Coming dissertations at Uppsala university
-
Equivariant Neural Networks for Biomedical Image Analysis
Link: http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-519611
While artificial intelligence and deep learning have revolutionized many fields in the last decade, one of the key drivers has been access to data. This is especially true in biomedical image analysis where expert annotated data is hard to come by. The combination of Convolutional Neural Networks (CNNs) with data augmentation has proven successful in increasing the amount of training data at the cost of overfitting. In this thesis, equivariant neural networks have been used to extend the equivariant properties of CNNs to more transformations than translations. The networks have been trained and evaluated on biomedical image datasets, including bright-field microscopy images of cytological samples indicating oral cancer, and transmission electron microscopy images of virus samples. By designing the networks to be equivariant to e.g. rotations, it is shown that the need for data augmentation is reduced, that less overfitting occurs, and that convergence during training is faster. Furthermore, equivariant neural networks are more data efficient than CNNs, as demonstrated by scaling laws. These benefits are not present in all problem settings and which benefits will occur is somewhat unpredictable. We have identified that the results to some extent depend on architectures, hyperparameters and datasets. Further research may broaden the performed studies to explain how the results occur with new theory.
-
The epidemiology of risk factors and short- and long-term outcome in the Swedish intensive care cohort
Link: http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-519461
The sepsis syndrome is present in ¼ to ⅓ of patients in intensive care units (ICUs) worldwide. The short-term prognosis is grim, with a 30-day mortality of 30–35%; however, the long- term outcomes are now being explored, as multi-professional follow-up after ICU care is increasingly being implemented. In 2020 the first and second waves of another severe infection, the Coronavirus disease 2019 (Covid-19) hit Sweden. The number of ICU beds were scaled up by several hundred percent while we simultaneously tried to understand the disease. Reports on risk factors for adverse outcomes in Covid-19 started to appear, but we needed to know more. Thus, we initiated this project aiming at assessing sepsis as an independent risk factor for later morbidity and mortality. Subsequently, with the onset of the pandemic, our focus shifted to identifying risk factors for adverse outcomes in Covid-19 and describing the functional recovery after severe Covid-19. We used the Swedish Intensive Care Registry and several governmental registries to this end.
In Cox regression, we compared one-year ICU sepsis survivors without previous dementia with ICU patients without sepsis, finding no increased risk of dementia during follow- up. In a similar cohort, we assessed the impact of sepsis on long-term mortality and causes of death in a series of Cox and multinomial models. We found a surprisingly small overall association between sepsis and mortality and a persistently increased risk of infectious causes of death in sepsis patients. We compared the prevalence of several common comorbidities and medications as risk factors for ICU admission and mortality in ICU patients with Covid-19 with that of age- and sex-matched population controls and in patients discharged alive with those that were deceased at discharge. We found associations between several comorbidities and medications with these adverse outcomes. To better understand the meaning of these comorbidities as risk factors for short-term mortality, we compared them in logistic regression models on patients with Covid-19, sepsis and acute respiratory distress syndrome (ARDS). We found very similar impacts from the comorbidities; however, greater age was more associated with mortality in Covid-19 than in either sepsis or ARDS. Finally, we investigated the long-term functional recovery in ICU patients with Covid-19 compared to hospital-admitted patients with Covid-19 and population controls matched to the ICU group. The ICU patients had a markedly impeded recovery that was not explained by demographics or comorbidities in statistical models.
-
Paranormal Sweden? : Paranormal beliefs and practices in contemporary Sweden
Link: http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-518522
This dissertation aims to study the contemporary occurrence of paranormal beliefs, activities and experiences in Sweden as well as a paraculture of practices dedicated to these issues, with a particular focus on issues of knowledge and authority. The study is placed against the backdrop of academic and popular claims that paranormal beliefs, activities and experiences are not only common but are also on the rise.
A quantitative survey targets a representative sample of Swedish adults (n=1101) and assesses the relation between demographic characteristics and paranormal beliefs and practices. The survey is complemented with fieldwork in the form of participant observation and semi-structured interviews in the paraculture. More specifically, settings and organizations related to mediumship, cryptozoology, ghost hunting, parapsychology and ufology are studied as cases of paranormal practice. The data is analyzed in relation to theorizations of occulture and epistemic authority, the latter by combining the concepts of epistemic capital and boundary-work. The findings point to gender (i.e., being female) and the number of recent occultural contact points as the strongest predictors of paranormal beliefs, activities and experiences. In the paraculture, participants are prone to appeal to different strategies of epistemic capital depending on the setting they were recruited from. One notable result is that while most participants and groups relied on counter-epistemic strategies, such strategies are combined with conventional strategies of science-like and traditional strategies of epistemic capital. Paranormal practice, distinctly epistemic, accordingly becomes a case of how knowledge-making on contested and controversial phenomena may take form in light of the tectonic social and cultural shifts implied by modernization.