Jonathan Siegel, M.A. - US grants

This grant is directed at the study of myocardial metabolism under conditions of surgical stress related to myocardial hypothermic preservation during cardiac bypass, low flow normothermic ischemia, and hyperdynamic sepsis. Studies are underway in evaluating the myocardial utilization of various substrates during the period of hypothermic cardiac preservation and in the immediate post perfusion rewarming period when the myocardium must take over circulatory demands and preliminary studies are being initiated in the other two areas. This study of myocardial metabolism has been developed experimentally in a myocardial pedicle system which enables complete control of arterial inflow and coronary venous outflow in a segment of the intact working left ventricle of a dog from which all collaterals have been excluded. Techniques also have been developed for the study of substrate metabolism using 13C labeled glucose which permits tracing of the metabolic byproducts of glucose through Krebs cycle of into ketone and fat metabolism. The regulation of glucose metabolsim under conditions of hypothermic preservation and during rewarm reperfusion during low flow ischemia, and in sepsis are being studied. The effects of various substrate mixtures is deing correlated with the degree of return of effective myocardial electrical activity in a coherent transmural propagation velocity sequence as an indicator of myocardial arrhythmogenic potential. These techniques are designed so as to be applicable to human studies. Thus, they involve the use of non radioactive 13C labeled substrates which can be ascertained by GC-mass spectrometry techniques in a 50 mg biopsy taken during cardiopulmonary bypass and electrical measurements which can be made with a very fine gauge needle electrode containing multiple ECG sensor tips. Studies of human myocardial metabolism are planned in cardiac surgical patients undergoing cardiopulmonary bypass and the clinical effect of various substrate mixtures with and without K+ cardioplegia will be evaluated in the setting of intrinsic myocardial disease. Finally, a non invasive method of the detection of focal areas of aberrant or delayed transmural conduction form body surface electrodes is being developed, which can be applied in man as a guide to the detection of abnormal areas of metabolic and electrical recovery from injury. These techniques will be used to study the use of alternative metabolic fuels in the management of acute myocardial failure states.

This is a multidisciplinary study of synaptic transmission between the inner hair cells (IHC) of the mammalian cochlea and the radial afferent neurons of the spiral ganglion. Its purpose is to describe for the first time the quantal release of transmitter by IHCs, to examine the role of calcium in mediating transmitter release and the initiation of postsynaptic action potentials in this system. Another component of the study will be to identify morphological correlates of synaptic transmission by the IHCs under conditions of varying rates of transmitter release by following the incorporation by IHCs of the tracer horseradish peroxidase (HRP) at the ultrastructural level. The physiology of the synapse will be assessed by intracellular recording from the radial afferents near their points of synaptic contact with the OHCs and by recording from single cochlear nerve afferents during simultaneous perfusion of the perilymphatic spaces with solutions containing altered divalent cation content. The number and ultrastructure of active zones supplying radial afferents will be assessed by serially reconstructing their terminations on the hair cells with high voltage electron microscopy (HVEM). The long range goal is to develop a physically-based model of this sensory synapse. The study has public health implications in that it will open the way for the study of drug and disease mechanisms which may specifically affect the synapse.

The cochlea is the receptor organ for hearing in mammals. One group of its mechanosensory cells, the inner hair cells, make unique functional synaptic contacts with auditory sensory neurons that carry information to the brain. Unlike many other synaptic sites, in the mammalian cochlea a single active zone of one inner hair cell usually constitutes the entire synaptic input to the afferent neurons, unlike the extensive branching common elsewhere. This exploratory project will develop a novel isolated preparation of the mammalian cochlea to exploit the unique opportunity to study the physiology of a single synaptic active zone. Results will be important to our understanding of synaptic mechanisms, and the isolated preparation will provide an important new approach with impact on auditory neuroscience.

This is a Shannon Award providing partial support for research projects that fall short of the assigned institute's funding range but are in the margin of excellence. The Shannon award is intended to provide support to test the feasibility of the approach; develop further tests and refine research techniques; perform secondary analysis of available data sets; or conduct discrete projects that can demonstrate the PI's research capabilities or lend additional weight to an already meritorious application. Further scientific data for the CRISP System are unavailable at this time.

DESCRIPTION: This project seeks to understand how the responses to acoustic stimuli by inner hair cells (IHC) of the mammalian hearing organ are influenced by ionic currents passing through specific ion channels in the hair cell's basolateral membrane. A biophysical model predicts that among the effects of voltage-activated ion channels are likely to be a contribution to the adaptation of the hair cell to continuous sinusoidal stimulation and a hair cell equivalent of two-tone suppression. Whole-cell patch-clamp recording techniques will be carried out in intact cochleas removed from Mongolian gerbils and maintained by perfusion. Hair cell viability will be assessed using fluorescent dyes that indicate intracellular pH, membrane potential and synaptic vesicle recycling. Patch clamp data will be integrated into the biophysical hair cell model. Changes in membrane capacitance resulting from fusion of synaptic vesicles will be measured to characterize transmitter release. In addition to examining the basic biophysics of hair cell function, this work also contributes to the understanding of hair cell ionic homeostasis and thereby has relevance to understanding of hearing loss resulting from metabolic or other trauma to hair cells.

This research connects two different fields, machine learning from data science and numerical partial differential equations from scientific and engineering computing, through the comparative study of the finite element method and finite neuron method. Finite element methods have undergone decades of study by mathematicians, scientists and engineers in many fields and there is a rich mathematical theory concerning them. They are widely used in scientific computing and modelling to generate accurate simulations of a wide variety of physical processes, most notably the deformation of materials and fluid mechanics. By contrast, deep neural networks are relatively new and have only been widely used in the last decade. In this short time, they have demonstrated remarkable empirical performance on a wide variety of machine learning tasks, most notably in computer vision and natural language processing. Despite this great empirical success, there is still a very limited mathematical understanding of why and how deep neural networks work so well. We hope to leverage the success of deep learning to improve numerical methods for partial differential equations and to leverage the theoretical understanding of the finite element method to better understand deep learning. The interdisciplinary nature of the research will also provide a good training experience for junior researchers. This project will support 1 graduate student each year of the three year project.

Piecewise polynomials represent one of the most important functional classes in approximation theory. In classical approximation theory and numerical methods for partial differential equations, these functional classes are often represented by linear functional spaces associated with a priori given grids, for example, by splines and finite element spaces. In deep learning, function classes are typically represented by a composition of a sequence of linear functions and coordinate-wise non-linearities. One important non-linearity is the rectified linear unit (ReLU) function and its powers (ReLUk). The resulting functional class, ReLUk-DNN, does not form a linear vector space but is rather parameterized non-linearly by a high-dimensional set of parameters. This function class can be used to solve partial differential equations and we call the resulting numerical algorithms the finite neuron method (FNM). Proposed research topics include: error estimates for the finite neuron method, universal construction of conforming finite elements for arbitrarily high order partial differential equations, an investigation into how and why the finite neuron method gives a much better asymptotic error estimate than the corresponding finite element method, and the development and analysis of efficient algorithms for using the finite neuron method.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

Deep neural networks (DNN) have become a core technology for building artificial intelligence (AI) systems, with numerous applications in critical domains such as manufacturing and medicine. Despite the phenomenal success of such networks, their usage has met resistance in many mission-critical tasks because it is difficult to explain the prediction model generated by the computer. For example, in some applications in public health and medicine, because only interpretable models are acceptable, users have stuck with classic machine-learning methods, which are often not as accurate as DNN models. Moreover, increased transparency in AI systems makes human inspection possible, a necessary trait for studying social justice and equity in AI. This project aims to develop theory, methods, and applications to achieve interpretability for machine-learning models based on DNN. Advancement in this project will broaden the usage of DNN in science, engineering, and industry, further unleashing its power. Besides developing fundamental methodologies, the investigators will advance the application area of automated emotion recognition. The ability to recognize and quantify emotion can help psychologists and clinical workers notice extreme distress and potential danger to oneself and others. Through this project, the research team will develop software packages for public access, graduate and undergraduate students will be trained to conduct interdisciplinary research, and the faculty members of the team will integrate the research results into their teaching activities.<br/><br/>Although various post-hoc methods have been developed to interpret the decision of DNNs, the explanation is often unstable and highly localized by construction. More importantly, the explanation model exists in separation from the prediction model, whose high complexity remains despite the explanation. In this project, inspired by source coding, the investigators will draw an analogy between explaining a complex model and transmitting signals with a limited channel capacity. Similar to vector quantization to enable data transmission at an allowable rate, the prediction mapping is quantized so that the prediction can be described at a desired level of interpretability. To formalize the idea, the investigators propose a mixture of discriminative models, trained as an embedded part of a neural network. Function approximation theory will be developed for interpretable models based on neural networks. Besides testing and evaluating the proposed framework using benchmark datasets, such as images, videos, text, and biomedical data, the investigators will explore in greater depth the application for emotion recognition based on body movements and, in collaboration with psychology researchers, evaluate the insight gained from interpreting the prediction model.<br/><br/>This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

Since 1987, the International Conferences of Domain Decomposition Methods (DDM) have been held in 15 countries throughout Asia, Europe and North America. The 27th edition of this prestigious conference will be held at the Czech Technical University in Prague, Czech Republic, on July 25-29, 2022. The primary topic of the conference concerns large scale scientific computing problems, which are an important and very general interdisciplinary research field and plays a crucial role in many application areas, for example weather forecast, seismic predictions, fluid simulations, material industry, oil field exploration, medical imaging, among others. This project will enable US based junior researchers to travel to the conference, network, and present their research.<br/><br/>The International Conference of Domain Decomposition Methods is the only regularly occurring international forum dedicated to interdisciplinary technical interactions between theoreticians and practitioners working in the development, analysis, software implementation, and application of domain decomposition methods, which are important in scientific computing. The complexity and size of such problems often lead to very large systems, where massive computing resources and special computational approaches must be employed to obtain approximate solutions. As we approach the dawn of exascale computing, scalable techniques such as DDM are vital tools for solving complex problems that would otherwise be intractable.<br/><br/>This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.