Gamma activity modulated by naming of ambiguous and unambiguous images: intracranial recording

Abstract

Objective: Humans sometimes need to recognize objects based on vague and ambiguous silhouettes. Recognition of such images may require an intuitive guess. We determined the spatial-temporal characteristics of intracranially-recorded gamma activity (at 50-120Hz) augmented differentially by naming of ambiguous and unambiguous images.

Methods: We studied 10 patients who underwent epilepsy surgery. Ambiguous and unambiguous images were presented during extraoperative electrocorticography recording, and patients were instructed to overtly name the object as it is first perceived.

Results: Both naming tasks were commonly associated with gamma-augmentation sequentially involving the occipital and occipital-temporal regions, bilaterally, within 200ms after the onset of image presentation. Naming of ambiguous images elicited gamma-augmentation specifically involving portions of the inferior-frontal, orbitofrontal, and inferior-parietal regions at 400ms and after. Unambiguous images were associated with more intense gamma-augmentation in portions of the occipital and occipital-temporal regions.

Conclusions: Frontal-parietal gamma-augmentation specific to ambiguous images may reflect the additional cortical processing involved in exerting intuitive guess. Occipital gamma-augmentation enhanced during naming of unambiguous images can be explained by visual processing of stimuli with richer detail.

Significance: Our results support the theoretical model that guessing processes in visual domain occur following the accumulation of sensory evidence resulting from the bottom-up processing in the occipital-temporal visual pathways.

Keywords: Blurred images; Electrocorticography (ECoG); Event-related potentials (ERPs); High-frequency oscillations (HFOs); Low spatial frequencies; Pediatric epilepsy surgery; Ripples; Top-down processing; Visual recognition.

Figure 1. Examples of visual stimuli

(A) and (B) were included in the unambiguous image set, while (C) and (D) in the ambiguous set. (C) was named, for example, as ‘fish’, ‘whale’ or ‘submarine’. (D) as ‘human’, ‘hair’, or ‘pineapple’.

Figure 2. Gamma-augmentation during naming tasks in patient #5

(A) Green electrodes: sites showing significant gamma-augmentation elicited by naming of unambiguous and ambiguous images. Blue electrodes: significant gamma-augmentation elicited by naming of unambiguous images alone. Red electrodes: significant gamma-augmentation elicited by naming of ambiguous images alone. White electrodes: no significant gamma-augmentation elicited by either naming task. (B) Time-frequency plots are presented with the zero point reflecting the onset of image presentation. Channel #1 in the left occipital pole showed significant gamma-augmentation within 100 ms following the onset of presentation of unambiguous and ambiguous images; naming of unambiguous images was associated with more sustained gamma-augmentation. Channel #2 in the lateral occipital region showed significant gamma-augmentation within 200 ms following the presentation of unambiguous images, while gamma-augmentation elicited by ambiguous images failed to reach significance. Channel #3 in the left inferior parietal lobule was associated with significant gamma-augmentation at 810–940 ms following the presentation of ambiguous images but no significant gamma-augmentation during naming of unambiguous ones. Channel #4 in the left orbitofrontal region was associated with significant gamma-augmentation at 860–1190 ms following the presentation of ambiguous images but no significant gamma-augmentation during naming of unambiguous ones. (C) Time-frequency plots are presented with the zero point reflecting the onset of responses. Channel #4 in the left orbitofrontal region and channel #5 in the left inferior-frontal region were associated with significant gamma-augmentation prior to responses to ambiguous images alone. Channel #6 in the left Rolandic region was associated with significant gamma-augmentation around the onset of responses to both ambiguous and unambiguous images. The mean response time was 1930 ms during naming of ambiguous images and 1140 ms during that of unambiguous ones.

Figure 3. Gamma-augmentation during naming tasks in patient #8

(A) Green electrodes: sites showing significant gamma-augmentation elicited by naming of unambiguous and ambiguous images. Blue electrodes: significant gamma-augmentation elicited by naming of unambiguous images alone. Red electrodes: significant gamma-augmentation elicited by naming of ambiguous images alone. White electrodes: no significant gamma-augmentation elicited by either naming task. (B) Time-frequency plots are presented with the zero point reflecting the onset of image presentation. Channel #1 in the right polar occipital region showed significant gamma-augmentation within 100 ms following the onset of presentation of unambiguous and ambiguous images; naming of unambiguous images was associated with more sustained gamma-augmentation. Channel #2 in the lateral occipital region showed significant gamma-augmentation within 200 ms following the presentation of unambiguous images, while gamma-augmentation elicited by ambiguous images failed to reach significance. Channel #3 in the right inferior frontal region was associated with significant gamma-augmentation at 410–530 ms following the presentation of ambiguous images, but gamma-augmentation during naming of unambiguous ones failed to reach significance. (C) Time-frequency plots are presented with the zero point reflecting the onset of responses. Channel #4 in the right Rolandic region was associated with significant gamma-augmentation around the onset of responses to both ambiguous and unambiguous images. The mean response time was 2580 ms during naming of ambiguous images and 1750 ms during that of unambiguous ones.

Figure 4. Gamma-augmentation elicited by naming of ambiguous and unambiguous images

(A) Presented is the definition of the anatomical regions of interest; this was employed in our previous study (Matsuzaki et al., 2012; Kojima et al., 2013a). Medial occipital region: medial portion of BA 17/18. Polar occipital region: polar portion of BA 17/18. Lateral occipital region: lateral portion of BA 19/37. Inferior occipital-temporal region: inferior portion of BA 19/37. Medial temporal region: BA 27/28/34/35/36. Inferior temporal region: inferior temporal gyrus involving BA 20/37. Middle temporal region: middle temporal gyrus involving BA 21/37. Superior temporal region: BA 22/41/42. Inferior parietal region: BA 39/40. Inferior Rolandic region: BA 4/3/1/2 not more than 4 cm superior from the sylvian fissure. Medial frontal region: medial portion of BA 6/8 and posterior portion of BA 24/32/33. Dorsolateral premotor region: dorsolateral portion of BA 6. Middle/superior frontal region: lateral portion of BA 46/9/8. Inferior frontal region: inferior frontal gyrus involving BA 44/45. Orbitofrontal region: BA 47/11. (B) Red dots: electrode sites showing significant gamma-augmentation elicited by naming of ambiguous images alone. Blue dots: sites showing significant gamma-augmentation elicited by naming of unambiguous images alone. Green dots: sites showing significant gamma-augmentation elicited by both naming tasks. Black dots: sites failing to show significant gamma-augmentation during either naming task.

Figure 5. Gamma-augmentation specifically elicited by naming of ambiguous images

Both on common average and bipolar montages, electrodes F1 and F2 showed significant gamma-augmentation immediately prior to the responses during naming of ambiguous images. It has been reported that bipolar montage can effectively eliminate ocular EMG-elicited gamma-augmentations (Jerbi et al., 2009; Nagasawa et al., 2011; Uematsu et al., 2013).

Figure 6. Models of intuitive guess

A theoretical model (upper) proposes that the lower-order visual region rapidly projects the information of blurred images to the orbitofrontal region possibly via the magnocellular pathway, and deliver a top-down signals back to the occipital-temporal region including the fusiform gyrus to facilitate visual recognition (Bar et al., 2006; Kveraga et al., 2007). The present study supports the model (lower) that guessing processes in the visual domain take place in the frontal-parietal regions (not necessarily confined to the orbitofrontal regions) following the accumulation of sensory evidence resulting from the bottom-up processing between the lower- and higher-order visual pathways (Ploran et al., 2007; Heekeren et al., 2008).