INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Dyspnea, an uncomfortable awareness of difficult breathing or the need to breathe, is a common, often distressing clinical symptom in a large number of pathologic conditions. Furthermore, like pain, dyspnea is a multidimensional sensory experience encompassing an important affective component and thus likely to involve complex central processing. Previous studies of putative underlying mechanisms of dyspnea (reviewed in 1- 3), mostly based on acutely induced unpleasant respiratory sensations in healthy volunteers, have investigated peripheral rather than central aspects of dyspnea. Overwhelming evidence now exists that the numerous putative receptors and neural afferent pathways all contribute to dyspnea, but that none are exclusive for this sensation (2). These findings suggest that dyspnea results from multiple and simultaneous sensory inputs, thus highlighting again the crucial importance of the central nervous system in the final integration of this complex sensation. Despite recognition of this issue, surprisingly little is currently known about the brain areas that are specifically involved in this processing. Previous studies using the evoked-potential technique identified cortical projections of basic unpleasant respiratory sensations induced by external loads (4) or out-of-phase chest-wall vibration (5). However, unless evoked-potential studies are combined with dipole tracing and brain imaging (5), the identification of subcortical structures involved in sensory processing is not possible. In the present study, we attempted to identify the neural substrates of dyspnea using functional brain imaging with positron emission tomography (PET), which allows identification of activated brain regions throughout the whole brain volume. We assessed brain activation specifically associated with respiratory discomfort during loaded breathing, which is a fundamental component of dyspnea in a large variety of respiratory diseases. Furthermore, we performed a separate assessment of brain activation associated with modulation of perceived intensity of respiratory discomfort that is not related to its main physical stimulus, the respiratory motor response. We speculated that the integration of factors underlying this perceptual modulation, which are likely to include affective/cognitive functions, may involve a specific brain region or regions.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

Eight healthy male volunteers (seven right-handed and one ambidextrous) with a mean age of 33 (± 3) yr participated in the study after informed consent. None of the subjects had any history of neurologic or respiratory disease. The study was approved by the national and institutional ethics committee.

Induction and Assessment of Respiratory Discomfort

Induction. We assessed respiratory discomfort related to loaded breathing, an important component of dyspnea, because it can be induced in healthy subjects by external loads, thereby allowing the induction of large and repeated changes in sensation intensity required by PET activation studies. Respiratory discomfort was induced by adding near linear resistive loads to an external breathing circuit during both inspiration and expiration. Loads consisted of hollow cylinders with different internal diameters but identical external diameters so that they could be adapted to the same tube of the respiratory circuit. The magnitude of the resistive loads was determined by the size of their internal diameter. The two loads used for the PET study were selected from a set of loads of different magnitudes during a preliminary study outlined subsequently. We used the amplitude of mouth pressure swings (Pm) as an index of the increase in command and execution of the respiratory motor response (referred to as the "main generating mechanism of respiratory discomfort" in our study), which is assumed to most closely reflect the stimulus of dyspnea during resistive loaded breathing (6, 7). Because during resistive loaded breathing there is a concomitant increase in respiratory discomfort and its main generating mechanism, we induced additional intrasubject variation in perceived intensity of respiratory discomfort for each given level of Pm in order to be able to assess specifically brain activity associated with sensory perception. We therefore repeated all experimental conditions with concomitant inhalation of menthol because this agent has been previously shown to be able to change the intensity of perceived respiratory discomfort during high loaded breathing without any significant change in mean inspiratory flow, another index of respiratory motor response (8). Menthol was delivered by a filter that was saturated 10 h before the PET study with a freshly prepared alcoholic solution of l-menthol and inserted in the external breathing circuit. The subjects were kept blind to the fact that some experimental conditions included menthol inhalation because we found in an earlier study (unpublished results) that the modulating effect on sensation intensity decreased in some subjects when menthol inhalations were repeated.

Assessment. The subjects were asked to rate perceived intensity of respiratory discomfort by selecting the appropriate word descriptor on the modified Borg scale, a validated sensation scale (9). Respiratory discomfort was defined in a standardized way as "the unpleasant, uncomfortable sensation of difficulties in breathing" by stressing the fact that its assessment should include the affective component of the sensory experience.

Apparatus and Recording

External breathing circuit. The same external breathing circuit was used for all experimental conditions. The device consisted of a mouthpiece connected in series with a first short rigid tube, the site of recording of mouth pressure and end-tidal CO₂ pressure (PET_CO2), followed by a pneumotachograph and a second short tube designed for the easy introduction and removal of the resistive loads. Throughout all experimental conditions mouth pressure, flow, volume, and PET_CO2 were continuously monitored and displayed on a paper chart by a multichannel recorder (Gould ES 2000). Mouth pressure was recorded using a pressure transducer (P ± 50 cm H₂O; Validyne Corp., Northridge, CA). Instant flow was assessed with a pneumotachograph (Fleisch No. 2; Lausanne, Switzerland) connected to a differential pressure transducer (P ± 2 cm H₂O; Validyne Corp.) and integrated to volume (Thomson Instruments, Montreal, PQ, Canada). PET_CO2 was analyzed with an infrared CO₂ analyzer (Beckman LB-2; Beckman Instruments, Fullerton, CA). For the experimental conditions with concomitant menthol inhalation, the filter with menthol was inserted between the first rigid tube and the pneumotachograph.

Functional brain imaging. We performed functional imaging in a three-dimensional mode with an ECAT-HR+ PET camera (Siemens, AG, Erlangen, Germany). Relative changes in regional cerebral blood flow (rCBF) were taken as an index of local neuronal activity. rCBF changes were determined from the distribution of cerebral radioactivity after a single bolus intravenous injection of 9 mCi of radiolabeled water (H₂¹⁵O) (10). Each rCBF measurement lasted 80 s and was separated from the subsequent one by a 10-min rest period to allow decay of radioactivity to a near background level (< 3% of the recorded peak). Each subject received eight injections of H₂¹⁵O (corresponding to the eight rCBF measurements performed in our study).

Experimental Protocols

Pilot study. Just before the PET study, we performed a session of repeated load presentations with concomitant recordings of mouth pressure in order to select for each subject the two loads used for the PET study. These two loads were chosen to induce a sensation that was rated as "slight" or "moderate" for the first one, and as "severe" in more than 50% of the runs for the second one, corresponding to a Borg score of 2 or 3 and  5, respectively. The aim was to obtain two different sets of Pm values and intensity ratings with sufficiently high intrasubject and between-subject variation within each set to assess correlations between rCBF and each of the two aforementioned variables. In addition, we verified for each subject that the unloaded condition, i.e., when breathing through the external breathing device without added load, did not induce any respiratory discomfort. A further objective of the pilot study was to familiarize the subjects with the experimental device and to explain how to use the Borg scale. The subjects were told that the loads presented during the PET study would include some of those presented during the pilot study but they were kept blind to the selected load magnitude as well as to the presentation order.

PET study. The subjects lied supine in the scanner with the head immobilized in a position so that the axial brain slices were parallel to the reference line, i.e., the line connecting the anterior and posterior commissure (AP-PC line). After a transmission scan, we performed the eight rCBF measurements corresponding to the different experimental conditions. Thirty seconds before the beginning of the rCBF measurement, the subject was connected to the breathing circuit, the noseclip applied, and the recording of respiratory parameters was started. For the loaded conditions the resistive load was added 10 s before data acquisition. Immediately after each rCBF measurement, the respiratory device and the nose-clip were removed and the subject asked to rate the intensity of the induced respiratory discomfort. We assessed three different experimental conditions, the unloaded control condition and two loaded conditions previously determined during the pilot study, referred to as "moderate" and "high load" conditions, respectively. The "unloaded" and "moderate load" conditions were each applied once, whereas the "high load" condition was applied twice because the effect of menthol on respiratory sensation has been previously shown to be mainly effective for respiratory discomfort related to high loaded breathing (8, personal results). Each condition was applied with and without concomitant inhalation of menthol. Thus, we performed eight rCBF measurements consisting of two sets of four consecutive measurements separated by 10-min intervals, one set with and one set without concomitant menthol inhalation.

We grouped all conditions with menthol together within the same set and allowed a time interval of 20 min between the two sets in order to avoid any carry-over effect of menthol when it was inhaled during the first set. The 20-min interval was chosen because we found in an earlier study (unpublished results) that 10 min after a loaded condition with menthol inhalation, i.e., the time between two subsequent rCBF measurements, a small taste of menthol remained in a few subjects, but that this effect disappeared 15 to 20 min after inhalation. Within the first set, the order of different experimental conditions was randomized for each subject and this order was reversed for the second set. The order of presentation of the sets, with or without concomitant inhalation of menthol, was also randomized. In addition, a T1-weighted magnetic resonance image (MRI) was acquired for each subject.

Data Analysis

Respiratory and sensory responses. Perceived intensity of respiratory discomfort was assessed in terms of Borg scores, i.e., the numbers corresponding to the descriptors chosen by the subjects on the Borg scale. For each subject, peak inspiratory and expiratory mouth pressure, tidal volume (VT) and respiratory cycle duration (Ttot) were obtained from the breath-by-breath paper record. The values of these respiratory parameters were used for the calculation of Pm (mean amplitude of mouth pressure swings, the index of the main generating mechanism of respiratory discomfort), respiratory frequency (f), and minute ventilation (E). For each subject, Pm, VT, f, E, and PET_CO2 were averaged for each experimental condition. The resulting mean values of these respiratory parameters as well as the Borg scores were then averaged for the whole subject group. We also determined the relative contribution of these parameters to intensity of respiratory discomfort for the whole subject group by multiple stepwise regression analysis. A p value of < 0.05 was considered as statistically significant.

Functional brain imaging. Attenuation-corrected data were reconstructed into 63 2.25 mm thick axial slices with a resulting resolution of 4.5 mm full-width at half maximum after reconstruction. PET data were analyzed with the SPM96 software package (Wellcome Department of Cognitive Neurology; http//http://www.fil.ion.ucl.ac.uk/spm). Statistical maps of significant relative rCBF (11) were created after the subsequent steps, including image realignment to the first scan for the correction of head movement, smoothing, and registration into the standardized coordinate system of Talairach and Tournoux (12). Given the small number of scans for each experimental condition, all imaging results were averaged for the whole study group. State-dependent differences in global flow were covaried out using proportional scaling. Change in rCBF was determined by a categorical design based on a voxel-by-voxel comparison for the pooled data of the eight subjects between the different conditions, i.e., high load versus unloaded control condition and moderate load versus the unloaded control condition. We also performed a parametric design to assess the correlation between brain activation, evaluated as increases in rCBF, and individual values of each of the two explanatory variables, intensity ratings of respiratory discomfort and Pm, by taking into account the correlation between these two variables (13). In addition, this design further allowed a separate assessment of brain regions where rCBF was positively and specifically correlated with the part of variance of perceived intensity of respiratory discomfort that was not related to Pm, and similarly, the part of variance of Pm that was not related to intensity of respiratory discomfort. Significance of the changes of rCBF were determined by t statistic further transformed into a normally distributed Z statistic. rCBF changes were considered as significant (threshold for display) for a Z value  3.09 (which corresponds to p < 0.001 uncorrected for multiple comparisons) with special regard for rCBF changes reaching a Z value  4.41 (which corresponds to p  0.05 when corrected for multiple comparisons).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Respiratory and Sensory Responses

Individual results concerning respiratory and sensory responses to the loads and to menthol inhalation are shown in Figure 1 and group results in Table 1. As compared with the unloaded control condition, there was an expected increase in Pm, the index of the main generating mechanism of dyspnea (Table 1, Figure 1). There was also a slight increase in PET_CO2 and VT, as well as a decrease in f and E (Table 1). Global comparison of all experimental conditions with and without concomitant menthol inhalations showed that menthol induced no significant change in the two indices of respiratory motor response, Pm, and E (Table 1). However, menthol inhalation induced a decrease in VT and an increase in f, resulting in a very small but significant increase in PET_CO2 (Table 1). Our results showed that in individual subjects (Figure 1), as well as in the whole group, intensity ratings of respiratory discomfort were positively related to Pm. Stepwise regression for the whole study group showed that only Pm made a major contribution (65%) to the variance of intensity ratings of respiratory discomfort, whereas f made only a weak contribution (2%), and volume, E, and PET_CO2, no significant contribution. Menthol inhalation had no significant effect on self-rated sensation intensity related to the unloaded and moderate load condition (Table 1) and a highly variable effect from one subject to another and even within subjects for the high load condition (Figure 1). Indeed for the latter condition, menthol induced a decrease in intensity ratings for similar levels of achieved mouth pressure in some subjects as previously described (8), but in others it failed to induce any change and even, in a few cases, induced an unexpected increase (Figure 1). Nevertheless, menthol contributed to the overall subject variability of perceived intensity for each given level of Pm (Figure 1).

View larger version (24K): [in this window] [in a new window]   Figure 1.   Relationship between individual ratings of perceived intensity of respiratory discomfort (Borg scores) and Pm, an index of respiratory motor response, the main generating mechanism of respiratory discomfort. Results are presented as individual plots for each of the eight subjects (S1-S8) with and without concomitant menthol inhalation (closed and open symbols, respectively) for all experimental conditions, i.e., the unloaded control condition (triangles), the moderate (boxes), and the high load (circles) condition.

Functional Brain Imaging

PET brain imaging results are group means and are presented as brain activation areas, i.e., areas with significant rCBF increase (clusters). These clusters are characterized by the anatomic localization (coordinates in millimeters of the projection in three dimensions of the standard anatomic space of Talairach) and the statistical significance (in terms of Z scores, significant if Z > 3.09, corresponding to p < 0.001 uncorrected for multiple comparisons) of their maximal peak of rCBF increase and their volume (cluster size) in terms of number of voxels (brain volume of 2 mm³).

Brain activation related to menthol. The comparison of all conditions with menthol versus all conditions without menthol revealed two areas of rCBF increase in the cerebellum, namely in the posterior vermis (Talairach coordinates: 4, 56, 22; Z = 4.98; cluster size: 1,678 voxels) and in the right posterior hemisphere (Talairach coordinates: 36, 78, 36; Z = 4.44; cluster size: 217 voxels). Areas of weaker activation were detected in the right prefrontal lobe (Brodmann area 11), corresponding probably to the olfactory bulb (Talairach coordinates: 10, 56, 24; Z = 3.89; cluster size = 102 voxels) as well as in the left cerebellar hemisphere (Talairach coordinates: 26, 64, 26; Z = 3.58, cluster size: 111 voxels). Because we did not detect any specific brain activation that was the result of an interaction between menthol and loaded conditions, we pooled the data with and without menthol of each experimental condition for further analysis.

Brain activation related to perception of respiratory discomfort and to its main generating mechanism. The comparison of the high load versus unloaded conditions (details in Table 2 and Figure 2) showed three main sites of significantly increased rCBF (p < 0.05 corrected for multiple comparisons), the anterior part of the right claustrum/insula extending into the right frontal operculum and the putamen (Z = 4.75), the posterior cerebellar vermis extending into the right cerebellar hemisphere (Z = 4.44), and the left medial pons (Z = 4.41) (Table 2, Figure 2). Furthermore, areas of less significant rCBF increase were seen in the right secondary somatosensory cortex (Brodmann area 43), and in the right hemisphere of the cerebellum, the left precentral gyrus, and the medial supplemental motor area (Brodmann area 6) (Table 2, Figure 2). Although there was some increase in rCBF in similar brain regions during the moderate load condition, this failed to reach statistical significance. Parametric analysis showed that the brain regions in which rCBF was positively correlated with individual ratings of perceived intensity of respiratory discomfort were very similar to those detected with the comparison of the high load versus unloaded conditions (Table 2; Figure 3, upper panel ). Indeed, main activation areas were again detected in the right insula, the cerebellar vermis, and medial pons. A similar analysis for individual values of Pm revealed activated areas that were almost identical to those where rCBF was positively correlated with intensity ratings (Table 2; Figure 3, lower panel ). Specific assessment of increase in rCBF associated with perceived intensity of respiratory discomfort that is not related to Pm revealed significant activation which centered in the right cingulate sulcus with a caudal extension into the posterior cingulate gyrus (Brodmann area 31) (Talairach coordinates: 8, 26, 46; Z = 4.19; cluster size: 213 voxels) (Figure 4). By contrast, no significant activation was related to the variance of Pm that was unrelated to sensation intensity ratings.

View larger version (22K): [in this window] [in a new window]   Figure 2.   Group results of brain activation associated with high loaded as compared with unloaded breathing. Results are represented as "glass brain views" consisting of the projection of statistical parametric maps (SPM), i.e., areas with a significant increase in the group average rCBF, in the three dimensions of the standard stereotactic space defined by the atlas of Talairach and Tournoux. The brain is viewed from the side (sagittal), from the back (coronal), and from the top (transverse), thus providing a three-dimensional view of the activation areas. The right hemisphere is on the right side of the coronal view and the lower part of the transverse view. Display for all maps is thresholded at Z = 3.09, p < 0.001 (uncorrected). Z values are coded by the color scale shown. Main activation areas (Z > 4.40, corrected p < 0.05) are seen in three distinct brain areas, the anterior part of the right claustrum/insula extending into the frontal operculum and the putamen, the cerebellar vermis extending into the right cerebellar hemisphere and the left medial pons. Minor activation areas (3.10 < Z < 4.41) are present in the right secondary somatosensory cortex (Brodmann area 43), and to a far lesser degree of significance, in the right hemisphere of the cerebellum, the left precentral gyrus, and the medial supplemental motor area (Brodmann area 6).

View larger version (39K): [in this window] [in a new window]   Figure 3.   Group results showing brain areas where neural activation was positively correlated with individual ratings of perceived intensity of respiratory discomfort (upper panel ) and with Pm (lower panel ). The presentation is the same as in Figure 2 with a three-dimensional view of areas of significant increase in the group average rCBF. Activation patterns associated with perceived intensity of respiratory discomfort are very similar to those associated with Pm, involving mainly the anterior part of the right claustrum/insula, the cerebellar vermis, and the left medial pons. These activation areas are also almost similar to those related to high loaded breathing.

View larger version (67K): [in this window] [in a new window]   Figure 4.   Group result of the brain region where neural activation was positively correlated with the part of variance of perceived intensity of respiratory discomfort that is not related to Pm, the index of its main generating mechanism. The result is presented as the statistical parametric map, i.e., the area with a significant increase in the group average rCBF superimposed onto a sagittal section (x = 8) (left panel ) and a coronal section (y = 26) (right panel ) of the group T1-weighted MRI image. Brain activation centered in the right cingulate sulcus (Z = 4.19) extending mainly into the posterior cingulate gyrus (Brodmann area 31). Same presentation as in Figures 2 and 3 concerning the color code for Z values and the threshold for display (Z = 3.09, p < 0.001 uncorrected). The right hemisphere of the brain is on the right side of the coronal view.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

By detecting specific brain activation patterns associated with acutely load-induced respiratory discomfort, the present study is one of the first to identify neural substrates for the conscious perception of dyspnea.

Methodological Considerations

Functional imaging with PET, based on the detection of changes in rCBF, is a powerful tool for the identification of neural activation throughout the whole brain volume. In addition, the scanner used (ECAT EXACT HR+) allows data acquisition in a three-dimensional mode, thereby increasing the sensitivity and decreasing the dose of radioactive tracer, and provides high spatial resolution (approximately 5 mm). However, the specific limitations of PET should be kept in mind. First, this technique is based on the detection of changes in rCBF by the distribution of cerebral radioactivity, which is a validated (10) but nevertheless indirect measurement of neural activation. Second, the low temporal resolution (approximately 1 min) of PET, as well as the relatively long time period between successive scans required for the decay of radioactivity, limit the number of scans that can be performed in one experimental session and thereby the possibility to analyze brain activation in single subjects.

Moreover, activation studies with PET impose specific constraints on the study design; in our case, requiring large and repeated changes in the intensity of dyspnea at precise time points, conditions which are of course difficult to obtain with disease-related dyspnea. This obliged us to assess acutely induced respiratory discomfort, i.e., an experimental model of dyspnea in healthy subjects, rather than disease-related dyspnea itself. However, the experimental protocol used, i.e., loaded breathing that induces an increase in respiratory effort related to that of motor command and execution, is a partial but nevertheless important component in a great variety of circumstances of disease-related dyspnea. The study of acute dyspnea is of clinical relevance because, even during chronic disease, perception of acute onset or rapid changes in the intensity of dyspnea are critical for the detection of changes in the underlying clinical condition.

A further specific difficulty of brain imaging studies of sensory perception is to differentiate neural activation of the latter from that related to its main underlying mechanical changes. We therefore used concomitant menthol inhalation, which has been shown to have a specific modulating effect on perceived intensity of load-induced respiratory discomfort presumably through stimulation of cold receptors (8), to induce additional uncoupling of these two highly related parameters. In our study, the effect of menthol was less important than that obtained in the study of Nishino and coworkers (8). This is probably due to the fact that, given the randomized order of experimental conditions which is mandatory in brain imaging studies, the first inhalation of menthol did not always coincide with the high load condition. Indeed, it has been previously shown (8, personal results) that the modulating effect of menthol on the intensity of perceived sensation is most important in the presence of high loads and during the first minutes of its presence. Nevertheless, menthol inhalation contributed to the overall variation in intensity ratings at each level of its main generating mechanism, achieved Pm, without any significant change in the latter and thus allowed us to perform a separate assessment of brain activation specifically related to perceived intensity of respiratory discomfort.

Results of Functional Brain Imaging

Main activation areas. We identified several highly activated brain areas during the high loaded condition with the greatest increase in rCBF occurring in three distinct brain regions, the right anterior insula, the cerebellar vermis, and the medial pons (Figure 2, Table 2). Our results also showed that brain activation correlated with ratings of perceived intensity of respiratory discomfort and with Pm in very similar brain areas, i.e., mainly the right anterior insula, the cerebellar vermis, and the medial pons (Figure 3, Table 2), which are almost the same regions as those detected for the high load condition. Given the strong interrelationship between respiratory discomfort and Pm, their relative contribution to brain activation cannot be precisely determined. There are, however, several arguments in favor of the hypothesis that the aforementioned brain areas may be concomitantly activated by both parameters and thus may be involved in sensorimotor integration of loaded breathing. Indeed, these brain regions have all been shown to be involved to various degrees in both perception of different unpleasant sensations and motor command and execution. Thus, activation in the anterior insula, a putative sensorimotor integration area of the limbic cortex, has been detected in virtually all imaging studies of acutely induced pain (14) and one study of chronic clinical pain (17). This result suggests that the similarities between dyspnea and pain also apply to brain structures involved in their central processing. Activation in the anterior part of the insula has also been shown to be associated with the perception of various other predominantly unpleasant centrally processed sensations, e.g., aversive emotions (18), panic attacks (19) as well as hunger (20) and thirst (21). Most importantly, in a parallel study to ours (22), despite a clearly different means of generating respiratory discomfort, namely by decreasing the VT at a constant high level of CO₂ in mechanically ventilated subjects, Banzett and coworkers also detected a bilateral but predominantly right-sided activation in the anterior/mid insulae. Furthermore, the insula has been shown to be involved in motor command (23) and, more specifically, in circumstances including respiration, where increased effort production is accompanied by unpleasant sensations. Indeed, activation in the insula has been shown in sensorimotor integration of exercise (24) and of CO₂-induced hyperventilation where discomfort has been mentioned without having been specifically assessed (25). Activation of the cerebellum has also been shown to be associated with increased respiratory motor activity during CO₂- induced and exercise-induced hyperventilation (25) and loaded breathing (28) as well as in several studies of unpleasant sensations, especially pain (14, 15, 17, 18). The peak of our activation area in the pons is located more centrally than the areas found in a previous functional MRI study of loaded breathing (28) and may correspond to medial reticular formation, which has been previously shown to have connections with respiratory-related neurons (29). Furthermore, as suggested by the large intersubject differences of achieved Pm for any given external load (Figure 1, Table 1), motor response to the loads may actually result from the prior integration of various load-related sensory inputs. It is therefore tempting to assume that our main activation areas, i.e., the right anterior insula, the cerebellar vermis, and the left medial pons are parts of a neural network involved in the concomitant processing of both the respiratory motor response to the load and the genesis of the respiratory discomfort and thus determine the main proportion of its perceived intensity.

In addition, concerning the insula, it is noteworthy that brain activation associated with respiratory discomfort induced by two clearly different mechanisms, i.e., resistive loading in our study, and restraining ventilation below spontaneous levels in the study of Banzett and coworkers (22), was located in almost the same area of the anterior right insula. This important finding strongly suggests that this brain structure may be of crucial importance for the quantitative perception of dyspnea irrespective of the underlying generating mechanisms.

Brain activation associated with perceived intensity of respiratory discomfort unrelated to Pm. We detected one area of brain activation that was specifically associated with the variance of perceived intensity of respiratory discomfort that was unrelated to the magnitude of Pm, the index of its main generating mechanism (Figure 4). The activation site was located in a multifunctional limbic integration area, the posterior cingulate gyrus. Thus, this brain area may be involved in the integration of the effect on perception of the various factors such as menthol (8) and chest wall vibration (30) that can change the perception of respiratory discomfort with little or no change in achieved mouth pressure or other indices of respiratory motor response. It should however be noted that in our study, as well as in the others mentioned previously (8, 30), an index, in our case Pm, rather than the actual physical generating mechanism has been measured. However, it seems unlikely that potential inaccuracy of this estimation accounts for the totality of the observed intrasubject variability of perceived intensity of respiratory discomfort for each given level of Pm, given the fact that this variability was very high in some subjects and also differed between subjects (Figure 1). Similarly, because respiratory discomfort was virtually unrelated to the other ventilatory parameters, it is unlikely that the latter play a prominent contributive role in the observed perceptual variability. We hypothesize that this variability may be due, at least in part, to affective/cognitive factors. Indeed, several regions in the posterior cingulate gyrus have previously been shown to be activated by various predominantly unpleasant sensory-affective functions such as acute and chronic pain (17, 31), and, most interestingly, emotion and memory (32) as well as anxiety (33) and its related symptoms (34) which are possible modulators of perceived intensity of respiratory discomfort. Moreover, as for the anterior insula, the activation area in the posterior cingulate gyrus was only present in the right hemisphere. A similar lateralization has been previously demonstrated for pain (16) and has been attributed to a preferential involvement of the right hemisphere in emotional response to aversive stimuli (16). Because affective/cognitive factors play an important contributive role in perceived intensity of respiratory discomfort in health (35, 36), and probably even more in disease-related dyspnea, the identification of a putative neural substrate of affective/cognitive processing of dyspnea is of potential clinical interest.

In summary, the results of the present functional brain imaging study suggest that conscious perception of acutely induced respiratory discomfort during loaded breathing results from two distinct, parallel integration processes. The first may be involved in sensorimotor integration of loaded breathing and thus in the perception of the predominant part of perceived intensity of respiratory discomfort, whereas the second may be involved in the modulation of perceived intensity which presumably includes emotional processing of this sensory experience. These results may contribute to a better understanding of dyspnea and thereby help to improve its clinical management.