In this study, I aim to achieve multimodal brain image fusion by referring to some intrinsic properties of data, e.g. geometry of embedding structures where the commonly used image features reside. Since the image features investigated in this study share an identical embedding space, i.e. either defined on a brain surface or brain atlas, where a graph structure is easy to define, it is straightforward to consider the mathematically meaningful properties of the shared structures from the geometry perspective.
I first introduce the background of multimodal fusion of brain image data and insights of geometric properties playing a potential role to link different modalities. Then, several proposed computational frameworks either using the solid and efficient geometric algorithms or current geometric deep learning models are be fully discussed. I show how these designed frameworks deal with distinct geometric properties respectively, and their applications in the real healthcare scenarios, e.g. to enhanced detections of fetal brain diseases or abnormal brain development.
Domestic dogs have assisted humans for millennia. However, the extent to which these helpful behaviors are prosocially motivated remains unclear. To assess the propensity of pet dogs to spontaneously and actively rescue distressed humans, this study tested whether sixty pet dogs would release their seemingly trapped owners from a large box. To examine the causal mechanisms that shaped this behavior, the readiness of each dog to open the box was tested in three conditions: 1) the owner sat in the box and called for help (“Distress” test), 2) an experimenter placed high-value food rewards in the box (“Food” test), and 3) the owner sat in the box and calmly read aloud (“Reading” test).
Dogs were as likely to release their distressed owner as to retrieve treats from inside the box, indicating that rescuing an owner may be a highly rewarding action for dogs. After accounting for ability, dogs released the owner more often when the owner called for help than when the owner read aloud calmly. In addition, opening latencies decreased with test number in the Distress test but not the Reading test. Thus, rescuing the owner could not be attributed solely to social facilitation, stimulus enhancement, or social contact-seeking behavior.
Dogs displayed more stress behaviors in the Distress test than in the Reading test, and stress scores decreased with test number in the Reading test but not in the Distress test. This evidence of emotional contagion supports the hypothesis that rescuing the distressed owner was an empathetically-motivated prosocial behavior. Success in the Food task and previous (in-home) experience opening objects were both strong predictors of releasing the owner. Thus, prosocial behavior tests for dogs should control for physical ability and previous experience.
etc. Given a low resolution image, it aims to reconstruct a high resolution
image. The problem is ill-posed since there can be more than one high resolution
image corresponding to the same low-resolution image. To address this problem, a
number of machine learning-based approaches have been proposed.
In this dissertation, I present my works on single image super-resolution (SISR)
and accelerated magnetic resonance imaging (MRI) (a.k.a. super-resolution on MR
images), followed by the investigation on transfer learning for accelerated MRI reconstruction.
For the SISR, a dictionary-based approach and two reconstruction based
approaches are presented. To be precise, a convex dictionary learning (CDL)
algorithm is proposed by constraining the dictionary atoms to be formed by nonnegative
linear combination of the training data, which is a natural, desired property.
Also, two reconstruction-based single methods are presented, which make use
of (i)the joint regularization, where a group-residual-based regularization (GRR) and
a ridge-regression-based regularization (3R) are combined; (ii)the collaborative representation
and non-local self-similarity. After that, two deep learning approaches
are proposed, aiming at reconstructing high-quality images from accelerated MRI
acquisition. Residual Dense Block (RDB) and feedback connection are introduced
in the proposed models. In the last chapter, the feasibility of transfer learning for
accelerated MRI reconstruction is discussed.
To overcome the above data scarcity and generalization issues, in my dissertation, I first propose two unsupervised conventional machine learning algorithms, hyperbolic stochastic coding, and multi-resemble multi-target low-rank coding, to solve the incomplete data and missing label problem. I further introduce a deep multi-domain adaptation network to leverage the power of deep learning by transferring the rich knowledge from a large-amount labeled source dataset. I also invent a novel time-sequence dynamically hierarchical network that adaptively simplifies the network to cope with the scarce data.
To learn a large number of unseen concepts, lifelong machine learning enjoys many advantages, including abstracting knowledge from prior learning and using the experience to help future learning, regardless of how much data is currently available. Incorporating this capability and making it versatile, I propose deep multi-task weight consolidation to accumulate knowledge continuously and significantly reduce data requirements in a variety of domains. Inspired by the recent breakthroughs in automatically learning suitable neural network architectures (AutoML), I develop a nonexpansive AutoML framework to train an online model without the abundance of labeled data. This work automatically expands the network to increase model capability when necessary, then compresses the model to maintain the model efficiency.
In my current ongoing work, I propose an alternative method of supervised learning that does not require direct labels. This could utilize various supervision from an image/object as a target value for supervising the target tasks without labels, and it turns out to be surprisingly effective. The proposed method only requires few-shot labeled data to train, and can self-supervised learn the information it needs and generalize to datasets not seen during training.
Human protein diversity arises as a result of alternative splicing, single nucleotide polymorphisms (SNPs) and posttranslational modifications. Because of these processes, each protein can exists as multiple variants in vivo. Tailored strategies are needed to study these protein variants and understand their role in health and disease. In this work we utilized quantitative mass spectrometric immunoassays to determine the protein variants concentration of beta-2-microglobulin, cystatin C, retinol binding protein, and transthyretin, in a population of 500 healthy individuals. Additionally, we determined the longitudinal concentration changes for the protein variants from four individuals over a 6 month period. Along with the native forms of the four proteins, 13 posttranslationally modified variants and 7 SNP-derived variants were detected and their concentration determined. Correlations of the variants concentration with geographical origin, gender, and age of the individuals were also examined. This work represents an important step toward building a catalog of protein variants concentrations and examining their longitudinal changes.
The apolipoprotein E (APOE) e4 allele is the most prevalent genetic risk factor for Alzheimer's disease (AD). Hippocampal volumes are generally smaller in AD patients carrying the e4 allele compared to e4 noncarriers. Here we examined the effect of APOE e4 on hippocampal morphometry in a large imaging database—the Alzheimer's Disease Neuroimaging Initiative (ADNI). We automatically segmented and constructed hippocampal surfaces from the baseline MR images of 725 subjects with known APOE genotype information including 167 with AD, 354 with mild cognitive impairment (MCI), and 204 normal controls. High-order correspondences between hippocampal surfaces were enforced across subjects with a novel inverse consistent surface fluid registration method. Multivariate statistics consisting of multivariate tensor-based morphometry (mTBM) and radial distance were computed for surface deformation analysis. Using Hotelling's T2 test, we found significant morphological deformation in APOE e4 carriers relative to noncarriers in the entire cohort as well as in the nondemented (pooled MCI and control) subjects, affecting the left hippocampus more than the right, and this effect was more pronounced in e4 homozygotes than heterozygotes. Our findings are consistent with previous studies that showed e4 carriers exhibit accelerated hippocampal atrophy; we extend these findings to a novel measure of hippocampal morphometry. Hippocampal morphometry has significant potential as an imaging biomarker of early stage AD.
Mild Cognitive Impairment (MCI) is a transitional stage between normal aging and dementia and people with MCI are at high risk of progression to dementia. MCI is attracting increasing attention, as it offers an opportunity to target the disease process during an early symptomatic stage. Structural magnetic resonance imaging (MRI) measures have been the mainstay of Alzheimer's disease (AD) imaging research, however, ventricular morphometry analysis remains challenging because of its complicated topological structure. Here we describe a novel ventricular morphometry system based on the hyperbolic Ricci flow method and tensor-based morphometry (TBM) statistics. Unlike prior ventricular surface parameterization methods, hyperbolic conformal parameterization is angle-preserving and does not have any singularities. Our system generates a one-to-one diffeomorphic mapping between ventricular surfaces with consistent boundary matching conditions. The TBM statistics encode a great deal of surface deformation information that could be inaccessible or overlooked by other methods. We applied our system to the baseline MRI scans of a set of MCI subjects from the Alzheimer's Disease Neuroimaging Initiative (ADNI: 71 MCI converters vs. 62 MCI stable). Although the combined ventricular area and volume features did not differ between the two groups, our fine-grained surface analysis revealed significant differences in the ventricular regions close to the temporal lobe and posterior cingulate, structures that are affected early in AD. Significant correlations were also detected between ventricular morphometry, neuropsychological measures, and a previously described imaging index based on fluorodeoxyglucose positron emission tomography (FDG-PET) scans. This novel ventricular morphometry method may offer a new and more sensitive approach to study preclinical and early symptomatic stage AD.
Studies on urban heat island (UHI) have been more than a century after the phenomenon was first discovered in the early 1800s. UHI emerges as the source of many urban environmental problems and exacerbates the living environment in cities. Under the challenges of increasing urbanization and future climate changes, there is a pressing need for sustainable adaptation/mitigation strategies for UHI effects, one popular option being the use of reflective materials. While it is introduced as an effective method to reduce temperature and energy consumption in cities, its impacts on environmental sustainability and large-scale non-local effect are inadequately explored. This paper provides a synthetic overview of potential environmental impacts of reflective materials at a variety of scales, ranging from energy load on a single building to regional hydroclimate. The review shows that mitigation potential of reflective materials depends on a set of factors, including building characteristics, urban environment, meteorological and geographical conditions, to name a few. Precaution needs to be exercised by city planners and policy makers for large-scale deployment of reflective materials before their environmental impacts, especially on regional hydroclimates, are better understood. In general, it is recommended that optimal strategy for UHI needs to be determined on a city-by-city basis, rather than adopting a “one-solution-fits-all” strategy.
Land surface energy balance in a built environment is widely modelled using urban canopy models with representation of building arrays as big street canyons. Modification of this simplified geometric representation, however, leads to challenging numerical difficulties in improving physical parameterization schemes that are deterministic in nature. In this paper, we develop a stochastic algorithm to estimate view factors between canyon facets in the presence of shade trees based on Monte Carlo simulation, where an analytical formulation is inhibited by the complex geometry. The model is validated against analytical solutions of benchmark radiative problems as well as field measurements in real street canyons. In conjunction with the matrix method resolving infinite number of reflections, the proposed model is capable of predicting the radiative exchange inside the street canyon with good accuracy. Modeling of transient evolution of thermal filed inside the street canyon using the proposed method demonstrate the potential of shade trees in mitigating canyon surface temperatures as well as saving of building energy use. This new numerical framework also deepens our insight into the fundamental physics of radiative heat transfer and surface energy balance for urban climate modeling.