Temporal lobe is one of the four major lobes in mammal brain and has several important functional regions, such as auditory cortex and fusiform face area. Specifically, the auditory cortex has been subdivided into multiple functionally specialized cortical fields ¹, and reveals cortical depth-dependent anatomical profiles ². However, functional characterization of the auditory cortex, such as the tonotopic structure, is still under debate, mostly due to small sizes of auditory cortical fields ³.

By tailoring a 24-channel temporal lobe array and pulse sequence, here we investigate the tonotopic mapping with high spatial resolution (1.5 mm isotropic) at 3T. Using laminar analysis, we found that the tonotopic maps change across laminar depths. The spatial distribution of frequency tuning at the middle layer of the auditory cortex gray matter matches the best to the result acquired by electrophysiological recordings on rodents ⁴.

A dense 24-channel surface coil array was designed and tested on a 3T MR system (Skyra, Siemens). All loop coils with 5 cm diameter were tuned to 123.25 MHz and connected to a low noise pre-amplifier integrated with a mixer. An active detuning circuit was formed using a variable inductor and a PIN diode. Coil loops were hexagonally arranged on the right-hand side of the mechanical housing (Fortus 400mc, Stratasys) fitted to the shape of a human head (Figure 1A). To mutually decouple between neighboring coils in the array, coils were critically overlapped. The noise correlation matrix was estimated using data acquired from a pulse sequence without RF excitation. SENSE g-factor maps were calculated using a GRE pulse sequence (FoV: 178x178 mm, Resolution: 0.7x0.7x5 mm, TR: 462 ms, TE: 10 ms, Flip angle: 25^o, Sagittal). SNR and time-domain SNR (tSNR) maps were calculated using an EPI pulse sequence with GRAPPA acceleration (FoV: 192x192 mm, Resolution: 1.5 mm isotropic, TR: 3000 ms, TE: 30 ms, Flip angle: 90^o, R: 3, Sagittal). The same EPI pulse sequence was used in the music and tonotopic mapping experiments, but a delay of 5 s was included in each TR in the tonotopic mapping experiment. In the music experiment, five “on” (music) and six “off” (silence) 30 s blocks were presented alternatively. In the tonotopic mapping experiment, a run consisting of five 5 s tone burst stimuli of different frequency (1/4, 1/2, 1, 2, 4 kHz) and one silence condition in a random order was repeated 20 times. All fMRI data were analyzed using the General Linear Model. Tonotopic map and laminar layer were analyzed using a previously published methods ^5,6, respectively.The Q_U/Q_L-ratio of a single isolated coil element was 237/64 = 3.70. Figure 1B shows the noise correlation matrix, which ranges between 0.4% and 56% with an average of 18%. Figure 2 shows maps of the inverse of g-factor for one-dimensional accelerations. Compared to a commercial 32-channel whole head array, our 24-channel temporal lobe array provided overall lower g-factors. Figures 3A and 3B show the SNR and tSNR maps using the 32-ch whole-head array and our temporal lobe array, respectively. Figure 3C shows the SNR vs. tSNR gain ratio maps. The results indicate that our coil array provides two-fold SNR improvement in the temporal lobe. This improvement was also found in tSNR maps as well. Figure 4A and 4B show that using the temporal lobe array, we can observe more significant BOLD signal elicited by music stimuli in the primary auditory cortex. Using the tone burst stimuli and a sparse EPI acquisition shown in Figure 5A, we obtain tonotopic maps and observe two mirror-symmetric frequency progressions (high-low-high) in the Heschl’s gyrus. Further projecting the tonotopic maps onto the intermediate cortical surface across five cortical depths, we observe the clearest frequency progressions in the middle layer of auditory cortex gray matter (Figure 5B), which is consistent with the result acquired by electrophysiological recordings on rodents.The dedicated 24-channel coil array leads to large gains in both SNR and tSNR at temporal lobe for high-resolution functional images. This SNR and tSNR advantage was traded-off for high spatial resolution in this study. Specifically, not only a more robust hemodynamic response, but also different tonotopic maps across the cortical layers were found in our study. Overall, we show that our 24-channel temporal lobe array can be a useful tool in high-resolution functional imaging for precisely delineating the functional properties of temporal lobe. In the future, we may further develop this coil array to cover bi-hemispheric temporal lobe by adding another 24 coils.