Final Report Summary - COMPLEX3D (Neural substrates of depth perception: from surfaces to complex 3D forms)
The overarching aim of this project is to develop a novel perspective to understand how binocular disparity signals are processed throughout the dorsal and ventral visual streams. The work relies on integrating advanced methods from psychophysics, brain stimulation, neuroimaging, and advanced mathematical and computational approaches.
This work aims to:
1) provide a global analysis of the contributions of dorsal and ventral regions in the processing of different types of disparity signals using classical simplistic stimuli – extending beyond recordings in a small number of areas.
2) provide a rigorous test of the plasticity of the depth processing network – testing whether modifying the behavioural relevance of depth stimuli results in a redistribution of the brain areas underlying 3D processing, thereby revealing whether current understanding of depth processing has been distorted by the use of highly trained observers and animals.
3) advance the use of complex, biologically-relevant stimuli for the study of 3D vision, testing the possibility that previous work has used sub-optimal stimulation paradigms. This has potential for significant impact on future work in the field that will provide a more realistic understanding of the way in which the brain processes relevant signals from the environment.
We summarize our main findings to date below:
Neural substrates of depth processing.
We tested the brain areas activated by coarse and fine depth information across the visual streams using simple plane stimuli similar to those used in traditional paradigms. We measured observers’ behavioural performance and brain (fMRI) activity while they made depth judgments. We found that both forms of depth information engaged activity along early visual cortical regions and bilateral activity along both dorsal (V3d to IPS) and the ventral (V3v to LO) pathways.
Learning and reorganization.
We used learning paradigms to increase the behavioural relevance of depth stimuli and tested for the influence of training on the neural substrates underlying depth perception, while exploiting a combination of psychophysical and brain stimulation (rTMS) techniques. We used extensive behavioural testing to first establish the generality of task training. Briefly, we found that training on the fine feature task transfers substantially to the coarse signal-in-noise task, and there is a comparatively smaller transfer between tasks vice versa. Additionally, we reveal that training on the coarse signal-in-noise task benefits signal-in-noise tasks involving other visual cues (e.g. motion, orientation). In subsequent experiments, we sought to test the effects of learning on dorsal and ventral cortex in the human brain. Specifically, we applied rTMS over dorsal (left and right PPC) or ventral (left and right LO) regions, and control site (Cz) while participants made fine (feature) and coarse (signal-in-noise) depth judgments. Critically, we found a functional dissociation between parietal and LO regions for the coarse and fine tasks: for the coarse task, we found that task performance was significantly worse with stimulation over the left PPC (but not bilateral LO) versus the control site suggesting specific, parietal involvement for this task. By contrast, performance for the fine task was worse with stimulation of right LO (but not parietal cortex) versus the control site. Interestingly, we found that training alters the functional roles of these regions. After training on a fine (feature) task, performance on the coarse task is no longer affected by stimulation of parietal cortex, but is instead affected by stimulation of right LO. Performance on the feature difference task remained affected by right LO stimulation following training. These findings suggest that fine-tuning feature templates through training may facilitate noise filtering mechanisms, reducing the contribution of parietal cortex and enhancing the role of ventral regions.
The outcomes of the work have important implications for understanding the organization of the visual system, with potential impacts on the rehabilitation of brain injured patients and robotic implementations.