Abstract: PURPOSE: When an observer is moving in the environment, objects in the world would project on the observer’s retina and generate a dynamic light motion pattern, named optic flow. Optic flow patterns induced by forward/backward self-motion in a rigid scene contain 3D structure information as well as 2D features such as a radial velocity field. Previous studies often confuse radial motion patterns that do not contain any 3D structure information about self-motion in a rigid scene the same as optic flow. Thus, it remains in question whether the cortical areas reported by these studies are specialized to process real optic flow or radial motion patterns in general. Here, we sought to address this question by finding the neural correlates of the detection of real optic flow using a new psychophysical method.
METHODS: Two types of visual stimuli were tested: (1) The real 3D-cloud optic flow consisted of dots randomly distributed in a 3D space that moved either away or toward the observer resulting in contraction or expansion optic flow patterns, and (2) the fake optic flow that was generated by shuffling the image velocities of the dots in the 3D-cloud optic flow while keeping their initial positions and motion directions intact. Accordingly, the fake and the real optic flow stimuli were matched regarding not only the static 2D features (e.g., the radial pattern and the CoM) but also the dynamic motion signals (e.g., speed and acceleration of dot motion), except that the dot image motion in the fake optic flow was not consistent with self-motion in a rigid scene. We first conducted a psychophysical experiment in which we varied the motion coherence level (0-60%) of the stimuli by perturbing the image motion direction of a percentage of randomly selected signal dots, and participants were asked to indicate whether they perceived any coherent motion pattern. This experiment had 20 conditions: 2 types of stimuli (fake vs. real optic flow) × 2 motion directions (contraction vs. expansion) × 5 coherence levels (0%-60% with the step of 15%). We then conducted an fMRI experiment to find the cortical areas whose responses can be related to behavioral performance. This experiment adopted a block design and used the same stimuli as the psychophysical experiment. Participants were scanned for 4 sessions, with 8 runs in each session. Each session corresponded to one of the four experimental conditions: 2 types of stimuli (fake vs. real optic flow) × 2 motion directions (contraction vs. expansion). Each run had 15 stimulus blocks (5 motion coherence levels × 3 times) and 4 fixation blocks. Each stimulus block contained 16 trials of a 1-s motion stimulus at one motion coherence level and the fixation block also lasted 16 s. The testing order of stimulus was randomized in each run. Participants made a task-irrelevant judgment (color discrimination) during scanning. For each participant, we identified their visual ROIs and the ROIs previously reported to respond to radial motion stimuli (e.g., V1, V2, V3d, V3a, V3b/KO, MT, MST, V6, V7, VIP, CSv, and Pc) using standard localizers. We performed ROI-based multivoxel pattern analysis (MVPA) to examine the brain responses to contraction and expansion in fake versus real optic flow stimuli.
RESULTS: The psychophysical results showed that the detection threshold for contraction patterns was lower than that for expansion patterns for fake optic flow, but the opposite trend was found for real optic flow. The MVPA results showed that only dorsal area MST showed significantly higher decoding accuracy for contraction than expansion for fake optic flow, but this trend was reversed for real optic flow, mirroring the behavioral data.
CONCLUSIONS: The visual system does not consider non-rigid radial motion patterns (e.g., fake optic flow) the same as real optic flow. Although previous studies have reported many cortical areas in the human brain respond to radial motion stimuli, only the dorsal area MST shows neural correlates of the detection of real optic flow.

Key words: fMRI, motion, optic flow, detection sensitivity