Direct Exploration of the Role of the Ventral Anterior Temporal Lobe in Semantic Memory: Cortical Stimulation and Local Field Potential Evidence From Subdural Grid Electrodes

Abstract

Semantic memory is a crucial higher cortical function that codes the meaning of objects and words, and when impaired after neurological damage, patients are left with significant disability. Investigations of semantic dementia have implicated the anterior temporal lobe (ATL) region, in general, as crucial for multimodal semantic memory. The potentially crucial role of the ventral ATL subregion has been emphasized by recent functional neuroimaging studies, but the necessity of this precise area has not been selectively tested. The implantation of subdural electrode grids over this subregion, for the presurgical assessment of patients with partial epilepsy or brain tumor, offers the dual yet rare opportunities to record cortical local field potentials while participants complete semantic tasks and to stimulate the functionally identified regions in the same participants to evaluate the necessity of these areas in semantic processing. Across 6 patients, and utilizing a variety of semantic assessments, we evaluated and confirmed that the anterior fusiform/inferior temporal gyrus is crucial in multimodal, receptive, and expressive, semantic processing.

Keywords: anterior fusiform; basal temporal language area; semantic memory; subdural electrodes; ventral anterior temporal lobe.

Figure 1.

Summaries of the strong convergence of results from the LFPs (A—Investigation 1), direct stimulation studies (B—Investigations 2–4), and previous functional neuroimaging investigations of verbal and nonverbal semantic processing (peak activations are plotted in C). A critical area within the vATL was located around the anterior fusiform gyrus (3.8–5.2 cm from the temporal pole). Representative LFP waveforms (average waveforms from all of the naming session) are shown for each patient (A) taken from the same vATL region as that identified in the direct stimulation Investigations 2–4 (B) and aligning with the peaks from semantic functional neuroimaging studies (Spitsyna et al. 2006; Binney et al. 2010; Visser et al. 2011) (C). The LFP waveform consisted of a negative activity with a peak latency of 230–300 (Patients 1, 3, and 5) and the subsequent positive/negative activity peaking at 310–450 ms (1–5). Patient 6 showed a negative activity with a peak latency around 300 ms, similar to left cases.

Figure 2.

(A) The results of language mapping with conventional 50 Hz cortical stimulation of vATL sites in Patients 1–5. Small black dots represent subdural electrodes investigated with cortical stimulation. Electrodes showing any language impairment are illustrated with large circles with 5 segments (one for each task). Filled segments denote impairment of the corresponding task by stimulation. A black fill denotes arrest by stimulation and gray fill denotes slowing or errors. The gray ovals mark the core language electrodes used as the site for the targeted stimulation Investigations 3 and 4 in these 3 patients. In this target area, high-frequency stimulation generated impairments across all the semantically related tasks (both receptive and expressive), but spared kana word reading (which requires visual processing and speech production but not access to word meaning, unlike the other tasks), consistent with the conclusion that this region is crucial for pan-modal semantic processing. Clinical stimulation of vATL in Patients 3 and 6 was partial, inconclusive and had to be discontinued due to dural pain or other symptoms. For clarity, only the subdural grid (4 × 5 electrodes) in the basal temporal area is shown. (B) The performance status for each task during vATL stimulation (Patients 1, 2, 4, and 5). Picture naming, kanji word reading, and spoken verbal command task were consistently impaired upon stimulation while kana word reading was rarely impaired.

Figure 3.

Effect of varying anterior fusiform stimulation intensity on naming and written word-picture matching performance. (A) The direct comparison of performance before and after stimulation for the 2 tasks in Patients 1 and 2. The same pattern was observed in both RT and accuracy, though RT was more sensitive overall to the effect of milder stimulation intensities (see the main text). As expected, performance declined (slower RTs and lower accuracy) as stimulation intensity increased. (B) A direct comparison (by normalization to the sham condition—see Experimental Procedures) of the relative effect of stimulation intensity on the 2 tasks for both patients. As expected, stimulation had a relatively greater impact on naming than word-picture matching (see the main text).

Figure 4.

LFPs and results of direct cortical stimulation on synonym judgment, naming, and number judgment in Patient 4 and 5. (A) Representative LFP waveforms (average waveforms from the first and second halves of the synonym judgment session to show the stability of the data) taken from the same vATL region as that used for the direct stimulation investigation (Patient 4—electrodes D13; Patient 5—electrodes E13) and aligning with the peaks from the naming LFP studies (Fig. 1). (B) Both patients' RT data across the 3 tasks with and without concurrent (7 mA/3 s) stimulation. Synonym decision times were substantially and significantly slowed for all 3 types of word [high-frequency concrete (HF), low-frequency concrete (LF), and low imageability/abstract (LI)]. Accuracy was also substantially reduced (see the main text). As found in Investigation 3 for Patients 1 and 2 (Fig. 3), stimulation at this same site also generated significant and considerable slowing of picture naming. In contrast, however, neither speed nor accuracy of difficulty (RT)-matched number judgments were affected, suggesting a semantically selective effect of stimulation at this site.