Coming dissertations at Uppsala university
-
Towards sustainable Ni-rich layered oxide cathodes : A synchrotron-based study
Link: http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-543384
Rechargeable Li-ion batteries (LIBs) are essential for portable electronic devices, electric vehicles and the development of large-scale energy storage for renewable sources. Among various cathode materials in LIBs, layered Ni-rich transition metal oxides are widely used due to their high energy density. Conventionally, the toxic N-methyl-2-pyrrolidone (NMP) solvent and fluorine-containing polyvinylidene fluoride (PVdF) binder are utilized during electrode manufacturing. However, it is desirable to replace NMP with an environmentally friendly solvent and also to aim for a fluorine-free binder. Thus, this thesis aims to develop the aqueous-processing methodology for LiNi0.8Mn0.1Co0.1O2 (NMC811) electrode production.
This thesis identifies the formation of carboxylate species as a product of the irreversible reaction between the NMC811 surface and H2O vapor. Furthermore, results show that aqueous processing generates a reactive electrode surface, with subsequent electrolyte decomposition. In addition, a NiO-like rock-salt phase forms in the near-surface regions, most likely due to Li-ion leaching and Li/Ni disorder. Also, increased charge transfer resistance is observed, which likely correlate to the rock-salt phase. Building on insights into H2O’s effects on the NMC811 surface, two aqueous-processing methods for producing NMC811 electrodes are studied. To mitigate these challenges, firstly H3PO4 is added to the aqueous slurry, primarily to lower the pH and limit Al current collector corrosion. This modification to some extent stabilizes the reactive electrode surface and alleviates Li/Ni disorder, leading to improved capacity retention and enhanced reversibility of the phase transition. Secondly, with the aim to stabilize the NMC811 surface during aqueous processing, Ti is incorporated within the structure. This effectively hinders rock-salt phase formation and reduce the Li-ion transfer resistance. With inspiration from a reaction heterogeneity detected in the aqueous-processed NMC811 electrode, the study further investigates particle-scale Li-ion heterogeneity in the commercially aged LixNi0.9Co0.05Al0.05 secondary particles, suggesting a significant Li-ion heterogeneity within the particles cycled to a high state of charge.
In conclusion, this thesis elucidates the degradation mechanisms of aqueous-processed NMC811 material and demonstrates the roles of material modifications in enhancing cycling performance, offering valuable insights into the manufacturing of sustainable batteries. Furthermore, it highlights the importance of employing X-ray-based techniques for in-depth studies of battery materials.
-
Representation Learning for Computational Pathology and Spatial Omics
Link: http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-542989
Artificial intelligence (AI) advancements have enhanced the analysis and interpretation of computational pathology. Through representation learning, deep learning models can automatically identify complex patterns and extract meaningful features from raw data, revealing subtle spatial relationships. Spatial omics, which captures spatially resolved molecular data, naturally aligns with these approaches, enabling a deeper examination of tissue architecture and cellular heterogeneity. However, early spatial omics methods often overlooked the morphological and spatial context inherent in tissues.
The integration of spatial omics with imaging AI and representation learning provides a comprehensive view for understanding complex tissue environments, providing deeper insights into disease mechanisms and molecular landscapes. This thesis investigates how deep learning-derived representations from biological images can be utilized in the context of spatial omics and disease processes.
Key contributions of this work include: (i) investigating the correlation between representations learned from models trained on hematoxylin-eosin (H&E)-stained images and underlying gene expression profiles; (ii) applying self-supervised learning to identify genetically relevant patterns across H&E and DAPI staining; and (iii) developing a framework that leverages self-supervised representations to refine cell-type assignments obtained from spatial transcriptomics deconvolution methods. As a culmination of this part of the thesis, this research introduces (iv) a conceptual framework for understanding representations within spatial omics and provides a survey of the current literature through this lens.
The thesis further includes practical applications such as (v) developing a tool for annotation of whole-slide images (WSI) using self-supervised representations and (vi) exploring the use of weakly-supervised learning to identify early tumor-indicating morphological changes in benign prostate biopsies.
-
Image and data analysis for understanding spatial transcriptomics and tissue architecture
Link: http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-543366
The human body is a complicated system, with its complex functions interpreted through interactions that scale from tissues down to individual genes. To unravel this complexity, we must look at the smallest functional compartments, specifically, the spatial organization of gene expression within tissues. This thesis focuses on computational methods in spatial transcriptomics, to precisely map gene expression patterns, providing insights into how cellular neighborhoods and protein interactions drive tissue functionality. Central to this research are computational tools and methodologies that tackle the spatial transcriptomics pipeline from start to finish, addressing data acquisition, pre-processing, decoding, classification, and spatial statistics. The contributions are presented across two perspectives: the technical advancements in the pipeline and their application to real-world biological scenarios.
On the technical side, this thesis introduces two novel classification methods. The first is a graph-based segmentation algorithm that groups molecular signals into cells without relying on nuclear stains, overcoming common imaging limitations. The second is a fast and interactive clustering method for imaging-based spatial omics data. Additionally, the thesis includes work on TissUUmaps 3, an interactive visualization tool designed for high-resolution exploration and quality assessment of large-scale spatial data, along with plugins to enhance the detailed analysis of spatial patterns in gene expression across tissue types. A comprehensive review of spatial statistics methods applicable to spatial omics data complements these technical contributions.
The biological application side demonstrates the utility of these tools in uncovering insights from real-world datasets. In the mouse olfactory epithelium, spatial gene expression mapping reveals cellular patterns critical for sensory processing. In bladder cancer, spatial transcriptomics illuminates immune cell behavior and interactions within the tumor microenvironment, showcasing how these methods bridge computational advancements with biological discovery.
Together, these contributions highlight the power of computational spatial transcriptomics to reveal critical insights into tissue architecture, gene co-expression, and cellular functionality. By refining and scaling analytical techniques, this thesis advances our understanding of tissue complexity and paves the way for discoveries in developmental biology, oncology, and precision medicine, demonstrating the transformative potential of spatial omics for both fundamental research and clinical applications.