Experimental Brain Research

, Volume 164, Issue 3, pp 357–364

Localization of the human female breast in primary somatosensory cortex

Authors

    • Department of Clinical and Cognitive Neuroscience at the University of HeidelbergCentral Institute of Mental Health, J5
    • University Hospital Charité Psychosomatic Medicine
    • University Hospital Charité Psychosomatic Medicine
  • Michael Schaefer
    • Human Cortical Physiology Section NINDSNational Institutes of Health
  • Sabine M. Grüsser
    • Institute for Medical PsychologyCenter for Humanities and Health Sciences
  • Herta Flor
    • Department of Clinical and Cognitive Neuroscience at the University of HeidelbergCentral Institute of Mental Health, J5
Research Article

DOI: 10.1007/s00221-005-2257-2

Cite this article as:
Rothemund, Y., Schaefer, M., Grüsser, S.M. et al. Exp Brain Res (2005) 164: 357. doi:10.1007/s00221-005-2257-2
  • 109 Views

Abstract

Rationale: Despite an extensive body of research on the topography of the primary somatosensory cortex (S1) little is known about the representation of the trunk. Aim: The aim of this study was to determine the representation of the breast in S1 in human females. Results: The representation of the human breast in primary somatosensory cortex was determined in ten healthy female subjects. Non-painful electrical stimulation of the mammilla (Th4 dermatome), groin (L1 dermatome) and the first digit of both sides of the body activated cutaneous receptors and thus elicited somatosensory evoked potentials. The representation of these body parts in primary somatosensory cortex (S1) was determined using neuroelectric source imaging. Equivalent current dipole localizations were overlaid with individual structural magnetic resonance images to account for individual cortical differences. The breast representation was localized between the representation of the groin and the first digit. In the medial–lateral direction the representation of the breast was approximately 15 mm lateral of the longitudinal fissure in the contralateral hemisphere. Source localizations were stable across subjects. However, one subject showed ipsilateral representation of the breast, which might be related to bilateral receptive fields of the ventral body midline representation. This study confirms the Penfield and Rasmussen (1950) invasive data by use of noninvasive source imaging.

Keywords

EEGTrunk representationElectrical stimulationSomatosensory evoked potential (SEP)Source localization

Introduction

Although many studies have been conducted on the topography of the primary somatosensory cortex in both monkeys and humans (Hari et al. 1993; Itomi et al. 2000; Nelson et al. 1980; Penfield and Boldrey 1937; Penfield and Rasmussen 1950; Woolsey 1964; Woolsey et al. 1979; Yang et al. 1993) little is known about the representation of the trunk. Studies that investigated the organization of the primary somatosensory cortex in animals revealed a small representation of the trunk region lateral to the representation of the thigh and leg (e.g. Felleman et al. 1983; Nelson et al. 1980); the trunk region was rarely investigated in detail, however (Conti et al. 1986; Taoka et al. 1998; Xerri et al. 1994). Conti et al. (1986) and Taoka et al. (1998) examined the trunk region in monkeys in the context of the existence of bilateral neurons. They found that both the ventral and dorsal midline of the trunk are represented bilaterally in addition to predominantly contralateral and ipsilateral representation. In macaques, little cortex seems to be devoted to the representation of the trunk, with large receptive fields on the lower and middle trunk and slightly smaller ones on the upper trunk (Rothemund et al. 2002). Rothemund et al. (2002) found the nipple representation at only two mapping sites in one female macaque. This finding is similar to results reported by Xerri et al. (1994) who studied the ventrum in rats and compared receptive field size in lactating rats with that in non-lactating and virgin control rats. The nipple and areolar skin were found to be weakly represented in both lactating and non-lactating rats.

Only a few human studies on the representation of the trunk in primary somatosensory cortex (S1) are available, probably because of its less obvious functional significance (Itomi et al. 2000; Nakamura et al. 1998). In the magnetoencephalographic (MEG) study of Nakamura et al. (1998) the authors reported no clear response to tactile stimulation of the trunk. Itomi et al. (2000) also used MEG to localize the representation of the trunk in SI after electrical stimulation of the dermatomes Th4–Th12 in 14 healthy subjects. They reported two relatively stable event-related potentials (ERPs) for stimulation of dermatomes Th6–Th10, one at 25 ms after stimulus onset and another at 40 ms after stimulus onset. The representation of dermatomes Th6, Th8, and Th10 followed a medial to lateral sequence with an assumed distance between the receptive fields of each dermatome less than a few millimeters. However, the difference between these values was not significant. Furthermore, consistent ERPs for stimulation of dermatomes Th4 and Th12 were not found. Th4 is the dermatome where the female human breast served by the intercostal nerve is located. Functional magnetic resonance imaging studies of the trunk have so far not been conducted, probably because of the weak representation of the trunk at lower field strengths. The existing studies did not yet reveal the localization of the breast representation in SI of healthy female subjects. Therefore, the goal of this study was to localize the representation of the human female breast in the primary somatosensory cortex. Some of the results have been presented in brief elsewhere (Rothemund et al. 2004).

Methods

Participants

Thirteen healthy female volunteers participated in the study. They had to meet the following inclusion criteria:

  • no history of malign or benign breast disease,

  • no history of severe pain in any body part,

  • no current psychiatric condition,

  • no polyneuropathic disorder such as diabetes, and

  • between 18 and 40 years of age.

Most of the volunteers were recruited through flyers containing information about the study. After participating in the electroencephalographic (EEG) examination and after obtaining individual magnetic resonance images (MRI), participants were reimbursed with €50 for participation in the study. One of the participants had one child. Five women used oral contraceptives. All of the participants had completed at least 12 years of school, most of the participants (n=8) were students.

Three subjects were excluded from further analysis because of poor signal-to-noise ratio of the EEG data. In addition, in one subject the EEG recording was interrupted because of the subject’s indisposition. For this subject no data were obtained for stimulation of the groin. The subjects’ mean age in this study was 25 years (SD=3.3, range=19–29).

Written informed consent was obtained from all participants. The study was approved by the local Human Subjects Committee and adhered to the Human Subjects Guidelines of the Declaration of Helsinki.

EEG assessment

A 61-channel EEG including horizontal and vertical electrooculograms to control ocular artifacts was derived. The Ag/AgCl electrodes were fixed into the electrode cap (Electrocap International) following the 10–20 system. Additional electrodes were placed in five concentric circles with an electrode distance of 1.5 cm. CZ was used as reference electrode and the ground electrode was mounted on the subject’s forehead. Electrode impedance was kept below 5 kΩ .

The positions of all electrodes, and anatomical landmarks such as the nasion and the left and right preauricular points were digitized using an infrared-based camera system (Optotrak, Northern Digital) and used for source localization. The electrical signal from cortical and peripheral channels was amplified using two SynAmps amplifiers (Neuroscan, Neurosoft) and recorded using the Scan (Acquire) software (Neuroscan, Neurosoft). To derive somatosensory evoked potentials (SEPs), non-painful electrical stimulation with 2,000 constant current bipolar square wave stimuli was applied at a randomized stimulation frequency of 3–4 Hz. Stimulus duration was 0.4 ms. The current applied for stimulation was between 1 and 10 mA. Before stimulation, the sensory threshold was determined. The current applied for stimulation was chosen above sensory and below pain threshold. The subject was asked to choose a stimulus intensity that was non-painful but readily distinguishable. The continuous signal was derived at a sampling rate of 1 kHz, and filtered online from DC to 200 Hz. The recordings were made in an electrically shielded room. For localization of the breast the skin area of the nipple (Th4-dermatome, innervated by the intercostal nerve), the skin area of the groin (L1-dermatome, innervated by the iliohypogastric nerve), and the tip of the first digit (D1, innervated by the median nerve) at both sides were stimulated (Fig. 1). The stimulation electrodes were not directly placed on the nerves but on skin areas (receptive fields) innervated by these nerves. Ring electrodes with a diameter of 1.6 cm were used for transcutaneous electrical stimulation.
Fig. 1

The location of the stimulation electrodes on one side of the body is shown. Subjects were stimulated on both sides of the body. Dashed lines indicate dermatomes

The signal was filtered offline from 1 to 60 Hz, segmented from −60 to 100 ms and baseline-corrected from −50 to 0 ms. Ocular artifacts outside −50 and 50 μV were rejected and the data were then averaged. The electrodes were transformed to a common average reference.

Neuroelectric source imaging

Magnetic resonance (MR) images (Siemens Vision MR 1.5T scanner, T1-weighted, TR=22 ms, TE=10 ms, alpha=30°, slice thickness=1 mm) of the subject’s head were obtained to account for individual differences of the anatomical structure of the cortex. The MR images were superimposed on the dipole localizations by using CURRY neuroimaging software (Neuroscan). The volume conductor was a spherical head model consisting of three conductive shells (brain, skull, and scalp) fitted to the three-dimensional (3D) coordinates of the EEG-electrodes. The origin of the 3D head coordinate system was at the left front bottom corner of the 3D magnetic resonance block. The x-axis corresponded to the medial–lateral, the y-axis to the anterior–posterior, and the z-axis to the inferior–superior dimension. For representation of the data (Fig. 2) this coordinate system was later transferred into a head-based coordinate system with its origin at the midpoint between the preauricular points. The multiple signal classification (MUSIC) algorithm (Mosher et al. 1992) was used to localize the current equivalent dipoles. MUSIC belongs to the so-called signal subspace methods. The algorithm scans a single dipole model through a 3D head volume and computes projections on to an estimated signal subspace. A grid is defined on this 3D head volume. For each point of this grid the forward model for a dipole at this location is projected against a signal subspace that has been computed from the EEG data. The source model gives the best projections at locations that correspond to dipole locations (Mosher and Leahy 1998). The process underlying the MUSIC algorithm is a singular value decomposition (SVD), which splits the information contained in the time courses of all channels into fixed, orthogonal spatial patterns of potentials, together with the signal-to-noise ratio (SNR) and time courses (loadings) for each of these patterns. The signal subspace is associated with the leading singular values of the spatiotemporal measured data matrix, the metric peaks at the locations of multiple, non-coherent sources for a given time range. The generator of S1 activity was determined by selecting the peak of maximum probability within the somatosensory cortex.
Fig. 2

Representations of the human female breast, the groin, and the first digit in primary somatosensory cortex. The figure shows the localizations averaged over the ten subjects. The left part of the figure displays the locations in a coronal plane and the right part refers to source localization results in a horizontal plane. Source localization reveals that the representations of the breast (Th4) and the groin (L1) were located close to each other. D1 refers to the first digit

In one subject localization of the breast could not be determined using MUSIC, because of artifacts. In this case a conventional localization procedure—principal-component analysis (PCA) was chosen. PCA is based on an SVD of a matrix containing the measured data. This matrix is decomposed into two orthonormal rows, one representing the temporal properties and a second representing the spatial properties. The diagonal row contains the singular values. Cortical representations of the stimulated locations were assessed by source modeling of the earliest prominent activity peak. The dipoles were fitted in a time window of 0–80 ms after stimulus onset when the primary somatosensory cortex (S1) is active (Forss et al. 1994; Hari et al. 1993). The spatial resolution of EEG is approximately 7–8 mm compared with 2–3 mm in MEG (Itomi et al. 2000; Leahy et al. 1998). However, both EEG and MEG yield comparable data (Leahy et al. 1998).

The hemispheric difference was determined intraindividually (Gallen et al. 1994) using the Euclidean distance (Elbert et al. 1994) measure to determine distances in 3D space, and then averaged over the ten subjects.

The statistical significance of localization differences between hemispheres and of cortical distances between representation of body parts was computed by applying two-tailed t-tests for paired samples. P-values less than 0.05 were considered statistically significant. The software package SPSS was used for all statistical analyses. Because of the indisposition of one subject during the experiment, results of the groin representation were calculated for nine subjects.

Results

Source localization showed that the breast representation (Th4) is localized between the representation of the groin (L1) and the first digit (Figs. 2 and 3). In the medial–lateral direction the breast representation was found approximately 15 mm lateral from the longitudinal cerebral fissure.

Digit 1 representation was always localized in the contralateral hemisphere, that is, the right D1 was represented in the left hemisphere. Breast representation was predominantly localized in the contralateral hemisphere with only one of the ten subjects showing an ipsilateral representation. Furthermore, one of the ten subjects showed an ipsilateral representation of the groin. Figure 4 shows typical SEP patterns for stimulation at Th4, L1 and the first digit.
Fig. 3

Averaged source localizations for the human female breast, the groin, and the first digit in primary somatosensory cortex after electrical stimulation of either side of the body. Displayed are medial–lateral and inferior–superior dimensions of the dipole localizations. The displayed ranges show the standard errors of the mean. The zero-value on the abscissa corresponds to the localization of the longitudinal fissure. The figure shows topographically arranged dipole localizations for the stimulated body parts with the groin representation located most medially, followed by breast and D1 representation

Fig. 4

Representative SEPs from a single subject to stimulation of the mammilla (Th4-dermatome, intercostal nerve), groin (L1-dermatome, iliohypogastric nerve), and first digit (median nerve). Neighboring electrode traces are shown. Arrows indicate the peaks used for source localization

The mean Euclidean distance between the vertex and the right breast was 41.1 mm (SD=7.9, range=26.5–52.3) compared with 44.2 mm for the left breast and the vertex (SD=10.6, range=26.4–61.8). The mean Euclidean distance between the vertex and the right groin was 45.7 mm (SD=9.6, range=31.2–61.8) whereas this difference was 47.8 mm between the left groin and the vertex (SD=13.1, range=26.0–69.3). The mean Euclidean distance between the right thumb (D1) and the vertex was 54.0 mm (SD=10.6, range=41.7–75.8) whereas the mean difference between the left D1 and the vertex was 55.0 mm (SD=12.8, range=28.1–76.4). The distance between the vertex and the localization in the left and right hemisphere was symmetrical with only few millimeters difference between the hemispheres (tbreast(9)=−0.81; P=0.44; tgroin(8)=−0.46; P=0.66; tD1(9)=−0.31; P=0.76).

Within the right and left hemispheres, the Euclidean distance between the localization of the breast and the vertex was not significantly smaller than the distance between the groin and the vertex (tright(8)=−1.37; P=0.21; tleft(8)=−0.93; P=0.38). The breast was located in a more superior position (Fig. 2), although not significantly. The mean Euclidean distance between the localization of the right groin and the right breast was 18.8 mm (SD=10.8, range=8.6–42.7) whereas the mean Euclidean distance of the left groin and left breast localization was 24.9 mm (SD=16.1, range=7.6–55.3; tgroin-breast(8)=1.81; P=0.11). The mean distance between right D1 and right breast was 25.9 mm (SD=12.5, range=6.2–47.0), and the mean difference between the localization of the left D1 and the left breast was 33.2 mm (SD=11.1, range=15.9–53.7; tD1-breast(9)=2.25; P=0.51). The dipole coordinates and the respective goodness of fit (GoF) values, means, and standard deviations are displayed in Tables 1, 2, and 3.
Table 1

Source localization in mm for the right and left breast in each subject, separately for the x-, y-, and z-directions

 

Right breast

Left breast

x

y

z

GoF

x

y

z

GoF

br01k

124

95

204

95.4

161

118

207

92.8

br02k

124

97

186

79.9

163

95

193

77.4

br03k

151

107

201

81.1

115

95

197

81.0

br04k

112

101

198

79.6

167

102

204

79.0

br05k

134

95

206

a

148

95

193

83.0

br06k

129

91

188

82.0

167

85

187

88.0

br07k

134

112

203

72.0

139

96

186

85.0

br08k

136

95

199

85.0

156

92

206

83.0

br09k

138

103

190

93.0

142

88

195

93.0

br10k

138

98

197

74.0

154

109

210

87.0

Mean

132

99.4

197.2

82.4

151.2

97.5

197.8

84.9

SD

10.5

6.4

7.0

7.8

16.1

9.8

8.5

5.3

x=medial–lateral, y=posterior–anterior, z=inferior–superior, GoF=goodness of fit

aSource localization was determined by computing principal component analysis. The GoF of this algorithm cannot be compared with the GoFs of the MUSIC algorithm

Table 2

Source localization in mm for right and left groin in each subject, separately for the x-, y-, and z-directions

 

Right groin

Left groin

x

y

z

GoF

x

y

z

GoF

br01k

116

85

200

91.8

149

111

205

91.1

br02k

127

79

188

76.0

157

85

171

95.3

br03k

124

113

196

86.5

158

93

181

83.9

br04k

152

91

209

63.0

112

101

198

91.0

br05ka

        

br06k

144

88

191

93.0

167

92

190

87.0

br07k

140

99

188

89.0

144

100

198

69.0

br08k

142

88

195

82.0

142

87

185

76.0

br09k

139

106

201

78.0

142

98

201

89.0

br10k

138

103

190

85.0

144

99

189

82.0

Mean

135.8

94.7

195.3

82.7

146.1

96.2

190.9

84.9

SD

11.2

11.1

7.0

9.4

15.6

7.9

10.8

8.3

x=medial–lateral, y=posterior–anterior, z=inferior–superior, GoF=goodness of fit

aDue to the subject’s indisposition during the EEG measurement no localization of the groin was obtained

Table 3

Source localization in mm for right and left first digit in each subject, separately for the x-, y-, and z-directions

 

Right D1

Left D1

x

y

z

GoF

x

y

z

GoF

br01k

106

89

201

93.0

167

88

200

91.5

br02k

96

93

173

93.6

162

83

165

97.4

br03k

111

99

196

83.0

147

123

196

87.0

br04k

92

105

197

92.6

184

100

190

93.2

br05k

105

114

197

60.0

182

115

190

78.0

br06k

114

106

188

87.0

187

106

198

91.0

br07k

91

96

193

90.0

191

107

194

92.0

br08k

115

95

200

89.0

179

92

196

66.0

br09k

137

97

189

73.0

181

98

191

92.0

br10k

124

95

204

80.0

167

102

204

84.0

Mean

109.1

98.9

193.8

84.1

174.7

101.4

192.4

87.2

SD

14.5

7.4

8.9

10.7

13.6

12.1

10.6

9.2

x=medial–lateral, y=posterior–anterior, z=inferior–superior, GoF=goodness of fit

The mean GoF for the localization of the dipoles was above the 80% criterion (Forss et al. 1994) although individual dipole fits did not meet this criterion (14 out of 58 dipole fits). The mean GoF of the localization of the right and left breast was 82.4% (SD=7.8, range=72.0–95.4) and 84.9% (SD=5.3, range=77.4–93.0), 82.7% (SD=9.4, range=63.0–93.0) for the right and 84.9% (SD=8.3, range=69.0–95.3) for the left groin, 84.1% (SD=10.7, range=60.0–93.6) for the right and 87.2% (SD=9.2, range=66.0–97.4) for the left D1.

With respect to the anterior–posterior line, the breast representation was localized within the primary somatosensory cortex and was found just lateral to the representation of the groin. With respect to localization in the inferior–superior dimension, the breast seemed to be located most superior, followed by D1 and the groin.

Discussion and conclusions

There is a lack of studies explicitly exploring the cortical representation of the trunk region. This might be because of its less obvious functional significance compared with other body parts such as the hand, and a greater difficulty in imaging its location because of its small representation. Many investigators who have studied the functional organization of S1 focused on the representation of hand, foot, or face (Baumgartner et al. 1998; Hari et al. 1996; Hashimoto et al. 1992; McCarthy et al. 1993; Sutherling et al. 1988). Although there are some extended reports on the organization of S1 (cf. Hari et al. 1993; Yang et al. 1993), only few human studies investigated the trunk region in detail (Itomi et al. 2000; Nakamura et al. 1998).

In the study reported here, the representation of the female breast at dermatome Th4 in primary somatosensory cortex was determined in ten healthy subjects using neuroelectric source imaging. EEG has been shown to be a low-cost but reliable method (Schaefer et al. 2002) to study the somatotopy of somatosensory cortex, even if sources are expected to be localized close to each other (Baumgartner et al. 1993) and may yield superior activation patterns compared with functional MR at low field strengths when the representation of the body part is small. MEG yields better resolution but was not available in this study.

Dipole source analysis revealed the representation of the breast (Th4) to be localized between the representation of the groin (L1) and the first digit. The representation of groin, breast and D1 showed a medial to lateral sequence, and thus a somatotopic order. In the inferior–superior dimension the breast representation was located most superiorly, followed by representations of D1 and the groin.

In contrast with our results, Itomi et al. (2000) did not find consistent ERPs subsequent to stimulation of Th4 and Th12. The authors explain this result in terms of insufficient stimulation because of problems of fixation of the stimulation electrodes caused by muscles and fat at the respective sites. Electrode fixation problems did not arise in the study we report here. Furthermore, the location of the stimulation electrode in the two studies was different. Itomi et al. (2000) stimulated Th4 around the mid-axillary line of the dermatome, parallel to the costal ribs whereas the female subjects of our study were stimulated directly at the nipple.

In the study presented here, the dipole locations obtained for the groin and the breast were found to be close to each other (Figs. 2 and 3). This result may be explained by a small representation of both the breast and groin in primary somatosensory cortex, because of large receptive fields. Animal studies further support the hypothesis of a small cortical representation of the breast, in particular of the nipple (Rothemund et al. 2002; Xerri et al. 1994). Possible plasticity effects such as increased representation of the nipple-bearing skin as induced by nursing behavior (Rosselet et al. 2004; Xerri et al. 1994) were unlikely to be seen in this human sample because all subjects except one were nullipara. In addition, plasticity effects related to breast feeding are reversed when the nursing behavior is terminated (Rosselet et al. 2004).

The relative position of the trunk localization in the study presented here differed from the results of Itomi et al. (2000). They found Th6–Th10 to be localized clearly distant of the longitudinal cerebral fissure whereas in this study the trunk representation was found to be adjacent to the groin representation and approximately 15 mm lateral of the longitudinal cerebral fissure (Figs. 2 and 3). One explanation might be the different location of the stimulation electrodes in the two studies. Although localization accuracy of EEG is less than that of MEG, it has the advantage that radially oriented dipoles can be detected whereas MEG enables detection of tangentially oriented dipoles only.

Two subjects in this study showed ipsilateral representation of either groin or breast. The phenomenon of ipsilateral dominance has been described for the localization of foot movement in motor cortex (Brunia and Van den Bosch 1984). The authors proposed that ipsilateral activation is caused by a contralateral source near the longitudinal fissure. The dipoles are thought to be directed oblique to the median plane, and thus seem to be stronger at the ipsilateral side. This might explain the ipsilateral localization of the groin about 5 mm from the longitudinal fissure (medial–lateral axis) in this study. The breast representation, however, was found lateral of the representation of the groin approximately 15 mm distant from the longitudinal fissure. Hence, contralateral sources near the longitudinal fissure as described by Brunia and Van den Bosch (1984) are unlikely to account for the ipsilateral localization of the breast. Ipsilateral breast representation may have been caused by neurons which have ipsilateral receptive fields on the body/trunk ventral midline. The phenomenon of ipsilateral and bilateral receptive fields in addition to contralateral receptive fields on the midline trunk has been shown by Conti et al. (1986) in macaque and by Taoka et al. (1998) in awake Japanese monkeys and may be similar in human primates.

In conclusion, after electrical stimulation source localization methods revealed the representation of the female human breast at the Th4 dermatome to be localized between the representation of the first digit (D1) and the groin at the L1 dermatome, relatively close to the latter.

Acknowledgements

This research was supported by a grant from the Deutsche Forschungsgemeinschaft to Herta Flor (FL 156/25) and Boehringer Ingelheim Funds to Yvonne Rothemund.

Copyright information

© Springer-Verlag 2005