A regularization algorithm for decoding perceptual temporal profiles from fMRI data

Designing individual-specific and trial-specific models to accurately predict the intensity of nociceptive pain from single-trial fMRI responses

Using machine learning to predict the intensity of pain from fMRI has attracted rapidly increasing interests. However, due to remarkable inter- and intra-individual variabilities in pain responses, the performance of existing fMRI-based pain prediction models is far from satisfactory. The present study proposed a new approach which can design a prediction model specific to each individual or each experimental trial so that the specific model can achieve more accurate prediction of the intensity of nociceptive pain from single-trial fMRI responses. More precisely, the new approach uses a supervised k-means method on nociceptive-evoked fMRI responses to cluster individuals or trials into a set of subgroups, each of which has similar and consistent fMRI activation patterns. Then, for a new test individual/trial, the proposed approach chooses one subgroup of individuals/trials, which has the closest fMRI patterns to the test individual/trial, as training samples to train an individual-specific or a trial-specific pain prediction model. The new approach was tested on a nociceptive-evoked fMRI dataset and achieved significantly higher prediction accuracy than conventional non-specific models, which used all available training samples to train a model. The generalizability of the proposed approach is further validated by training specific models on one dataset and testing these models on an independent new dataset. This proposed individual-specific and trial-specific pain prediction approach has the potential to be used for the development of individualized and precise pain assessment tools in clinical practice.

Neuroimaging-based biomarkers for pain: state of the field and current directions

Chronic pain is an endemic problem involving both peripheral and brain pathophysiology. Although biomarkers have revolutionized many areas of medicine, biomarkers for pain have remained controversial and relatively underdeveloped. With the realization that biomarkers can reveal pain-causing mechanisms of disease in brain circuits and in the periphery, this situation is poised to change. In particular, brain pathophysiology may be diagnosable with human brain imaging, particularly when imaging is combined with machine learning techniques designed to identify predictive measures embedded in complex data sets. In this review, we explicate the need for brain-based biomarkers for pain, some of their potential uses, and some of the most popular machine learning approaches that have been brought to bear. Then, we evaluate the current state of pain biomarkers developed with several commonly used methods, including structural magnetic resonance imaging, functional magnetic resonance imaging and electroencephalography. The field is in the early stages of biomarker development, but these complementary methodologies have already produced some encouraging predictive models that must be tested more extensively across laboratories and clinical populations.

Quantifying and Characterizing Tonic Thermal Pain Across Subjects From EEG Data Using Random Forest Models

OBJECTIVE

Effective pain assessment and management strategies are needed to better manage pain. In addition to self-report, an objective pain assessment system can provide a more complete picture of the neurophysiological basis for pain. In this study, a robust and accurate machine learning approach is developed to quantify tonic thermal pain across healthy subjects into a maximum of ten distinct classes.

METHODS

A random forest model was trained to predict pain scores using time-frequency wavelet representations of independent components obtained from electroencephalography (EEG) data, and the relative importance of each frequency band to pain quantification is assessed.

RESULTS

The mean classification accuracy for predicting pain on an independent test subject for a range of 1-10 is 89.45%, highest among existing state of the art quantification algorithms for EEG. The gamma band is the most important to both intersubject and intrasubject classification accuracy.

CONCLUSION

The robustness and generalizability of the classifier are demonstrated.

SIGNIFICANCE

Our results demonstrate the potential of this tool to be used clinically to help us to improve chronic pain treatment and establish spectral biomarkers for future pain-related studies using EEG.

Painful Issues in Pain Prediction

How perception of pain emerges from neural activity is largely unknown. Identifying a neural 'pain signature' and deriving a way to predict perceived pain from brain activity would have enormous basic and clinical implications. Researchers are increasingly turning to functional brain imaging, often applying machine-learning algorithms to infer that pain perception occurred. Yet, such sophisticated analyses are fraught with interpretive difficulties. Here, we highlight some common and troublesome problems in the literature, and suggest methods to ensure researchers draw accurate conclusions from their results. Since functional brain imaging is increasingly finding practical applications with real-world consequences, it is critical to interpret brain scans accurately, because decisions based on neural data will only be as good as the science behind them.

A tool for classifying individuals with chronic back pain: using multivariate pattern analysis with functional magnetic resonance imaging data

Chronic pain is one of the most prevalent health problems in the world today, yet neurological markers, critical to diagnosis of chronic pain, are still largely unknown. The ability to objectively identify individuals with chronic pain using functional magnetic resonance imaging (fMRI) data is important for the advancement of diagnosis, treatment, and theoretical knowledge of brain processes associated with chronic pain. The purpose of our research is to investigate specific neurological markers that could be used to diagnose individuals experiencing chronic pain by using multivariate pattern analysis with fMRI data. We hypothesize that individuals with chronic pain have different patterns of brain activity in response to induced pain. This pattern can be used to classify the presence or absence of chronic pain. The fMRI experiment consisted of alternating 14 seconds of painful electric stimulation (applied to the lower back) with 14 seconds of rest. We analyzed contrast fMRI images in stimulation versus rest in pain-related brain regions to distinguish between the groups of participants: 1) chronic pain and 2) normal controls. We employed supervised machine learning techniques, specifically sparse logistic regression, to train a classifier based on these contrast images using a leave-one-out cross-validation procedure. We correctly classified 92.3% of the chronic pain group (N = 13) and 92.3% of the normal control group (N = 13) by recognizing multivariate patterns of activity in the somatosensory and inferior parietal cortex. This technique demonstrates that differences in the pattern of brain activity to induced pain can be used as a neurological marker to distinguish between individuals with and without chronic pain. Medical, legal and business professionals have recognized the importance of this research topic and of developing objective measures of chronic pain. This method of data analysis was very successful in correctly classifying each of the two groups.

Ranking brain areas encoding the perceived level of pain from fMRI data

Pain perception is thought to emerge from the integrated activity of a distributed brain system, but the relative contribution of the different network nodes is still incompletely understood. In the present functional magnetic resonance imaging (fMRI) study, we aimed to identify the more relevant brain regions to explain the time profile of the perceived pain intensity in healthy volunteers, during noxious chemical stimulation (ascorbic acid injection) of the left hand. To this end, we performed multi-way partial least squares regression of fMRI data from twenty-two a-priori defined brain regions of interest (ROI) in each hemisphere, to build a model that could efficiently reproduce the psychophysical pain profiles in the same individuals; moreover, we applied a novel three-way extension of the variable importance in projection (VIP) method to summarize each ROI contribution to the model. Brain regions showing the highest VIP scores included the bilateral mid-cingulate, anterior and posterior insular, and parietal operculum cortices, the contralateral paracentral lobule, bilateral putamen and ipsilateral medial thalamus. Most of these regions, with the exception of medial thalamus, were also identified by a statistical analysis on mean ROI beta values estimated using the time course of the psychophysical rating as a regressor at the voxel level. Our results provide the first rank-ordering of brain regions involved in coding the perceived level of pain. These findings in a model of acute prolonged pain confirm and extend previous data, suggesting that a bilateral array of cortical areas and subcortical structures is involved in pain perception.

Decoding the perception of pain from fMRI using multivariate pattern analysis

Pain is known to comprise sensory, cognitive, and affective aspects. Despite numerous previous fMRI studies, however, it remains open which spatial distribution of activity is sufficient to encode whether a stimulus is perceived as painful or not. In this study, we analyzed fMRI data from a perceptual decision-making task in which participants were exposed to near-threshold laser pulses. Using multivariate analyses on different spatial scales, we investigated the predictive capacity of fMRI data for decoding whether a stimulus had been perceived as painful. Our analysis yielded a rank order of brain regions: during pain anticipation, activity in the periaqueductal gray (PAG) and orbitofrontal cortex (OFC) afforded the most accurate trial-by-trial discrimination between painful and non-painful experiences; whereas during the actual stimulation, primary and secondary somatosensory cortex, anterior insula, dorsolateral and ventrolateral prefrontal cortex, and OFC were most discriminative. The most accurate prediction of pain perception from the stimulation period, however, was enabled by the combined activity in pain regions commonly referred to as the ‘pain matrix’. Our results demonstrate that the neural representation of (near-threshold) pain is spatially distributed and can be best described at an intermediate spatial scale. In addition to its utility in establishing structure-function mappings, our approach affords trial-by-trial predictions and thus represents a step towards the goal of establishing an objective neuronal marker of pain perception.

Understanding chronic inflammatory and neuropathic pain

This meeting report highlights the main topics presented at the conference "Chronic Inflammatory and Neuropathic Pain," convened jointly by the New York Academy of Sciences, MedImmune, and Grünenthal GmbH, on June 2-3, 2011, with the goal of providing a conducive environment for lively, informed, and synergistic conversation among participants from academia, industry, clinical practice, and government to explore new frontiers in our understanding and treatment of chronic and neuropathic pain. The program included leading and emerging investigators studying the pathophysiological mechanisms underlying neuropathic and chronic pain, and experts in the clinical development of pain therapies. Discussion included novel issues, current challenges, and future directions of basic research in pain and preclinical and clinical development of new therapies for chronic pain.

Visualizing the complex brain dynamics of chronic pain

Chronic pain is now recognized as a disease state that involves changes in brain function. This concept is reinforced by data that document structural and morphological remapping of brain circuitry under conditions of chronic pain. Evidence for aberrant neurophysiology in the brain further confirms neuroplasticity at cellular and molecular levels. Proper detection of pain-induced changes using emerging non-invasive and cost-effective technologies, such as analytical electroencephalography methods, could yield objective diagnostic measures and may guide therapeutic interventions targeting the brain for effective management of chronic pain.