| Start Date: | 4/5/2016 | Start Time: | 11:00 AM |
| End Date: | 4/5/2016 | End Time: | 1:00 PM |
|
Event Description
BIOMED PhD Thesis Defense
Title:
Contour Integration via Cortical Interactions in Visual Cortex
Speaker:
Xiajing Gong, PhD Candidate, School of Biomedical Engineering, Science and Health Systems
Advisor:
Hualou Liang, PhD, School of Biomedical Engineering, Science and Health Systems
Abstract:
The visual system possesses a remarkable ability to group fragmented line segments into coherent contours and to segregate them from background. This process, known as contour integration, is critical to identifying object boundaries in complex visual scenes, and thus particularly important for performing shape discrimination, image segmentation and ultimately object recognition. Current evidence supports the idea that long-range horizontal connections in early visual cortex contribute to the process of contour integration, but the underling cortical circuitry, particularly the top-down feedback influence from higher visual areas, is not fully understood. Throughout the thesis, I took computational approaches to systematically examine how contour information is represented across the network of cortical areas and the circuitry by which this information is encoded.
Three closely related projects, each having new methods development and hypothesis testing, were performed to analyze and interpret a very large set of neural data. The data set consists of recently acquired multi-electrode multi-unit spikes and local field potentials (LFP) simultaneously recorded in visual areas V1 and V4 of monkeys performing a visual contour detection task. In the first project, I extended the well-established Granger causality measure to the analysis of spiking trains data, which enabled me to quantify the causal interactions within and between areas V1 and V4. Our findings provided the first demonstration that there is a top-down V4 feedback influence upon early visual area V1 during contour integration. In the second project, I investigated whether the contour signals in V1 are derived from feedback inputs alone, or whether they are mediated by an intimate interaction between feedback and horizontal connections within V1. I developed conditional causality measure to dissect the respective contributions of V1 horizontal connections and V4 feedback to contour grouping. Our results suggest that feedback and lateral connections closely interact to mediate the contour integration process. In the third project, I proposed a novel Granger causality measure for the analysis of mixed neural data of spikes and LFP. Spikes and LFP are generated by separate sources with distinct signal characteristics. A joint analysis of spikes and LFP was performed to address the fundamental question about how contour regulates cortical communication between individual neurons and local network activity. The results conform to the general input-output relationship between LFP and spikes within an area. Importantly, I found that contour-related activity is only observed from spikes to LFP, but not in the opposite direction. These findings suggest that Granger causality from spikes to LFP, rather than that from LFP to spikes, carries contour-related information.
Taken together, these results indicate that cortical interactions underlie contour integration, thus contribute to a better understanding of the cortical circuitry for parsing visual images and for sensory processing in general. Given the increasing use of multi-electrode recordings in multiple cortical areas, the methodology developed in this thesis should also have a broad impact. |
|
Location:
Monell Chemical Center, 1st Floor Conference Room, located at 3508 Market Street. |
Audience:
Undergraduate StudentsGraduate StudentsFacultyStaff |
|
