No description available.

No description available.

Three different void content calculation techniques using optical microscopy werecompared in multiple-user trials. The three methods studied comprised of a selection,thresholding, and semi-automatic machine learning method. The techniques wereapplied to micrographs of three carbon fiber-epoxy composite plates manufacturedin-house, where one plate had reduced void content by means of debulking priorto curing. The users performed the techniques on the sets of micrographs and thestandard deviation between the users void content results were measured.The advantages of the three methods were discussed and their practical applications wereproposed. The trials showed agreement between users on what are voids and not as well asshowing that uncertainties in void content are specimen-specific and not attributed todifferent users or methods applied. All three methods showed satisfying precision incalculating void content compared to void content quality levels provided by literature.It was found that thresholding, which is the current standard method of void contentcalculation using microscopy, inhabits an unscientific bias which compromises the legitimacyof the method. The study formulates a manual selection-based method usingedge-detection selection tools intended to benchmark void content in images, as wellas proposing a route to the automation of void content analysis using microscopy. / Tre olika beräkningstekniker för kavitetshalter med hjälp av mikroskopi jämfördes genom fleranvändar-tester. De tre metoderna innefattade en selektions-metod, tröskelvärdesmetod, och en övervakad maskininlärningsmetod. Metoderna applicerades på mikrografer av tre kolfiber-epoxi kompositplattor tillverkade internt, varav en platta hade reducerad kavitetshalt genom en avbulkningsprocess innan härdning. Användarna genomförde metoderna på mikrograferna och standardavvikelsen mellan användarnas resulterande kavitetshalter mättes. För- och nackdelarna hos de tre metoderna diskuterades och deras praktiska applikationer föreslogs. Testerna visade en överensstämmelse mellan användare om vad som omfattar kaviteter och inte, samt en påvisning på att osäkerheter kring kavitetshalter är provbitberoende och inte användar- eller metodberoende. Alla tre metoder uppvisade en tillfredsställande precision i kavitethaltsberäkning jämfört med kvalitetsnivåer av kavitethalter erhållna från litteratur. Det konstaterades att tröskelvärdesmetoden, vilket är nuvarande standardmetoden för kavitethaltsberäkning med mikroskopi, innehar en bias som sätter validiteten av metoden i fråga. Studien formulerar även en manuell selektions-metod som använder selektions-verktyg för randdetektering, ämnad för att hitta referensvärden för kavitetshalter. Förslag ges även kring tillvägagångssättet till att uppnå automatiserade metoder för kavitethaltsberäkning.

A key challenge in the imaging of spinal cord injury (SCI) patients is the ability to accurately determine structural or functional abnormality as well as level and severity of injury. Over the years a substantial number of studies have addressed this issue, however most of them utilized qualitative analysis of the acquired imaging data. Quantitative analysis of patients with SCI is an important issue in both diagnostic and treatment planning. Hence in this work new multispectral magnetic resonance (MR) image based approaches were developed for high-throughput extraction of quantitative features from pediatric spinal cord MR images and subsequent analysis using decision support algorithms. This may potentially improve diagnostic, prognostic, and predictive accuracy between typically developing (TD) pediatric spinal cord subjects and patients with SCI. The technique extracts information from both axial structural MRI images (such as T2-weighted gradient echo images) and functional MRI images (such as diffusion tensor images). The extracted data contains first order statistics (diffusion tensor tractography and histogram based texture descriptors), second order (co-occurrence indices) and high order (wavelet primitives) statistics. MRI data from total of 43 subjects that includes 23 healthy TD subjects with the age range of 6-16 (11.94±3.26 (mean ±standard deviation)) who had no evidence of SCI or pathology and 20 SCI subjects with the age range of 7-16 (11.28±3.00 (mean ±standard deviation)) were recruited and scanned using 3.0T Siemens Verio MR scanner. Standard 4-channel neck matrix and 8-channel spine array RF coils were used for data collection. After data collection various post processing methods were used to improve the data quality. A novel ghost artifact suppression technique was implemented and tested. Initially, 168 quantitative measures of multi-spectral images (functional and structural) were calculated by using regions of interest (ROIs) manually drawn on the whole cord along the entire spinal cord being anatomically localized by an independent board certified neuroradiologist. These measures were then statistically compared between TD and SCI groups using standard least squared linear regression model based on restricted maximum likelihood (REML) method. Statistically, significant changes have been shown in 44 features: 30 features obtained from functional images and 14 features selected from structural images. Also, it has been shown that the quantitative measures of the spinal cord in DTI and T2W-GRE images above and below injury level were altered significantly. Finally, tractography measures were also obtained on a subset of the patients to demonstrate quantitative analysis of the extracted white matter structures. Overall the results show that the proposed techniques may have potential to be used as surrogate biomarkers for detection of the injured spinal cord. These measures enable us to quantify the functional and structural plasticity in chronic SCI and consequently has the potential to improve our understanding of damage and recovery in diseased states of the spinal cord. / Bioengineering

Bioindicators of the nanoscale structural quality of bone were investigated using ion milling and transmission electron microscopy of osteoporotic bone from human female donors. / Bone is a complex hierarchical biomaterial constantly undergoing remodeling events initiated by cell signaling and fulfilled by migratory bone cells. In osteoporosis, a multitude of signaling factors cause bone resorption to proceed quicker than bone reformation, resulting in a lower bone mineral density (BMD) and porosity as seen by thinning of the cortex and trabeculae. However, the structural motifs of these altered regions of the skeleton have not been understood on the nanoscale. In this thesis, transmission electron microscopy (TEM) was used with an image analysis technique termed nanomorphometry, developed to enable the measurement of nanoscale structural features in human bone. Several nanoscale bone quality bioindicators relevant to the collagen fibrils and bone mineral (mineral lamellae, ML) components were defined and tested (collagen fibril diameter, interfibrillar spacing, ML thickness & ML stack thickness) among two donor cohorts of post-menopausal osteoporotic female patients and age- and sex-matched controls. In one cohort, the anatomical region investigated was the intertrochanteric crest of the femur, while in the second, the femoral neck was studied. The bone sections were prepared using an ion milling workflow yielding electron-transparent views of the bone ultrastructure. Blinded image analysis of the ultrastructure revealed that in both cohorts, the thickness of the MLs was significantly larger in osteoporotic samples versus their controls. In the former cohort, it was found that anti-resorptive drug use in the treated group did not return the ML thickness back to control levels. In the latter cohort, the ML thickness correlated more closely with the proximal femur bone mineral density (BMD) than the age of the patient. These findings suggest that the morphology of the nanoscale mineral phase is affected by osteoporosis, an effect indirectly observed by other techniques, and warrants further exploration into the implications of this effect on bone quality, fragility and strength. / Thesis / Master of Applied Science (MASc) / Human bone is a biomaterial with many levels of organization from the macroscale down to the nanoscale. The material consists of roughly 30 weight % organic components (collagen, non-collagenous proteins) and 67 weight % inorganic components (calcium phosphate minerals) deposited by bone cells. Osteoporosis is a bone disease commonly associated with increased bone porosity and bone fragility. In this study, the effect of osteoporosis on the nanoscale structure of bone was directly imaged and investigated using transmission electron microscopy (TEM). Two advanced ion milling techniques (broad beam and focused ion beam) were used to thin the bone specimens for TEM. Bioindicators relating to the structure and size of collagen and mineral components in osteoporotic versus control bone were quantified in an unbiased image analysis workflow. Findings indicated an increase in the thickness of poly-crystalline bone mineral lamellae in the nanoscale structure of human osteoporotic bone from two human donor cohorts.

No / Clinical morphological analysis of histopathological specimens is a successful manner for diagnosing benign and malignant diseases. Analysis of glandular architecture is a major challenge for colon histopathologists as a result of the difficulty of identifying morphological structures in glandular malignant tumours due to the distortion of glands boundaries, furthermore the variation in the appearance of staining specimens. For reliable analysis of colon specimens, several deep learning methods have exhibited encouraging performance in the glands automatic segmentation despite the challenges. In the histopathology field, the vast number of annotation images for training the deep learning algorithms is the major challenge. In this work, we propose a trainable Convolutional Neural Network (CNN) from end to end for detecting the glands automatically. More specifically, the Modified Res-U-Net is employed for segmenting the colorectal glands in Haematoxylin and Eosin (H&E) stained images for challenging Gland Segmentation (GlaS) dataset. The proposed Res-U-Net outperformed the prior methods that utilise U-Net architecture on the images of the GlaS dataset.

Autonomous vehicles are the subject of intense research because they are a safe and convenient alternative to present-day vehicles. Human drivers base their navigational decisions primarily on visual information and researchers have been attempting to use computers to do the same. The current challenge in using computer vision lies not in the collection or transmission of visual data, but in the perception of visual data to extract from it useful information. The focus of this thesis is on the use of computer vision to navigate an autonomous vehicle that will participate in the Intelligent Ground Vehicle Competition (IGVC.) This document starts with a description of the IGVC and the software design of an autonomous vehicle. This thesis then focuses on the weakest link in the system - the computer vision module. Vehicles at the IGVC are expected to autonomously navigate an obstacle course. Competing vehicles need to recognize and stay between lines painted on grass or pavement. The research presented in this document describes two methods used for boundary line extraction: color-based object extraction, and shape analysis for line recognition. This is the first time a combination of these methods is being applied to the problem of line recognition in the context of the IGVC. The most significant contribution of this work is a method for extracting lines in a binary image even when the line is attached to a shape that is not a line. Novel methods have been used to simplify camera calibration, and for perspective correction of the image. The results give promise of vastly improved autonomous vehicle performance. / Master of Science

Properties of aggregates are affected by their morphological characteristics, including shape factors, angularity and texture. These morphological characteristics influence the aggregate's mutual interactions and strengths of bonds between the aggregates and the binder. The interactions between aggregates and bond strengths between the aggregate and the binder are vital to rheological properties, related to workability and friction resistance of mixtures. As a consequence, quantification of the aggregate's morphological characteristics is essential for better quality control and performance improvement of aggregates. With advancement of hardware and software, the computation capability has reached the stage to rapidly quantify morphological characteristics at multiple scales using digital imaging techniques. Various computational algorithms have been developed, including Hough transform, Fourier transform, and wavelet analysis, etc. Among the aforementioned computational algorithms, Fourier transform has been implemented in various areas by representing the original image/signal in the spatial domain as a summation of representing functions of varying magnitudes, frequencies and phases in the frequency domain. This dissertation is dedicated to developing the two-dimensional Fourier transform (FFT2) method using the Fourier Transform Interferometry (FTI) system that is capable to quantify aggregate morphological characteristics at different scales. In this dissertation, FFT2 method is adopted to quantify angularity and texture of aggregates based on surface coordinates acquired from digital images in the FTI system. This is followed by a comprehensive review on prevalent aggregate imaging techniques for the quantification of aggregate morphological characteristics, including the second generation of Aggregate Image Measurement System (AIMS II), University of Illinois Aggregate Image Analyzer (UIAIA), the FTI system, etc. Recommendations are made on the usage of aggregate imaging system in the measurements of morphological parameters that are interested. After that, the influence of parent rock, crushing, and abrasion/polishing on aggregate morphological characteristics are evaluated. Atomic-scale roughness is calculated for crystal structures of five representative minerals in four types of minerals (i.e., α-quartz for quartzite/granite/gravel/aplite, dolomite for dolomite, calcite for limestone, haematite and magnetite for iron ore); roughness ranking at atomic-scale is further compared with surface texture ranking at macroscale based on measurement results using the FTI system and AIMS II. Morphological characteristics of aggregates before and after crushing test and micro-deval test are measured to quantitatively evaluate the influences of the crushing process and the abrasion/polishing process on morphological characteristics of aggregates, respectively. / Ph. D.

This thesis tackles the challenge of real-time image analysis in resource-constrained embedded systems, focusing specifically on Micro Aerial Vehicle (MAV) applications. The primary objective of this research is to design a software architecture that integrates features like modularity, real-time capabilities, robustness, and adaptability to meet the demands. The study proposes a unique software architecture based on blackboard and microservices architectures, that facilitates the key strengths from both paradigms, while mitigating their individual limitations. Additionally, it leverages shared memory inter-process communication mechanism for implementing centralized knowledge base of the blackboard, and realizing the API of the microservices architecture. The computer vision system tasks are decomposed into smaller pieces, and developed and implemented as loosely coupled individual software components. The thesis contribution lies in an efficient architecture for real-time image analysis on safety-critical and resource-constrained MAV platforms. The architecture provides an efficient and real-time-capable backbone and offers modularity and reusability for diverse applications.:1. Introduction 2. Fundamentals 3. Literature Review 4. Conceptualization of Real-Time Software Architecture 5. Implementation 6. Test and Evaluation 7. Conclusion and Future Scope Appendix

To analyse the large amounts of image data that are generated by biologists with modern microscopes, machine learning algorithms became increasingly popular. In collaboration with Frank Schnorrer and Cl ́ement Rodier at Institut de Biologie du Developpement de Marseille, as well as Ian Estabrook at Physics of Life, TU Dresden, this thesis applies the supervised machine learning algorithms ‘Support Vector Machine’ and ‘Random Forest’ to data obtained from fluorescence microscope images of myofibrillogenesis in Drosophila pupae with the aim to identify sarcomeres, the structures that makeup the highly regular myofibrils. For the implementation in MATLAB, methods such as ‘feature engineering’ are used to increase the performance by reinterpreting the input data and using physical characteristics of the sample system. The project also identifies the problem of class imbalance between positive and negative examples in the input data and counters it with a redefined learning cost. In conclusion, the use of machine learning algorithms for image analysis in biophysics is a very promising way to reduce manual labour. The choice of the best learning algorithm depends on the purpose the obtained output data should serve.