Particle Image Velocimetry (PIV) for flow visualization has advanced with the integration of deep learning algorithms. These methods enable end-to-end processing, extracting dense flow fields directly from raw particle images. However, conventional deep learning-based PIV models, which predominantly rely on convolutional architectures, are limited in their ability to utilize contextual information and capture dependencies between pixels across sequential images, impacting the prediction accuracy. We introduce Twins-PIVNet, a deep learning framework for PIV optical flow estimation that leverages a spatial attention-based vision transformer architecture. Its self-attention mechanism captures multi-scale features of particle motion, significantly improving the dense flow field estimation. Trained on synthetic PIV datasets covering a wide range of flow conditions, Twins-PIVNet has been evaluated on both synthetic and experimental datasets, demonstrating superior accuracy and performance. In comparative studies, Twins-PIVNet outperforms existing optical flow and conventional methods, achieving accuracy improvements of 51% for backstep flow, 42% for DNS-turbulence, and 33% for surface quasi-geostrophic flow. Additionally, it also exhibits strong generalization on experimental PIV data, demonstrating robustness in handling real-world PIV uncertainties. Despite its attention mechanism, Twins-PIVNet maintains faster inference and training times compared to other PIV models, offering an optimal balance between complexity, efficiency, and performance.

A bi-directional digital holographic imaging system is presented that is designed to take images automatically out in the field. The main objective is to acquire sufficient information to be able to estimate the scattering phase function for different type of snowfall. The imaging system consists of a 3D-printed frame and two orthogonal telecentric imaging arms, one in the forward direction and one in the side scattering direction for which the reference arms are directed along different paths. All images are acquired using pulsed visible light.

Current optical flow-based neural networks for particle image velocimetry (PIV) are largely trained on synthetic datasets emulating real-world scenarios. While synthetic datasets provide greater control and variation than what can be achieved using experimental datasets for supervised learning, it requires a deeper understanding of how or what factors dictate the learning behaviors of deep neural networks for PIV. In this study, we investigate the performance of the recurrent all-pairs field transforms-PIV (RAFTs-PIV) network, the current state-of-the-art deep learning architecture for PIV, by testing it on unseen experimentally generated datasets. The results from RAFT-PIV are compared with a conventional cross-correlation-based method, Adaptive PIV. The experimental PIV datasets were generated for a typical scenario of flow past a circular cylinder in a rectangular channel. These test datasets encompassed variations in particle diameters, particle seeding densities, and flow speeds, all falling within the parameter range used for training RAFT-PIV. We also explore how different image pre-processing techniques can impact and potentially enhance the performance of RAFT-PIV on real-world datasets. Thorough testing with real-world experimental PIV datasets reveals the resilience of the optical flow-based method's variations to PIV hyperparameters, in contrast to the conventional PIV technique. The ensemble-averaged root mean squared errors between the RAFT-PIV and Adaptive PIV estimations generally range between 0.5–2 (px) and show a slight reduction as particle densities increase or Reynolds numbers decrease. Furthermore, findings indicate that employing image pre-processing techniques to enhance input particle image quality does not improve RAFT-PIV predictions; instead, it incurs higher computational costs and impacts estimations of small-scale structures.

The depth-gating capacity of a spatially quasi-incoherent imaging interferometer is investigated in relation to the 3D correlation properties of diffraction field laser speckles. The system exploits a phase-stepped imaging Michelson-type interferometer in which spatially quasi-incoherent illumination is generated by passing an unexpanded laser beam through a rotating diffuser. Numerical simulations and optical experiments both verify that the depth-gating capacity of the imaging interferometer scales as 𝜆/2NA2𝑝λ/2NAp2, where 𝜆λ is the wavelength of the laser and NA𝑝NAp is the numerical aperture of the illumination. For a set depth gate of 150 µm, the depth-gating capacity of the interferometer is demonstrated by scanning a standard USAF target through the measurement volume. The results obtained show that an imaging tool of this kind is expected to provide useful capabilities for imaging through disturbing media and where a single wavelength is required.

An interferometric technique that utilize a spatially quasi-incoherent light source to perform interferometric measurements involving diffusely scattering objects is presented. The proposed technique is demonstrated with settings that give a depth gate of 90 µm.

Mitochondrial Cytochrome c Oxidase Subunit 4 Isoform 2 (Cox4i2) is essential for acute oxygen sensing and signaling in pulmonary arterial smooth muscle cells (PASMCs) by triggering the production of superoxide during acute hypoxia [1]. However, the molecular mechanism underlying Cox4i2-dependent oxygen sensing remains elusive. We analysed oxygen-dependent respiration by high resolution respirometry, redox changes of the electron transport chain (ETC) by RAMAN spectroscopy, and supercomplex formation by blue native gel analysis of PASMCs isolated from wild type (WT) and Cox4i2-/- mice. To understand the role of Cox4i2-specific cysteine residues we determined hypoxia-induced superoxide production and oxygen affinity in a mouse epithelial cell line (CMT167 cells) overexpressing either Cox4i1, or WT Cox4i2, or Cox4i2 mutants (C41S, C55A, C109S). Respiration and supercomplex formation were similar in WT and Cox4i2-/- PASMCs. Interestingly, hypoxia-induced reduction of ETC components (NADH, ubiquinol, and reduced cytochrome c) was prevented in Cox4i2-/- PASMCs. CMT167 cells expressing either Cox4i1, or Cox4i2 mutants lacked hypoxia-induced superoxide release, which was detected only in cells expressing WT Cox4i2. In contrast, overexpression of Cox4i1, or Cox4i2, or Cox4i2 mutants did not affect oxygen affinity. Our findings suggest that Cox4i2 does not alter superoxide production by rearrangement of supercomplexes, whereas its specific cysteines are needed for the superoxide release. In conclusion, Cox4i2 plays a major role in the hypoxia-induced reduction of ETC components, likely mediated through its redox-active cysteine residues.

Delineating cancer tissue while leaving functional tissue intact is crucial in brain tumor resection. Despite several available aids, surgeons are limited by preoperative or subjective tools. Raman spectroscopy is a label-free optical technique with promising indications for tumor tissue identification. To allow direct comparisons between measurements preprocessing of the Raman signal is required. There are many recognized methods for preprocessing Raman spectra; however, there is no universal standard. In this paper, six different preprocessing methods were tested on Raman spectra (n > 900) from fresh brain tissue samples (n = 34). The sample cohort included both primary brain tumors, such as adult-type diffuse gliomas and meningiomas, as well as metastases of breast cancer. Each tissue sample was classified according to the CNS WHO 2021 guidelines. The six methods include both direct and iterative polynomial fitting, mathematical morphology, signal derivative, commercial software, and a neural network. Data exploration was performed using principal component analysis, t-distributed stochastic neighbor embedding, and k-means clustering. For each of the six methods, the parameter combination that explained the most variance in the data, i.e., resulting in the highest Gap-statistic, was chosen and compared to the other five methods. Depending on the preprocessing method, the resulting clusters varied in number, size, and associated spectral features. The detected features were associated with hemoglobin, neuroglobin, carotenoid, water, and protoporphyrin, as well as proteins and lipids. However, the spectral features seen in the Raman spectra could not be unambiguously assigned to tissue labels, regardless of preprocessing method. We have illustrated that depending on the chosen preprocessing method, the spectral appearance of Raman features from brain tumor tissue can change. Therefore, we argue both for caution in comparing spectral features from different Raman studies, as well as the importance of transparency of methodology and implementation of the preprocessing. As discussed in this study, Raman spectroscopy for in vivo guidance in neurosurgery requires fast and adaptive preprocessing. On this basis, a pre-trained neural network appears to be a promising approach for the operating room.

Biology can often be explained by structures and interactions between molecules. It is therefore of importance for medical sciences that the chemical reactions and molecular compositions of biological samples can be measured. The purpose of this thesis is twofold. Firstly, to develop methods for measuring the underlying mechanisms of the lung disorder pulmonary hypertension. Secondly, to investigating the possibility of classifying brain tissue for safer resection of brain tumors. A multimodal approach is often motivated, as there exist many different measurement techniques. In this thesis Raman spectroscopy has been applied both as the main measurement modality and in cooperation with other methods.

Raman spectroscopy is a label free optical measurement technique that measure inelastic scattering from materials that are illuminated by a monochromatic light source. Raman scattering results in a weak signal that is uniquely proportional to the chemical structure of the sample. When measuring biological samples, Raman scattering is accompanied by a strong intrinsic fluorescent signature that can overshadow the signal. To elucidate the underlying Raman spectrum, it is generally preprocessed to allow further analysis. There are many methods available for preprocessing; a selection of commonly used methodologies has been included here, as well as methods of my own design. The most novel approach being a neural network that was trained on synthetic spectra to perform preprocessing without relying on user defined variables, which is common for other methods (Paper A). The neural network resulted in improved preprocessing when compared to a control predictor, with test data from paraffin, ethanol, and polyethylene, as well as spectra based on simulations.

Hypoxic pulmonary hypertension (PH) is a condition where the arteries in the lung-walls are blocked due to prolonged oxygen deprivation or lung disease. People who suffer from PH, often experience shortness of breath and fatigue. If the condition persists the added strain upon the heart from the increased resistance in the arteries will result in right-heart failure. Hypoxic PH is the result of permanently constricted pulmonary arterial smooth muscle cells (PASMCs). PASMCs reside in the arterial walls and react locally to reduced oxygen content by constricting. This effect is called pulmonary vasoconstriction (HPV) and results in the regulation of deoxygenated blood to areas of the lung that have more oxygen available. Full understanding of the mechanisms of oxygen sensing in PASMCs has importance for the development of new treatments against PH. To this end a sealed microfluidic system was designed with the purpose to enable multimodal measurements of the response of cultivated PASMCs to acute hypoxia including Raman spectroscopy, patch clamp, and imaging (Paper B).  The microfluidic system was initially tested using Raman spectroscopy and oxygen sensing to investigate the reaction of PASMCs to hypoxia. The results were compared to an open flow system that showed a higher variation of the desired oxygen content (21% or 4%) compared to the designed closed microfluidic system. The system was later reworked and tested with simultaneous measurements using Raman spectroscopy, oxygen sensing, imaging, and patch-clamp (Paper C). With this setup it was possible to track the molecular response in the mitochondria in correlation with the activity of the calcium-ion channels and the mechanical response of the PASMCs.

The main modality used in clinics for brain tumor imaging is magnetic resonance imaging (MRI). Structural MRI gives neurosurgeons information regarding the size and mass effects of tumors. However, during surgery it can be difficult to assess the marginal zone of tumors. Improvements have been made by incorporating fluorescence guided resection (FGR) in the standard practice of many operating rooms in Europe. FGR relies on measuring the emission from metabolized precursor molecules. However, the drawback of FGR is that it is not tumor specific and has reduced sensitivity in low-grade tumors and children. In this thesis the option of incorporating a Raman probe setup, to fill in the gaps of other methods has been discussed. During this preliminary discussion it was noted that there is no standard approach for preprocessing and many different methodologies have been employed by various researcher. Therefore, measurement on fresh brain tumor tissue from a Raman microscopy setup was preprocessed using commonly applied methods, in addition to a pretrained neural network, to investigate the variations of the outcome (Paper D). It became apparent that different methods and variable choices can alter the distinctive spectral features. The conclusion being that it is important to be both transparent and specific when explaining how data has been prepared prior to analysis to enable reproducible results.