Research in the laboratory focuses on three areas: (1) biophysics of computation in nerve cells; (2) understanding selective, visual attention and perceptual consciousness at the neuronal, behavioral and computational levels; and (3) develop biological-motivated algorithms that predict where humans and monkeys will look at in natural scenes (gaze prediction). Publications are found here.
Research carried out as part of a "Biophysics of Computation" focuses on how the electrophysiology, synaptic architecture, and morphology of groups of individual neurons subserve information processing. What can we infer about a neuron by listening to its extracellular recorded action potential? How does the constantly fluctuating extracellular membrane potential influence the intracellular membrane potential? Can the extracellular membrane potential carry information of relevance to neural computation? We are carrying out experimental tests as well as large-scale simulations of the LFP with Henry Markram of the Blue Brain Project in Lausanne. With Guyri Buzsaki at Rutgers, we investigate the relationship between extra- and intracellular spike waveforms and the genesis of the local field potential.
For more than a decade, our laboratory collaborates with the neurosurgeon and neuroscientist Itzhak Fried at UCLA, recording from 128 electrodes in the medial temporal lobe of awake patients with pharmacologically intractable epilepsy who are implanted with depth electrodes to localize the focus of seizure onset. This unique setting allows us to observe invariant recognition, imagery and representation of familiar objects and famous individuals in conscious humans by listening in on the spiking activity of many individual neurons - complemented by local field analysis. This work requires sophisticated data processing skills and the careful design of the appropriate behavioral-physiological paradigms that will work in a clinical context.
Understanding the action of selective, visual attention (both saliency-driven, bottom-up as well as task-dependent, top-down forms) requires a firm grasp of how visual object recognition in natural scenes can be solved at the computational level, and how the resulting algorithms can be mapped onto the known architecture of the visual cortex and associated cortical and sub-cortical areas. We use analytical methods, coupled with computer simulations of the appropriate circuitry in the primate visual system, visual psychophysics, eye tracking and functional brain imaging at Caltech's 3.0 T Trio scanner to investigate human attentional selection (via saliency) and object recognition in the near-absence of focal attention, in visual search, in natural scene perception and as reward is modulated. Aspects of this work are done in collaboration with Pietro Perona, Antonio Rangel and Ralph Adolphs of Caltech and with Laurent Itti at USC. We are studying how deciding where to move the eyes can be described using decision-theoretical models. Using 'Continuous Flash Suppression' (CFS) and other techniques we have invented to hide images from conscious perception, we can show that visual, selective attention is a distinct process from visual consciousness. In collaboration with my Doktorvater, Tomaso Poggio at MIT, we are investigating neurobiologically plausible models of object recognition in static and dynamic natural scenes on the basis of both the ventral, object-recognition and the dorsal, attention visual streams. These models are also being tested for real-time performance.
We study the neuronal correlates of consciousness, elaborating a neurobiological framework to understand how subjective feelings (in particular, conscious visual perception) can arise in the mammalian forebrain. We are doing theoretical work, using integrated information theory, with Giulio Tononi at University of Madison and with Chris Adami at the Keck Graduate School.
For questions, contact our assistant,
Prof. Koch's syncretic photo (left) would be masked by the image in the middle. Adding the two pictures results in a perfectly grey image (right; except for all skin hues that remain unchanged for aesthetic reasons).
This shows the voltages in 28 neurons over 1000 times steps. The dark blue colors indicate high voltages (spikes). The large vertical yellow band indicates a neuron that has not fired. This image appears on the website for the American Physics Society (www.aps.org), courtesy of Lyle N. Long.
A neuron in the entorhinal cortex of a human patient that fired selectively to three pictures of Saddam Hussein, his written name, and his spoken name. The black and gray horizontal lines show the mean baseline firing and the threshold for defining significant responses, respectively. From Quian Quiroga et al. (2009)