Positron Emission Tomography in Cochlear Implant and Auditory Brain Stem Implant Recipients

Abstract

Objective

The purpose of this study was to determine whether similar cortical regions are activated by speech signals in profoundly deaf patients who have received a multichannel cochlear implant (CI) or auditory brain stem implant (ABI) as in normal-hearing subjects.

Study Design

Positron emission tomography (PET) studies were performed using a variety of discrete stimulus conditions. Images obtained were superimposed on standard anatomic magnetic resonance imaging (MRI) for the CI subjects. The PET images were superimposed on the ABI subject’s own MRI.

Setting

Academic, tertiary referral center.

Patients

Five subjects who have received a multichannel CI and one who had received an ABI.

Intervention

Multichannel CI and ABI.

Main Outcome Measure

PET images.

Results

Similar cortical regions are activated by speech stimuli in subjects who have received an auditory prosthesis.

Conclusions

Neuroimaging provides a new approach to the study of speech processing in CI and ABI subjects.

Positron emission tomography (PET) is an imaging technique that delineates functional neurochemical activity in the brain. Positron-emitting isotopes are injected into the blood stream and are concentrated in areas of the body where metabolic activity is high. The PET devices detect the patterns of photon emission for the target organ.

A number of cortical regions, in addition to the primary auditory cortex, are activated by speech stimuli in normal-hearing subjects. The purpose of this study was to determine whether similar networks are activated in profoundly deaf patients who have received a multichannel cochlear implant (CI) or an auditory brain stem implant (ABI). The PET studies were conducted in five normal-hearing subjects, five CI subjects, and one ABI subject to measure changes in regional cerebral blood flow evoked by acoustic stimulation.

MATERIALS AND METHODS

Patients

Five normal-hearing adults and five adult CI patients and one ABI patient served as subjects. All CI subjects used a multi-channel CI device. Three subjects used a Nucleus 22-channel CI (Cochlear Corporation, Englewood, CO, U.S.A.) and two subjects used a Clarion CI (Advanced Bionics, Sylmar, CA, U.S.A.). All of the CI users have significant open-set abilities, as shown on the NU-6 word lists (1) and CI device everyday sentences (2). Inclusion in this study was based on scores greater than 15% on the NU-6 and greater than 20% on the CI device tests. The ABI subject used a Nucleus ABI (Cochlear Corporation, Englewood, CO, U.S.A.).

Positron emission tomography scanning procedure

The PET studies were performed using a Siemens 951/31R, which measures 31 image planes simultaneously over an axial field-of-view of 10.8 cm. The intrinsic image resolution of this system is approximately 6.0-mm full width at half maximum in plane and 5.0-mm full width at half maximum in the axial direction.

The subjects were blindfolded during each scan when acoustic stimuli (70–75-dB sound pressure level) were presented monaurally to the right ear in the normal-hearing subjects or to the implanted ear in the CI subjects. Stimuli were delivered to the right ear of the normal-hearing subjects through an earphone with the contralateral ear occluded with a foam earplug. Stimuli were delivered from a loudspeaker to the microphone of CI subjects listening with their speech processor turned on.

The PET scanning was carried out using discrete stimulus conditions, with each condition repeated up to a total of eight scanning periods. Repetition of stimulus conditions permits averaging data within subjects. In each PET session, the first scan was obtained with no acoustic stimulation. This silence (baseline) condition was subtracted from all acoustic-stimulus conditions. The following sequence of acoustic-stimulus conditions was used: broadband noise, multitalker babble (3), words from the multisyllabic lexical neighborhood test (4), and the common phrases test (5).

Fifteen seconds after each stimulation task was begun, 50 mCi of H2150 was injected intravenously as a bolus, and tomographic image acquisition was begun concurrently with bolus injection and continued for 3.5 minutes. A rapid sequence of scans was performed over the 3.5-minute interval to enable the selection of a 90-second time window beginning 35 seconds to 40 seconds after the bolus arrived in the brain. The subject continued the task for an additional 15 seconds following the end of the PET data acquisition. The time between scanning periods was approximately 12 minutes to allow for radioactive decay to less than 2% of administered levels.

The identification of regions of significant brain activation was indicated using a four-step process that included 1) image registration, 2) global normalization of the image volume data, 3) identification of the intercommissural (anterior commissure-posterior commissure) line on either the magnetic resonance imaging (MRI) data or on an intrasubject averaged PET image set for stereotaxic transformation and alignment, and 4) averaging of subtraction images across subjects and statistical testing for identification of brain regions demonstrating task-specific significant changes in blood flow. For data interpretation, a peak analysis based on the t-statistic is performed using a pooled variance derived from the image data.

Images for each condition were averaged across all subjects in the same group and analyzed using the image subtraction method. Regions of statistically significant increases in activation were mapped in Talairach stereotaxic coordinates. The image-analysis software displays the distribution pattern of activated regions for a subject group after filter smoothing by at least 10 mm. The centroids of peak activity have a spatial resolution of 1 mm to 2 mm.

To spatially align the PET and MRI, standard anatomic MRIs were segmented to identify gray matter, white matter, and cerebrospinal fluid. Pixel intensity values were set in each of the segmented regions to expected PET radionuclide levels for each tissue type and the data smoothed to the same resolution as the PET data. The image volumes are then registered as rigid bodies (three translation parameters, three rotation parameters, and one scale parameter) using Newton’s method based on a least-squares cost function. The registration parameters derived from this algorithm were then applied to the original MRI data to map the PET results onto the anatomic images.

RESULTS

Stimuli perceived as speech by both normal and postlingually deaf CI subjects evoked strong bilateral activation in the superior temporal gyri, which are regions that include the primary and secondary auditory cortices (Figs. 1–3). Broadband noise did not significantly activate the auditory cortex in normal-hearing subjects. Bilateral activation of the immediate and secondary association areas in the auditory cortices was demonstrated in the ABI subject (Fig. 4).

FIG. 1.

Word identification. Subtracted composite positron emission tomography images superimposed on standard magnetic resonance images. (A) Normal-hearing subjects listening to words demonstrated bilateral activation of the superior temporal gyrus (area 22), with a larger activation on the left side. Cross-hairs at peak activation on the left (−57, −13, 2−), Z-score = +5.2. (B) Cochlear implant subjects listening to words demonstrated strong bilateral activation of the superior temporal gyrus (area 22). Cross-hairs at peak activation on the right (55, −22, 4), Z-score = +6.2. A similar activation of this cortical area was found on the left, Z-score = 6.0.

FIG. 3.

Multitalker babble. Subtracted composite positron emission tomography images superimposed on standard magnetic resonance images. (A) Normal-hearing subjects listening to multitalker babble demonstrated activation of the left superior temporal gyrus (area 22). Cross-hairs at peak activation on the left (−55, −19, 2), Z-score = +3.7. This is not statistically significant. (B) Cochlear implant subjects listening to multitalker babble demonstrated strong bilateral activation of the superior temporal gyrus (area 22). Cross-hairs at peak activation on the left (−55, −22, 4) between area 42 (auditory association area) and area 22, Z-score = +6.3. Activation of the right superior temporal gyrus is in a similar location (53, −19, 4), Z-score = +6.2.

FIG. 4.

Word identification. Auditory brain stem implant subject. Positron emission tomography image, overlaying patient’s own magnetic resonance imaging scan, shows bilateral activation of the immediate and secondary association areas in the auditory cortices.

DISCUSSION

Complex speech is initially processed by undergoing acoustical analysis in the primary auditory area and in the immediate auditory association cortex and is then relayed to the lexicon areas of the brain to be matched with encoded entries for familiar words (6). Bilateral activation of extensive areas of both the primary auditory area and the immediate auditory association area has been documented in the PET study by Howard et al. (7).

Despite the nonphysiologic speech coding strategies used by current CI systems, postlingually deaf CI recipients appear to process speech sound in an analogous manner to normal-hearing subjects. As shown by the current study and previous studies by Ito et al. (8) and Naito et al. (9, 10), similar auditory cortical areas are stimulated by speech and speechlike stimuli in postlingually deaf patients who have received a multichannel CI as in normal-hearing subjects.

In the current study, although activation included the auditory cortex, the peak activation was found in higher auditory association areas in the superior temporal gyrus (e.g., Brodmann area 22). Interestingly, multitalker babble was more effective in activating the higher auditory cortical areas in CI than in normal-hearing subjects. Apparently, normal-hearing subjects can readily distinguish a cacophony of sound from a meaningful speech signal, whereas CI subjects found enough similarity in multitalker babble to process the signal as speechlike. Normal-hearing subjects were likewise able to ignore noise to the extent that the PET image was not statistically significant. Cortical areas unique to the perception of isolated words or common phrases appear to be similar.

Although the ABI patient does not demonstrate open speech understanding abilities, activation of the auditory cortical regions was shown. This suggests that he not only processes speech and nonspeech signals in a similar manner to normal-hearing subjects but is capable of accomplishing this with minimal acoustic information.

CONCLUSION

Despite the inherent limitation of providing indirect information regarding the metabolic processes in the central nervous system, neuroimaging provides a new approach to the study of speech processing in CI and ABI subjects.

FIG. 2.

Common phrases. Subtracted composite positron emission tomography images superimposed on standard magnetic resonance images. (A) Normal-hearing subjects listening to common phrases demonstrated strong bilateral activation of the superior temporal gyrus (area 22), with somewhat larger activation on the left. Cross-hairs at peak activation on the left (−55, −24, 2), Z score = +5.6 (Z-score = 4.6 on the right). (B) Cochlear implant subjects listening to common phrases demonstrated strong bilateral activation of the superior temporal gyrus (area 22). Cross-hairs at peak activation on the right rostral half of the superior temporal gyrus (51, 3, −7), Z-score = +6.1. Similar activation was seen on the left side, Z-score = +5.8.

TABLE 1.

Clinical features and speech recognition performance

Patient	Gender	Age (years)	Device	Cause of deafness	Speech recognition
					NU-6 words	CID sentences
1	F	48	Nucleus CI	Progressive NHL	60%	98%
2	M	52	Nucleus CI	Meniere’s disease	50%	97%
3	M	37	Nucleus CI	Meningitis	20%	83%
4	F	63	Clarion CI	Progressive NHL	68%	85%
5	F	48	Clarion CI	Progressive NHL	42%	72%
6	M	22	Nucleus ABI	Neurofibromatosis II	MTS	CUNY sentences
					79% (stress)	45% visual
					42% (word)	61% auditory + visual