Date: Thursday, Feb. 2nd
Time: 4pm
Location: HNB small conference room.
Speaker: Susmita Chatterjee
Paper: Ishikane H, Gaugi H, Honda S, Tachibana. Synchronized retinal oscillations encode essential information for escape behavior in frogs. Nat Neurosci. 2005 Aug 8(8):1087-95.

Synchronized oscillatory activity is generated among visual neurons in a manner that depends on certain key features of visual stimulation. Although this activity may be important for perceptual integration, its functional significance has yet to be explained. Here we find a very strong correlation between synchronized oscillatory activity in a class of frog retinal ganglion cells (dimming detectors) and a well-known escape response, as shown by behavioral tests and multi-electrode recordings from isolated retinas. Escape behavior elicited by an expanding dark spot was suppressed and potentiated by intraocular injection of GABA(A) receptor and GABA(C) receptor antagonists, respectively. Changes in escape behavior correlated with antagonist-evoked changes in synchronized oscillatory activity but not with changes in the discharge rate of dimming detectors. These antagonists did not affect the expanding dark spot-induced responses in retinal ganglion cells other than dimming detectors. Thus, synchronized oscillations in the retina are likely to encode escape-related information in frogs.

Date: Thursday, Feb. 9th
Time: 4pm
Location: HNB small conference room.
Speaker: Junkwan Lee
Paper: Tu Z., Chen X., Yuille A.L., Zhu S.C. Image Parsing: Unifying Segmentation, Detection, and Recognition. Int’l J. of Computer Vision, Marr Prize Issue, 2005.

We propose a general framework for parsing images into regions and objects. In this framework, the detection and recognition of objects proceed simultaneously with image segmentation in a competitive and cooperative manner. We illustrate our approach on natural images of complex city scenes where the objects of primary interest are faces and text. This method makes use of bottom-up proposals combined with top-down generative models using the Data Driven Markov Chain Monte Carlo (DDMCMC) algorithm which is guaranteed to converge to the optimal estimate asymptotically. More precisely, we define generative models for faces, text, and generic regions- e.g. shading, texture, and clutter. These models are activated by bottom-up proposals. The proposals for faces and text are learnt using a probabilistic version of AdaBoost. The DDMCMC combines reversible jump and diffusion dynamics to enable the generative models to explain the input images in a competitive and cooperative manner. Our experiments illustrate the advantages and importance of combining bottom-up and top-down models and of performing segmentation and object detection/recognition simultaneously.

Longer version of the paper with math

Date: Thursday, Feb. 16th
Time: 4pm
Location: HNB small conference room.
Speaker: Alan Horsager

A.Horsager1,2,3*; S.H.Greenwald4; M.S.Humayun1,3; M.J.McMahon4; I.Fine1,2

1. Dept. of Ophthalmology, 2. Zilkha Neurogenetic Inst., 3. Doheny Retina Inst., Univ. of Southern California, LA, CA, USA 4. Second Sight Med. Products, LLC, Sylmar, CA, USA

Purpose: Retinitis pigmentosa (RP) and age-related macular degeneration are diseases involving progressive degeneration of photoreceptors. However, the bipolar and ganglion cells remain relatively spared, though neurally disorganized (Marc, 2004). Recently, a collaborative effort between Second Sight LLC and Doheny Eye Institute has implanted six advanced RP patients with a 4×4 array of epiretinal electrodes. Achieving functional perception using electrical stimulation will require relatively high frequency pulse trains that produce continuous percepts. We report here psychophysical data
from three patients examining the perceptual effect of pulse train
stimulation. Methods: Stimuli were pulse trains consisting of 0.075 or 0.975 ms cathodic-first charge balanced biphasic pulses. The inter-phase interval between the two pulses varied between 0.075 and 0.975 ms. The frequency of the pulse trains varied between 5 and 225 Hz. Current amplitude thresholds were measured using a yes-no paradigm where 50% of the trials were catch trials containing no stimulus. Pulse train amplitude was varied using a three down one up staircase. Results: An increase in pulse frequency leads to a decrease in the current needed to reach perceptual threshold. However the amount of total charge needed to reach threshold increases linearly as a function of frequency. This decrease in efficiency as a function of frequency is more pronounced for long than for short pulses, and does not depend on the inter-phase interval. Conclusions: One possibility is that short pulses result in less sub-threshold deactivation of ion channels. Alternatively, this effect of frequency may be due to inhibition mediated by bipolar/amacrine cells. Inhibition from bipolar and amacrine cells might be weaker for short pulses since these cells tend to be preferentially stimulated by longer pulses (Jensen, 2005).
Support Contributed By: Second Sight Medical Products, LLC

Date: Thursday, Feb. 23rd
Time: 4pm
Location: DRB conference room (note the location change).
Speaker: Jeff Wurfel
Paper: Zaksas D, Pasternak T. Area MT neurons respond to visual motion distant from their receptive field J Neurophysiol. 2005, Dec 94(6), 4156-67.

Neurons in cortical area MT have localized receptive fields (RF) representing the contralateral hemifield and play an important role in processing visual motion. We recorded the activity of these neurons during a behavioral task in which two monkeys were required to discriminate and remember visual motion presented in the ipsilateral hemifield. During the task, the monkeys viewed two stimuli, sample and test, separated by a brief delay and reported whether they contained motion in the same or in opposite directions. Fifty to 70% of MT neurons were activated by the motion stimuli presented in the ipsilateral hemifield at locations far removed from their classical receptive fields. These responses were in the form of excitation or suppression and were delayed relative to conventional MT responses. Both excitatory and suppressive responses were direction selective, but the nature and the time course of their directionality differed from the conventional excitatory responses recorded with stimuli in the RF. Direction selectivity of the excitatory remote response was transient and early, whereas the suppressive response developed later and persisted after stimulus offset. The presence or absence of these unusual responses on error trials, as well as their magnitude, was affected by the behavioral significance of stimuli used in the task. We hypothesize that these responses represent top-down signals from brain region(s) accessing information about stimuli in the entire visual field and about the behavioral state of the animal. The recruitment of neurons in the opposite hemisphere during processing of behaviorally relevant visual signals reveals a mechanism by which sensory processing can be affected by cognitive task demands.