filtered according to the signal processing principles in order to obtain a uniform 3D behaviour. The methods properties have been assessed with both real and synthetic datasets; and due to the complex waveform shapes required for their testing, some datasets were produced using a novel methodology based on Gaussian splatters. All methods proposed in this thesis and their corresponding articles have been mathematically defined and accompanied by a large amount of supporting material files in order to ensure their reproducibility. ​
​L'accés als continguts d'aquesta tesi queda condicionat a l'acceptació de les condicions d'ús establertes per la següent llicència Creative Commons: