Doris Y. Tsao - US grants

This award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5).

A three year old infant can recognize visual objects better than any machine. Despite decades of work on the problem of complex object recognition, the principles used by the brain to solve this problem are still unknown. One major experimental difficulty has been the diversity of feature preferences exhibited by visual neurons: For any given cell, it is unknown which set of visual features will result in a robust neural response. Over the past three decades, evidence has gradually accumulated for a system in the temporal lobe that may allow researchers to overcome this difficulty, consisting of a specialized set of areas dedicated to detecting and recognizing faces. The face patch system creates unprecedented possibilities for dissecting the principles of information flow in inferotemporal cortex, by giving direct access to anatomically distinct components of a unified object processing network. The stunning selectivity of this system for faces suggests that face detection is one of its major functions. Understanding in detail the strategies for face detection used by this system should illuminate the most difficult problem in object recognition: how to recognize a visual form despite substantial changes in appearance. With the support of the National Science Foundation, Dr. Doris Tsao and colleagues at the California Institute of Technology will tackle the problem of how cells in face-selective areas of the brain are able to detect faces. The work will use functional magnetic resonance imaging (fMRI) to first identify these areas. Then, the feature selectivity of cells in these regions will be characterized. This will be followed by experiments in which subjects perform tasks in which they actively detect faces, while neural activity is recorded. Neural activity in specific face areas will be artifically increased or decreased, and the effect on face detection behavior will be observed.

Results from the present project will likely spawn fertile exchange between neuroscience, computer science, and psychology. The results may lead to new insights into the brain circuits that are altered in prosopagnosia, a selective inability to recognize faces that afflicts a surprisingly large percentage of the population. The results may also provide new insights into social disorders such as autism and social anxiety disorder, since faces are by far the most socially significant class of visual stimuli that we perceive. Understanding the design principles used by the world's best face detection system may motivate design of better artificial face recognition algorithms (which have broad applications in security and entertainment). Finally, the research will offer training opportunities for graduate students and post-doctoral fellows. Pedagogical activities included in Dr. Tsao's CAREER grant include teaching an interdisciplinary course combining computational modeling of object recognition with biological experiments. Face perception is intriguing and accessible to almost everyone--likely because almost everyone has a set of specialized face areas. Thus the results of these investigations will likely be disseminated broadly, to enhance scientific understanding in society.

Description (provided by applicant): The goal of this proposal is to understand the neural circuits of the macaque brain mediating face perception. Faces are among the most meaningful visual forms processed by the primate brain, conveying information about identity, expression, gender, age, and direction of attention. Functional magnetic resonance imaging (fMRI) reveals a specialization for face processing in the macaque temporal lobe: six discrete regions of cortex respond much more strongly to images of faces than to images of other objects. The degree of specialization of these regions for faces is breathtaking: when targeted for single-unit recording, all four regions tested extensively so far turned out to consist almost entirely of face-selective cells. These regions, termed "face patches", are located at similar positions in the two hemispheres and across individuals. Experiments combining fMRI with microstimulation demonstrate that the face patches are strongly and specifically interconnected, yet electrophysiological experiments show that different patches are functionally distinct. The face patch system offers a unique opportunity to dissect the neural mechanisms underlying form perception, because the system is specialized to process one class of complex forms, and because its computational components are spatially segregated. We believe the central experimental challenge to understanding the face patch system is to understand the functional specialization of each patch and how this specialization comes about. To address this challenge, we plan to interrogate the system from four directions: 1) Representations: What are the selectivity and invariance properties of cells in each face patch?, 2) Behavioral Role: What is the effect of inactivating specific face patches on various face-related behaviors?, 3) Connectivity: What is the wiring diagram of the face patches?, and 4) Transformations: What are the key functional differences between input and output cells within each face patch? Addressing these questions will elucidate the principles of information flow within the face patch system and shed light on the general organizing principles of inferotemporal cortex. PUBLIC HEALTH RELEVANCE: A sizable fraction of humans suffer from "prosopagnosia", a selective inability to recognize faces;our basic research on the face system in macaques aims to gain new insights concerning the brain circuits that are altered in prosopagnosics. Faces are by far the most socially significant class of visual stimuli that we perceive, and our results may shed new light on disorders of social cognition such as autism and social anxiety disorder.

This three-year grant will purchase two pieces of equipment for magnetic resonance imaging of the brain at the California Institute of Technology. One equipment piece provides the best resolution in a horizontal orientation; the second provides imaging of behaving monkeys in vertical position. This will provide state-of-the-art tools for investigating brain structure and function in monkeys with noninvasive methods, and also provide opportunities for imaging post-mortem human brains. The technology will make possible a set of research studies on how the brain processes information, including how it sees faces, how it weighs different choices, and how it makes decisions and guides action. These are important questions in neuroscience, and the new equipment will greatly enhance science at the Caltech Brain Imaging Center. The grant will also provide opportunities for training of students and post-docs on the new equipment. This will include classes taught at Caltech as well as participation in individual research projects. The development of these new scientific tools will lead to a better understanding of how the brain works, and how it is "wired up." That knowledge, in turn, will contribute to efforts to build artificially intelligent systems. Taken together, the cutting-edge science enabled by the new equipment, and the training of the next generation of young scientists on it, will contribute substantially to cognitive neuroscience in America and worldwide.

DESCRIPTION Abstract: Understanding the circuit for topological object tracking (Science Area: Neuroscience) The problem I want to solve is how an object first arises in the brain. Light hitting the retina is caried by over a million axons of the optic nerve into primary visual cortex. These are the pixels that drive visual experience. But when we look around us, we don't see pixels. We see invariant objects in space--invariant in that we perceive the objects as unchanged despite severe changes in appearance as we move around them. How does the brain stitch together pixels into invariant, discrete objects in space? The time is ripe for a fresh attack on this problem due to a critical theoretical advance, and a host of experimental advances. I believe the essential reason why no one has solved the problem of invariant object perception until now is that no one has realized the answer could be very simple. A new mathematical theory explains how the representation of objects in the 3D visual world as surfaces enables a complete and fundamentally simple solution to the problem of object segmentation and tracking, i.e., labeling all the pixels belonging to a single object over space and time, regardless of object shape. The theory strongly suggests that a powerful topological engine is churning away within very early stages of the visual cortex, to generate invariant labels for the different objects in the environment over space and time, and specifies the computations that must be performed in order to generate these invariant labels. Motivated by this new theory, and taking advantage of several key recent experimental advances in monkey and rodent vision research, I describe a set of experiments to: 1) identify the neural signature of the topological object label in early macaque visual cortical areas, 2) behaviorally test whether rodents also generate a surface representation, and if so, then 3) dissect the circuit by which this label is generated through two photon imaging an

? DESCRIPTION (provided by applicant): A dream of neuroscience is to be able to non-invasively modulate any given region of the human brain with high spatial resolution. This would open new horizons for understanding human brain function and connectivity, and create completely new options for the non-invasive treatment of brain diseases such as intractable epilepsy, depression, and Parkinson's disease. Current non-invasive brain stimulation methods such as transcranial magnetic stimulation (TMS) and transcranial electrical stimulation (TES) can be applied only to superficial cortical areas, with crude 1 cm-scale resolution, limits placed upon these techniques by fundamental physics. Ultrasonic neuromodulation, the use of ultrasound as an energy modality to affect the activity of the brain, could overcome these limitations and thereby transform both basic and clinical human neuroscience. In fact, the engineering challenge of non-invasively focusing ultrasound to mm-sized regions, either shallow or deep in the brain, has been solved: clinical studies have already demonstrated the feasibility of making focal (~ 3 mm diameter) brain lesions in subcortical regions through transcranial high intensity ultrasound. Furthermore, recent human studies have documented enhanced sensory discrimination following relatively mild ultrasound stimulation. These two findings suggest that ultrasonic neuromodulation has the potential to serve as a game-changing new tool for functional dissection of the human brain, and development of non-invasive therapies for human brain disorders. However, we believe three major questions need to be addressed before ultrasound can be used as an effective and safe tool for modulating human brain activity: (1) What are the basic biophysical mechanisms through which ultrasound acts to affect neural activity? (2) What are the optimal ultrasound parameters for maximally modulating neural activity in the primate brain? (3) How does ultrasound targeted to specific brain areas affect the spatiotemporal pattern of activity across the entire brain to causally modify behavior? We will address these three fundamental questions through a systematic effort spanning in vitro preparations, rodents, macaques, and human subjects. First, we will elucidate the endogenous mechanisms by which ultrasound produces changes in neural activity through biophysical experiments in oocytes, purified lipid bilayers, and cell cultures (Shapiro). Second, we will identify the optimal parameters for eliciting ultrasonic neuromodulation in the macaque, the closest animal model of the human brain, through EEG, fMRI, and single-unit recordings (Tsao). Finally, following initial macaque studies, we will test the effects of ultrasound stimulation on te human brain, both spatially through fMRI (O'Doherty) and temporally through EEG (Makeig), examining effects both during rest and during performance of decision-making tasks. The innovations this project will provide are exactly those called for by RFA-MH-14-217: development of breakthrough technology to measure brain processes that were formerly inaccessible to imaging, including...local and micro-circuits in the nervous system and mechanisms linking single cell or circuit activity to hemodynamic or macro-electromagnetic signals. Ultimately it's the combination of local circuit perturbation with non-invasive imaging that will give us the greatest insights into brain function. The pairing of focal ultrasound with fMRI/EEG has potential to reveal human brain circuits with unprecedented spatial resolution and create a new bridge for linking circuit activity to non -invasively measured brain signals. Our approach is only possible through intense collaboration among a unique multidisciplinary team working across model systems, and prepares the necessary experimental foundations to test whether ultrasound is the answer to the long -held dream for a technique to focally stimulate any part of the human brain at will.

? DESCRIPTION (provided by applicant): The brain is a massively interconnected network of specialized circuits. Even primary sensory areas, once thought to support relatively simple, feed-forward processing, are now known to be parts of complex feedback circuits. All brain functions depend on millisecond timescale interactions across these brain networks, but current approaches cannot measure or manipulate these interactions with sufficient resolution to resolve them. We need the capacity to measure and manipulate the activity large ensembles of neurons distributed across anatomically or functionally connected circuits. That technology does not yet exist, a lack that motivates our efforts to develop a new system for large scale, multisite recording and manipulation that takes integrates biocompatible polymer electrodes, new headstage amplifiers, a new Ethernet-based data transmission system and open source, real-time cross-platform software. This system will support recordings and manipulations across thousands of channels in awake, behaving animals as well as closed loop feedback for the next generation of experiments. Our Specific Aims are 1) To develop new high-density, double-sided polymer recording/manipulation probes, 2) To develop new high-density headstage chips, integrated electrode- headstage assemblies and surgical techniques for implanting them, and 3) To develop a low-cost, powerful data acquisition system with open-source software and real-time capabilities. We have assembled a unique team of scientists and engineers with expertise spanning polymer electrode technology, integrated electronics, real-time systems, large-scale recording, and commercial experience. Our combined expertise will allow us to create and provide to the neuroscience community an integrated system that will allow for large scale, distributed measurements and manipulation of neural activity across many sites in awake, behaving animals.

Neuronal Substrates of Hemodynamic Signals in the Prefrontal Cortex    PIs: Dr. John P. O'Doherty and Dr. Doris Tsao  Institution: California Institute of Technology PROJECT SUMMARY fMRI is the dominant technique for probing human prefrontal cortex functions in cognition, learning and decision-making. This work is predicated on the assumption that fMRI activation relates in a principled manner to the underlying neuronal activity in a given area of prefrontal cortex. Yet, virtually nothing is known about how fMRI activations relate to the underlying neural computations within the prefrontal cortex. The absence of knowledge in this domain is in contrast to burgeoning work on the relationship between measured fMRI signals and neural responses in visual areas of the brain, illuminating for instance how neuronal responses in face responsive areas give rise to fMRI activations in the temporal lobes. Compared to visual cortical areas, neurons in prefrontal cortex have more sparse, heterogeneous, and functionally distributed response characteristics, thereby rendering the relationship between neuronal and fMRI responses more enigmatic. The overarching goal of this proposal is to elucidate the relationship between neuronal computations and fMRI responses in the same areas of the prefrontal cortex. To achieve this goal we will measure fMRI activity to identify the locus of activations in prefrontal cortex while separately recording neuronal activity using a multi- electrode recording system whose placement is guided by those fMRI activations. We will also probe the neurophysiological basis of functional connectivity typically found between regions of prefrontal cortex in human fMRI studies, by recording simultaneously from multiple regions identified as being functionally connected through our fMRI measurements. We will first address these questions in macaque monkeys, and then extend our findings directly to humans, scanning healthy human participants with fMRI, and making use of a rare opportunity to obtain both intracranial electrophysiological signals and fMRI scans from the prefrontal cortex in a group of human patients undergoing evaluation for surgical treatment of epilepsy. For behavioral tasks we will draw from the domain of value-based decision-making, because (a) such tasks involve multiple regions of prefrontal cortex in both monkey electrophysiology and human fMRI studies, (b) we can use virtually identical tasks with well constrained behavior in both species, and (c) most importantly, stark discrepancies exist between what is currently known about the response properties of single neurons in the prefrontal cortex of monkeys and activations measured with fMRI during decision-making tasks in humans, which are ripe for resolving with our proposed approach. By combining across these different techniques and methodologies in both humans and monkeys, we will be able to address the question of which aspects of underlying neuronal responses gives rise to the fMRI signal in prefrontal cortex. The work will provide an essential bridge between fMRI and finer-scaled electrophysiologically-based methods for studying high order cognitive function.

Project Abstract One of the major goals of the BRAIN initiative is to develop technologies capable of interfacing with specific neural circuits in the human brain. Ultrasonic neuromodulation (UNM) is among the most significant new technologies being developed for this purpose because it has the potential to non-invasively modulate neural activity in deep-brain regions with millimeter spatial precision. This unique capability would complement existing neuromodulation and imaging techniques in basic and clinical applications. However, despite a surge of interest in UNM, the lack of knowledge about its mechanisms and recent findings of off-target sensory effects accompanying direct neuromodulation pose significant challenges to the use of this technology in human neuroscience. In particular, while most groups working on UNM are racing ahead with device development and applications, we have uncovered a major issue with this technology that must be addressed before it can be used reliably by the neuroscience community: at the ultrasonic parameters used in most UNM studies, ultrasound causes not only direct neuromodulation of the targeted region, but also strong activation of auditory cortical circuits. This significantly confounds the interpretation of UNM-evoked electrical and motor responses seen in animal models, our understanding of efficacious doses and parameters, and most importantly, potential applications in humans. To overcome this issue, we will (1) establish an understanding of the mechanisms and parameters of both direct and indirect effects of ultrasound on neural circuits, (2) identify parameters for maximizing direct modulation, and (3) develop sham stimuli enabling properly controlled use of UNM as a tool for human neuroscience. To tackle these problems, we have assembled a multidisciplinary team of scientists and engineers with expertise in tissue mechanics, acoustics, biophysics, systems neuroscience, and human psychophysics who will use unique experimental approaches ranging from computational models to specially-developed transgenic rodents and human volunteers. If successful, this project will help resolve a key issue preventing focused ultrasound from serving as a reliable, interpretable modality for non-invasive neuromodulation, and lay the groundwork for the development of optimized devices and appropriate controls for widespread use of UNM in the study of brain circuits. 

SUMMARY Controlling specific neural circuits across large areas of the brain is a major technology goal of the BRAIN Initiative. To achieve this goal, technologies should ideally provide a combination of spatial, temporal and cell- type specificity and be noninvasive to facilitate their translation across animal models and, ultimately, human patients. Here, we propose an approach to modulating neural circuits noninvasively with spatial, cell-type and temporal specificity. This approach, which we have named Acoustically Targeted Chemogenetics, or ATAC, uses transient focused ultrasound (FUS) blood brain barrier opening (BBBO) to transduce neurons at specific locations in the brain with virally-encoded engineered receptors, which subsequently respond to systemically administered bio-inert compounds to activate or inhibit the activity of these neurons. This technology allows a brief, noninvasive procedure to make one or more specific brain regions capable of being selectively modulated using orally bioavailable compounds. In preliminary experiments, we have implemented this concept in mice by using ATAC to noninvasively target AAV9 viral vectors encoding chemogenetic DREADD receptors to excitatory neurons in the hippocampus, and showing that this enables pharmacological inhibition of memory formation. Building on this proof of concept, we will now scale ATAC to work in non-human primates. This goal is particularly important given the relatively limited success of existing technologies, including optogenetics and conventional chemogenetics, in robust behavioral neuromodulation in larger animals. Scaling ATAC to larger animals requires several innovations beyond the core concept, including evolving viral vectors for more efficient and intersectional transfection of neurons with FUS-BBBO, developing ultrasound methods to overcome skull aberrations and enable precise targeting in large animals, establishing ways of confirming the functionality of ATAC non- invasively with functional imaging, and optimizing the selection and pharmacological administration of chemogenetic ligands for large-animal behavioral studies. In this project, we will first establish the basic capabilities of ATAC in NHPs and integrate them with non-invasive functional imaging, setting a baseline for ATAC performance. Then, we will use a pioneering technology for in vivo evolution of viral vectors to develop AAV viruses specifically optimized to efficiently deliver chemogenetic receptors to brain regions targeted with FUS-BBBO. In parallel, we will develop non-clinical image guidance and aberration correction methods to enable precise targeting and verification of FUS-BBBO in NHPs. This will make it possible for academic groups without access to expensive clinical FUS systems to perform ATAC in larger organisms. Finally, as motivating example applications, we will demonstrate that the optimized ATAC paradigm can be used to inhibit multiple distinct brain regions in macaques, reversibly and repeatably modulating their ability to recognize faces and also apply it in a sensorimotor circuit to alter functional connectivity. We will also show its stability, reliability and non-toxicity.

Project Summary How is the representation of complex visual objects organized in inferotemporal (IT) cortex, the large brain region responsible for object recognition? To date, areas selective for a few categories such as faces, bodies, and scenes have been found, but the vast majority of IT cortex is ?wild,? lacking any known specialization. Various schemes have been proposed for parceling IT, but a comprehensive understanding of IT organization remains elusive. Here, we propose to use fMRI, microstimulation, and electrophysiology to develop a unified understanding of the organization and coding principles of macaque IT. The experiments are motivated by a major advance in computer vision and two key preliminary results from our lab. First, the advent of deep networks trained for object classification makes it possible to generate a parametric object space, providing a quantitative framework to decipher the feature selectivity of single IT cells. Second, our preliminary results suggest that a large portion of macaque IT cortex is topographically organized according to the first two principal components of object space. This topography encompasses at least four distinct networks, each with at least three hierarchical nodes of increasing view invariance, and includes the previously described face and body patch networks. Furthermore, single cells within each network are projecting incoming objects, formatted as vectors in the object space, onto specific preferred axes. Taken together, these results suggest a new hypothesis for IT organization: IT cortex is tiled by networks (i.e., sets of functionally connected nodes, where a node is a patch of IT cells) whose organization and coding principles are very similar to that of the face patch network, and the layout of these networks follows a regular topography specified by the statistical structure of object space. We propose three Specific Aims to rigorously test this hypothesis. In Aim 1, we will systematically map all networks within IT of individual animals. In Aim 2, we will record responses of cells in each identified network to a large, common set of object stimuli and determine their coding scheme. In Aim 3, we will perturb activity in each network and quantitatively assess effect on object recognition behavior. Together, these three Aims seek to build a comprehensive understanding of IT organization that bridges fMRI, single units, and behavior. Our lab has developed powerful experimental techniques to tackle each of these Aims and has previously applied them to the macaque face patch system. We believe the time is ripe to apply these techniques to the larger problem of how all objects are represented--not just faces. In the same way that Mendeleev?s arrangement of chemical elements according to their atomic mass and chemical properties helped elucidate the electronic structure of atoms, we believe systematic mapping and characterization of all networks in IT will help elucidate the fundamental neural mechanism for object recognition.