Paradoxical heat (PH), the illusion of skin heat, accompanies many neurological disorders. Using the technique of percept-related functional MRI, we found a region of the right insular cortex specifically activated when subjects perceive a heat sensation in their right hand even though their skin temperature is cool or at neutral. This region was suppressed during mild skin cooling. We propose that this differential response is a manifestation of the role of the insula in signaling temperature perceptions regardless of the actual temperature of the skin. These findings suggest that a region within the insula has a complex role in heat perception, perhaps contributing to a specific, rather than general, thermosensory perception. These data provide insight to our basic understanding of normal and pathological thermosensory perceptions.
The phenomenon of “paradoxical heat” (PH) refers to a perception of heat when the actual skin temperature is at or below neutral. PH sensations can occur during prolonged intense skin cooling and when skin temperature returns to a neutral level after intense cooling (e.g., on moving indoors after being outside on a very cold day). PH can also accompany frostbite, multiple sclerosis (Hansen et al. 1996), fibromyalgia (Berglund et al. 2002), and neuropathic pain (Ochoa and Yarnitsky 1994; Yosipovitch et al. 1995). The central mechanisms underlying PH are unknown.
Classically, the spinothalamic tract has been associated with pain and temperature sensation. Although specific heat- and cold-sensitive receptors have been identified (Craig 2002; McKemy et al. 2002; Reid and Flonta 2001), the cortical mechanisms underlying thermal sensation remain illusive. Previous imaging studies of thermal sensations have reported several cortical areas of activation, including the insula, evoked by innocuous and noxious thermal stimuli (Craig 2002; Craig et al. 2000; Iannetti et al. 2003; Peyron et al. 2000). However, the perceptual response to a thermal stimulus can vary across subjects. For instance, an intense cold stimulus can evoke many sensations including cold, pain, prickle, and paradoxically heat, each with a characteristic temporal signature (Davis and Pope 2002). Therefore a powerful approach to imaging brain activity related to sensations requires on-line information about evoked perceptions during the imaging experiment. This approach has been used previously to study emotions (Breiter et al. 1997). We adopted this approach to isolate central correlates of cold-evoked sensations with unique temporal signatures (Davis et al. 2002). In our previous study, percept-related functional MRI (fMRI) detected cortical responses associated with the perception of prickle. In the present study, we used percept-related fMRI to identify cortical responses tightly linked to the experience of PH. We now report that activity within a region of the insular cortex is tightly linked to the experience of PH.
All subjects gave informed written consent to experimental procedures approved by the University Health Network research ethics board. Because cold-evoked PH sensation varies across individuals, a group of 28 normal healthy potential subjects (15 male, 13 female; 21–61 yr old) were prescreened for consistently experienced cold-evoked heat sensations. From this group, 10 right handed normal healthy subjects (5 male, 5 female; 21–42 yr) were selected. Four of these subjects participated in a previous study of cold-evoked prickle (Davis et al. 2002).
Thermal stimulation and ratings
A contact thermal stimulator (20 × 25 cm probe, TSA 2001, Medoc) was used to deliver cold stimuli to the thenar eminence of the right hand. The temperature of the probe was held at 32°C for 40 s between each of five cold stimuli. During each period of stimulation, the probe temperature was cooled to 3°C at 0.5°C/s, held at 3°C for 10 s, and then returned to 32°C at 10°C/s. Subjects used a trackball to rate heat sensations on a visually presented 100-point scale. Only very small (<10 mm) movements of the trackball with the left index finger were necessary to register the ratings.
Imaging and analysis
Whole brain images were obtained on a 1.5-T Signa Echospeed MRI (GE Medical Systems, Milwaukee, WI) system. Anatomical images were acquired for each subject using a T1-weighted gradient echo high-resolution three-dimensional anatomic scan (TE = 5 ms, TR = 25 ms, flip = 45°, 124 1.5-mm-thick slices, 256 × 256 matrix, 24 × 24 cm field of view). Functional images were acquired using a T2*-weighted scan (25 contiguous 4-mm slices, single-shot spiral trajectory through k-space, flip angle = 85°, TR = 2,000 ms, TE = 40 ms, 64 × 64 matrix, 20 × 20 cm field of view) (Glover and Lai 1998). Each functional run consisted of 333 volumes, the first 3 of which were discarded for signal equilibration.
Data were analyzed with BrainVoyager v4.94 (Brain Innovation, Maastricht, Netherlands) as previously described (Davis et al. 2002). Briefly, preprocessing included trilinear interpolation to 1 × 1 × 1-mm resolution, slice acquisition time correction, motion correction, and also linear trend removal, high-pass filtering to attenuate frequencies <0.003 Hz (periods >330 s), spatial smoothing with a 6-mm-diam full-width half-maximum Gaussian kernel, and adjustment for temporal autocorrelation. Functional and anatomical images were co-registered and transformed into Talairach space (Talairach and Tournoux 1988).
The maximum value of the heat rating was taken within each 2-s fMRI data frame. The resulting heat rating curves were convolved with a standard Brain Voyager hemodynamic response function and entered into a general linear model (GLM). To identify activations that were common to all subjects, a conjunction analysis was performed. The conjunction analysis makes a single subject GLM for each subject then takes the minimum t value at each voxel. We calculated a statistical threshold using the following procedure. Alphasim (B. Douglas Ward, Biophysics Research Institute, Medical College of Wisconsin) was used to create Monte-Carlo simulations of our particular scanning and processing protocol. These simulations verified that for the final map, an image-wide (across voxels), conjoint (across subjects) P value of <0.05 could be achieved by using a voxel-wise threshold P value of 0.0001 and cluster size threshold of 140 mm3. Therefore the resulting map was thresholded at these parameters (i.e., voxel-wise P value threshold of 0.0001 and cluster size threshold of 140 mm3). We have successfully used this approach, combining threshold and cluster size previously (Davis et al. 2002; Downar et al. 2002, 2003).
Maps of motor-related activations due to trackball movement were constructed as described previously (Davis et al. 2002). In general, we used the first derivative of the heat ratings curve in each subject as a quantitative measure of trackball movement since this represents the velocity of the trackball. Because similar brain regions are likely involved in generating both upward and downward finger movements, we took the absolute value of this velocity curve. The maximum value of the velocity curve was taken within each 2-s fMRI data frame. The velocity curves were then convolved, and entered into a GLM as individual predictor curves. Autocorrelation was measured and removed, and the GLM was submitted to the same conjunction analysis (and the same thresholds) used for the original heat ratings data.
We evoked PH sensations with alternating periods of cold and neutral temperature stimuli applied to the hand (see methods). In each subject, we tracked the occurrence of PH sensations on-line during fMRI. Subjects typically reported feeling heat sensations when the thermode temperature was returning to a neutral level or held at this level between cooling pulses (Fig. 1). There was both within- and between-subject variability in the precise timing of the PH response with respect to the skin temperature. This variability was essential to isolate the heat percepts from other percepts (such as pain and cold) whose temporal signatures are tightly linked with skin temperature and thus occur at regular intervals (see Davis and Pope 2002; Davis et al. 2002; Harrison and Davis 1999).
Isolated perceptual ratings of heat were used to extract heat percept-related activations during fMRI. In each subject, we used the time signature of their perceptual ratings as a search tool to locate brain regions whose behavior tightly correlated to the percepts. We next performed a conjunction analysis across subjects to find only those brain responses that were activated in each and every subject. Using this strict criterion, we found a single cortical region whose activity tightly correlated to the PH reports in all subjects. This region was primarily located in the right anterior-mid insular cortex ipsilateral to the thermal stimulation site (Fig. 2) with some extension into the adjacent frontal operculum (BA 6/44). To ensure that the insular activation was related to the percept itself and not the motor act of rating, we performed an additional conjunction analysis of the responses associated with the finger movements required to register the ratings. For this analysis, we used the first derivative of the heat ratings curves as the search tool because this provides the temporal signature of the active finger movements (see methods). This analysis revealed activations in the sensorimotor cortex and supplemental motor area, but not in the insula; and is consistent with our previous report of rating-related activations being confined to motor regions (see supplemental material in Davis et al. 2002).
A region-of-interest analysis (ROI) (see methods) revealed a sharp increase in activity time-locked to the occurrence of the reported PH percepts (Fig. 3). This ROI was further interrogated for activity time-locked to the thermode temperature. This analysis revealed a period of reduced activity coincident with the initial period of skin cooling during which time the temperature fell from 32 to ∼12°C, followed by a clear period of activation at and just after the thermode temperature was returned to neutral (Fig. 3).
The insula has historically been associated with a variety of homeostatic functions, cool perception, interoception, and pain (Augustine 1996; Craig 2002), and the present data now show that it is also a key area for the perception of heat.
Our findings are consistent with the recent report of an evoked warm sensation during electrical stimulation of this region in an epilepsy patient surgery (Ostrowsky et al. 2002). Laser heat-evoked potentials have also been localized close to this insular region during excitation of C warm receptors (Iannetti et al. 2003) that produce warm sensations. Our findings are also consistent with the insula activation evoked by warm and noxious thermal contact stimulation (Craig et al. 1996; Strupp 1996), including the heat-evoked ipsilateral insula/frontal opercular response reported by Coghill et al. (1999). However, previous studies should be interpreted cautiously because they typically identify activations related to periods of stimulation during which there may have been mixed qualities of sensation. Given that the insula is composed of a variety of subregions, each of which may subserve different sensory functions, we also tested whether this region was active during other levels of the cooling stimulus. The interesting insular response (see Fig. 3) may impact on previous imaging studies of thermal stimulation-evoked responses (e.g., Coghill et al. 1999; Craig et al. 2000; Davis et al. 1998). For instance, the long duration of PET scans may preclude distinguishing whether a cold-evoked response occurred when subjects perceived cold or heat. Furthermore, until now, subjects have not been asked to report PH during cold stimulation.
The occurrence of PH sensations has puzzled basic scientists and clinicians despite efforts to try to understand its etiology. Several studies have shown that when A-fiber conduction is temporarily blocked, a cold stimulus feels hot (Davis 1998; Fruhstorfer 1984; Fruhstorfer et al. 2003; Mackenzie et al. 1975; Wahren et al. 1989). Patients with either compromised myelinated function or central abnormalities also experience PH sensations (Berglund et al. 2002; Hansen et al. 1996; Ochoa and Yarnitsky 1994; Yosipovitch et al. 1995). These findings led to the hypothesis that unmyelinated primary afferent fibers and central components of the heat pathway play a critical role in PH (Fruhstorfer 1984; Fruhstorfer et al. 2003; Hamalainen et al. 1982; Hansen et al. 1996; Mackenzie et al. 1975; Susser et al. 1999; Wahren et al. 1989; Yarnitsky and Ochoa 1991). A related phenomenon, the thermal grill illusion, is a perception of burning created by alternating strips of cool and warm stimuli. A region of the insula, not dissimilar to the region reported here, was also activated during this perceptual experience (Craig et al. 1996).
The lateralization of the PH response to the right insula is interesting in light of several recent lines of study. Craig (2002, 2003) and Critchley et al. (2004) have highlighted the role of the right insula in interoception and “subjective feeling states.” Other studies have linked the right insula to aspects of attention, awareness, and salience (Downar et al. 2000, 2002). Therefore it is important to note that the PH percept inherently draws ones attention to the PH feeling and that the percept seems to arise somewhat deeply.
Most imaging studies of pain- and thermal-related brain responses employ the general linear model to identify activations that occur “on average” across all subjects. Such imaging studies typically list many cortical areas (including the insula) activated during pain or thermal stimulation (for review, see Peyron et al. 2000). One possible confound in this approach is the possibility of a large contribution to the average from a subset of subjects (Friston et al. 1999). Therefore our use of a conjunction analysis that imposes the criterion that the activation must be significant in each and every subject ensured that the insula activation was a key region for PH perception in all our subjects. This finding strongly supports our supposition that the insula is a critical component for PH perception. The cool-related decrease in activity in this region may serve to sharpen or amplify the effect of the temperature reversal, even though it never actually reaches the hot range. Therefore the behavior of the insula during the cold stimulation indicate that it is not simply a general thermal detection area but has a more complex and perhaps specific role in heat perception.
This study was supported by the Canadian Institutes of Health Research and funds provided by the Canada Research Chair program to K. D. Davis.
The authors thank J. Downar for writing the “peak-finder” software plug-in and G. Detzler and C. Grace for technical assistance.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2004 by the American Physiological Society