Neurophysiological and anatomical considerations in functional imaging of pain.

Over the last decade, there has been an explosion of neuroscience research using functional brain imaging techniques. Advances in these modern imaging technologies have sparked great interest amongst pain researchers, resulting in numerous imaging studies of various aspects of pain and creation of the ‘pain measurement and imaging’ section in this journal. Given this era of rapid developments in imaging designs and analysis approaches, it is prudent to review some basic neurophysiological features of the nociceptive system that could impact the outcome of pain imaging experiments. I encourage ‘pain imagers’ to consider these factors when constructing experimental design and statistical approaches and in their data interpretation. The first imaging studies were applied to simple systems (e.g. vision, motor) and sought signal differences during stimulation and control conditions to identify task-related responses. Subsequently, this basic subtraction approach was extended to studies in more complex systems such as language, memory, attention and pain. The ideal statistical ‘subtraction’ requires two similar conditions that differ only in the specific process of interest (e.g. cognitive subtraction) (Price and Friston, 1997). This subtraction makes perfect sense for many systems. However, in some circumstances, the pain system is not particularly well suited to simple subtraction analyses. The critical assumption in functional magnetic resonance imaging (fMRI) studies of pain is that the blood oxygen level detection (BOLD) signal is sensitive to a nociceptive or pain signal. Extracting pain-specific activations is typically done by comparing BOLD signals during two states: either pain state minus baseline (no stimulation) or pain state minus innocuous stimulation state. The key question is whether the BOLD signal (or blood flow measure in the case of positron emission tomography (PET) imaging) can detect the difference between pain and non-pain state. First, let us consider neurophysiological and anatomical issues. There are three basic categories of neurons that are responsive to somatosensory stimuli—low threshold mechanoreceptive (LTM), wide dynamic range (WDR) and nociceptive specific (NS); and there are also thermoreceptors (warm, cold) which are a kind of low threshold type of neurons that respond to innocuous thermal stimuli. Classically, the LTMs are characterized by their response to innocuous stimuli, the NSs by their sensitivity only to noxious stimuli and the WDRs by their increasing response magnitude to stimuli of increasing stimulus intensity from the innocuous to noxious range. Despite a wealth of information about the subcortical distribution and properties of these neurons, we have only limited knowledge of the whereabouts and properties of these neurons in the primate cortex. The relevance to imaging pain should consider the following. (1) Multiple neuronal populations are activated by experimental pain (Besson and Chaouch, 1987). For instance, contact stimulation devices (thermodes, mechanical probes) can excite both nociceptive and non-nociceptive neurons. WDRs and slowly adapting LTMs will be active during maintained contact and rapidly adapting LTMs will be recruited when the probe is moved. As the stimulus intensity is increased into the noxious range, WDRs will increase their firing rate and NSs will be recruited. Innocuous thermal stimuli will excite innocuous thermoreceptors whereas noxious thermal stimuli will excite temperature-sensitive WDRs and NSs. Non-contact laser stimuli will selectively activate temperature-sensitive neurons. (2) Cortical regions presumed to be involved in pain contain a mixed population of neurons. Primate studies indicate that S1 and other regions implicated in pain perception contain a mixture of pain signalling neurons (NS, WDR) and non-nociceptive neurons (LTM) (Kenshalo and Isensee, 1983; Chudler et al., 1990; Dong et al., 1994; Augustine, 1996; Kenshalo et al., 2000). (3) The numbers of non-nociceptive neurons outnumber the nociceptive neurons. Therefore, the overall effect of increasing the stimulus intensity into the painful range may be modest and make detectability difficult (e.g. partial volume effect). Now, let us consider imaging issues that impact on extracting pain-related activations. (1) Imaging signals are derived from three-dimensional volumes called voxels.