Main Work Visual Neuroscience
Research Interests | PDF version
My thesis research interests are focused on the neurophysiology of primary visual cortex (V1) in alert and behaving monkeys. I am particularly interested in how visual stimuli are encoded in the output of cortical neurons in presence of fixational eye movements, and in receptive field analysis. The three main topics of my research are:
1. Effects of fixational eye movements on cortical visual responses
We address this issue by studying neuronal firing patterns during presentation of stationary stimuli while fixational eye movements move the neuron's receptive field on and around the stimulus. The receptive field properties were quantified using mapping with sweeping and flashing bars and gratings. We also apply a reverse correlation technique that allows localizing the receptive fields in eye position coordinates. Rigorous analysis of correlations between eye movement trajectories and spike trains, combined with characterization of receptive field properties, demonstrated that most effects of fixational eye movements on neuronal activity could be explained by interactions of a stationary stimulus and receptive field motion/position imposed by these eye movements. Some cells, that have high-pass temporal properties, respond only to fast movements imposed by fixational saccades ("saccade" cells), other, low-pass cells do not respond to brief crossing saccades but fire continuously while the stimulus is on their receptive field ("position/drift" cells), and "mixed" cells with properties intermediate between the two extremes fire bursts of activity immediately following the saccade and continue to fire at a lower rate during intersaccadic intervals. The tendency of each neuron to fire transient bursts or sustained trains of impulses following saccades is strongly correlated with the transiency of its response to stationary flashed stimuli. These results suggest that these groups of cells may carry out different functions: while "position/drift" neurons are well suited for coding spatial details of the visual scene because of their small RF size and their selectivity for sign of contrast and retinal position, saccade neurons may be more involved in temporal processing of stimulus salient features, as well as in constructing a stable world in spite of incessant retinal image motion.
We also found that fixational eye movements have a significant impact on responses of cortical neurons to sinusoidal gratings. Fast fixational saccades introduced spurious bursts or caused missing responses, while shifts in gaze, as well as slow drift, influenced the spatial relationship between the receptive field and the stimulus, subsequently causing alterations of the response temporal phase. We are presently investigating possible effects of fixational eye movements on temporal properties of the response, including interaction of eye movements with response adaptation and modulation.
We are currently extending this project to explore small voluntary saccades that are evoked by changing the position of the fixation point. Regulating the amplitude and direction of evoked saccades, we can compare neuronal responses to these saccades to responses produces by moving or flashing stimuli during steady fixation, in a more controlled manner.
2. (Pseudo)linear and nonlinear properties of V1 neurons
In this project, we stimulated cortical receptive fields using an "image stabilization" technique. The eye position signal from the eye tracker was added to the stimulus position signal to compensate for changes in eye position during stable fixation. Although this feedback loop is too slow to compensate for the fast fixational saccades, it minimizes the eye movement effects, allowing us to study the detailed receptive fields' structure during normal vision, in spite of inevitable gaze shifts. Further (off-line) corrections for eye movements include an automatic selection of relatively stable inter-saccadic fixation periods, and alignment of the responses in time or frequency domain to compensate for residual phase shifts.
We were interested in the relationship between spatial organization of receptive fields and responses to sinusoidal luminance gratings. Using spatial mapping with stabilized sweeping and flashing bars, we confirmed the basic simple/complex dichotomy previously established in anesthetized monkeys and cats. We also found that the majority of cells (about 80%) in V1 of alert monkeys are complex cells with overlapping increment- and decrement-responsive activating regions (ARs). However, complex cells exhibited a diverse mixture of pseudolinear and nonlinear responses to drifting sinusoidal gratings that can not be predicted from their spatial organization. Many of complex cells had a significant pseudolinear (F1, fundamental) component in the response to drifting gratings. In fact, a subset of complex cells had the relative modulation RM=F1/F0>1, the traditional criterion for identifying simple cells, and corrections for eye movements further increased the RM ratio. However, unlike simple cells, even those complex cells with high F1 modulation could exhibit diverse nonlinear responses when the spatial frequency, temporal frequency, or window size was changed. Furthermore, the responses of complex cells to counterphase (contrast-reversal) gratings were predominantly nonlinear even harmonics, and complex cells gave on-off responses to stationary flashing bars and moving edges. Thus, complex cells generate pseudolinear responses only within a certain range of parameters space, in contrast to quasilinear simple cells, which exhibit robust F1 modulation to all effective gratings.
These results demonstrate that in complex cells, the form of the response (and not only the response amplitude) is strongly parameter- and context-dependent, and they raise interesting questions about underlying mechanisms and their functional significance. They also indicate that complex cells are the most frequently encountered neurons in primate V1, and their behavior needs to receive more emphasis in models of visual function.
3. Modeling characteristic responses of V1 neurons in alert monkey
The analysis of the findings described in the previous section is incomplete without an attempt to model the observed diversity of cortical responses. The main objective of this model is to capture the characteristic properties of complex cells in response to a broad range of visual stimuli, including sweeping and flashing bars and drifting, stationary and contrast-reversal gratings, and to generate testable predictions and clues for further analysis of their receptive fields. This iterative process will hopefully lead to a better understanding of complex cells, which are the major physiological cell type in V1 of alert monkey. Ideally, the model should not only describe the "local" complex cell properties, but also be considered within a global framework of the functional processing in the primate visual system.
The model's main challenge is to explain the dependence of grating responses on stimulus attributes and how the pseudolinear responses can be generated in fundamentally nonlinear complex receptive fields. No existing complex cell model, including modifications of a popular energy model, can account for the variety of behaviors exhibited by complex cells. Our results suggest an alternative complex cell model based on elaborate interactions between increment and decrement ARs and surround.
Other developing project
1. Analysis of multi-units recordings, obtained with single electrodes (we use PCA spike sorting algorithm to extract single-unit spike trains). We are trying to explore patterns of connectivity between neighboring cells and effects of eye movements on correlation of the responses. We employ standard cross-correlation analysis as well as joint peristimulus time histograms (JPSTH).
2. Receptive fields "white noise" analysis in alert monkeys. This type of analysis, which had proven very informative in anesthetized preparation, is very difficult to implement in alert monkey, because of eye movements and time limitations imposed on data collection. We will attempt to overcome these problems by using online estimation of kernels that would provide a feedback to the stimulus generation process, yielding most efficient pseudorandom stimulus sequences.
3. Extension of simple stimuli to more realistic "natural" images and contextual effects. We are aiming to predict/explain responses to natural scenes using the detailed characterization of receptive field spatiotemporal properties. As a variant, we will embed a simple stimulus into a pseudo-natural surround and will gradually modify the complexity or/and "naturalness" (using natural scenes statistics) of the surround, in order to see how responses would change as we go from a simple stimulus to the extended complex scene.
General/future research interests
My research interests are not limited to primary visual cortex, and include many aspects of visual functions that are not directly related to my thesis work. For example, I am interested in:
My other scientific interests include programming experimental systems, data analysis and visualization. In order to accomplish my research goals, I have developed a complete visual neuroscience system that includes spikes and eye positions data acquisition with online visualization, monkey behavior control and feedback, and visual stimulus presentation. For data analysis I use custom software, mostly written in MATLAB.
My "scientific hobby" is social and behavioral aspects of primatology.