Five cLBP patients (one female) with implanted SCS were compared to five healthy control subjects (two females). Patient information (ex. demographics, cLBP onset, pharmacological pain control) is presented in Table 1. Patients were on stable doses of pain medications and SCS parameters for at least 3 months prior to image acquisition. Informed consent was obtained from all subjects in this study. This study was approved by the Thomas Jefferson University Hospital Institutional Review Board, and all methods were performed in accordance with the Declaration of Helsinki.

All subjects were scanned using a 3T Siemens Prisma MRI scanner (Siemens Healthcare, Erlangen, Germany), with a transmit/receive (Tx/Rx) head coil. Appropriate MRI safety measures were undertaken to scan the subjects. Patients were scanned with the SCS device switched off. All subjects were asked to relax and think of nothing in particular during the rs-fMRI scan.

Axial 3D magnetization prepared rapid gradient echo imaging (MPRAGE). A structural scan was acquired at a 0.9 × 0.9 × 1.5 mm³ voxel size, number of averages = 1, repetition time (TR) = 1,660 ms, echo time (TE) = 2.47 ms, inversion time (TI) = 836 ms, radiofrequency flip angle = 8°, and field-of-view (FOV) = 25 cm. The total acquisition time was 5.62 min. Resting-state fMRI images were acquired axially (35 slices) at a 3.5 × 3.5 × 3.5 mm³ voxel size, number of averages = 1, 300 volumes, TR = 2,000 ms, TE = 26 ms, flip angle = 90°, and FOV = 25 cm. The total acquisition time was 10 min. Multi-delay 3D gradient and spin echo (GRASE) PCASL was acquired using five post-label delays (0.5/1.0/1.5/2.0/2.5 s), at a voxel size of 2.5 × 2.5 × 3.0 mm³. The acquisition parameters were as follows: number of averages = 1, TR = 4,100 ms, TE = 36.48 ms, TI = 1,800 ms, bolus duration = 700 ms, and 13 pairs of tags and control for each delay. The total acquisition time was 7.18 min.

Resting-state fMRI images were preprocessed in MATLAB 2023a (MathWorks, Natick, MA, USA) using the Data Processing and Analysis for Brain Imaging (DPABI) package (Yan et al., 2016). Main steps included the conversion of Digital Imaging and Communications in Medicine (DICOM) to Neuroimaging Informatics Technology Initiative (NIFTI) files, image realignment, T1 image co-registration to functional images, slice timing correction, nuisance covariates (white matter, cerebrospinal fluid, and global motion) regression, head movement correction using Friston 24 model, and spatial normalization. Frequencies were bandpass filtered in the range of 0.01–0.1 Hz, and individual data were normalized to a Montreal Neurological Institute template and smoothed using a full width at half maximum degree of 7 × 7 × 7 mm³. From preprocessed resting-state data, four quantitative maps were derived, namely ALFF, fractional ALFF (fALFF), degree of centrality (DC), and regional homogeneity (ReHo) maps for each individual.

From PCASL data, the quantitative maps of relative CBF, cerebral blood volume (CBV), and arterial transit time (ATT) were derived for each subject using CereFlow software (Translational MRI, LLC, Los Angeles, CA) as described in previous studies (Hu et al., 2019; Benninger et al., 2022).

To examine the specific brain areas involved with the pain Neuromatrix, 26 regions of interest (ROI) from a custom automated anatomical labeling (AAL) atlas (Table 2) were extracted and registered for each patient's rs-fMRI and PCASL data using the Medical Image Registration Toolkit (mirtk.github.io). The regions were comprised of structures detailed in the pain Neuromatrix (Ailes et al., 2023). These were then co-registered into patient space, and average values of rs-fMRI and PCASL maps were collected per ROI for all subjects and were compared between cLBP patients and HCs (Figure 1).

Several different qualitative self-assessments were collected for each cLBP patient on the day of image acquisition. These are the Visual Analog Scale (VAS), Numerical Pain Rating Scale (NPRS), Central Sensitization Inventory (CSI), Oswestry Disability Index (ODI), Pain Catastrophizing Scale (PCS), Pittsburgh Sleep Quality Index (PSQI), and the Hospital Anxiety and Depression Scale (HADS).

Pain intensity was assessed using a VAS from 0 to 10 for current pain and NPRS, which averages the patient's worst, current, and lowest pain in the past day. One patient was not administered the pain VAS due to an oversight in providing the assessment. Pain sensitization was assessed using the CSI, with a score of >40 indicating symptoms of central pain sensitization.

We also administered assessments looking at the degree of disability, quality of life, as well as anxiety and depression. The ODI, which determines a patient's self-reported disability due to their pain, ranges from 0 to 5 for each question. A score of 5 indicates complete disability. The PCS assesses the extent to which pain elicits negative thoughts and feelings, with scores ranging from 0 to 52. A score >30, represents significant pain catastrophizing. The PSQI evaluates sleep quality and its impact on daily activities through various questions, with total scores ranging from 0 to 21. A higher score indicates a worse sleep quality. Finally, the HADS provides information on anxiety and depression subscales, with a score of >8 of 21 indicating significant symptoms of anxiety and/or depression.

All statistics were computed using John's Macintosh Project (JMP) Version 17 (SAS Institute Inc., Cary, NC, USA). Imaging variables per ROI were normalized using the Box–Cox transformation method for both rs-fMRI and PCASL quantitative maps before comparisons between patients and HCs (Sakia, 1992). Two-sample t-tests were used to compare the means for the variables between patients and HCs, using p < 0.05. False discovery rate (FDR) correction was applied to reduce the risk of multiple-comparison false positive errors using t-tests. To explore the association between subjective clinical measures and objective quantitative imaging, linear regression analyses were performed, correlating patient-level quantitative maps for each ROI with clinical assessment scores. Significant associations were those with a p < 0.01.

Figure 3 denotes significant correlations between rs-fMRI and PCASL quantitative measures and clinical scores of pain in the cLBP patients implanted with SCS. The fALFF in the amygdala (AMYG), a region involved in the fear and stress response, was negatively correlated with CSI scores (R² = 0.939, p = 0.0065). The ATT of the DLPC, involved in the cognitive component of pain, was positively correlated with the NPRS scores (R² = 0.92, p = 0.0098). The CBF showed significant correlations in two regions; VPL, the somatosensory relay of the thalamus, was positively correlated with VAS (R² = 0.989, p = 0.0056); and PCC, associated with emotional appraisal, was also positively correlated with VAS (R² = 0.988, p = 0.0059). Finally, the CBV in the PS, a nociceptive input region, was positively correlated with VAS (R² = 0.996, p = 0.0018).

The correlation of imaging-derived functional pain ROIs to clinical measures of pain in cLBP provides evidence of the utility of rs-fMRI and PCASL as measures of pain prognosis, along with quantifying the effects of SCS as a modulator of central pain processing. Despite a small patient population, analyses showed several strong statistically significant linear associations. VAS pain ratings were positively correlated with VPL CBF and PS CBV. As previously described, the VPL thalamus is associated with afferent nociceptive input and the insular PS with pain intensity coding (Afif et al., 2008; Ab Aziz and Ahmad, 2006). Both regions were also found to have increased rs-fMRI/PCASL measures when comparing cLBP patients to HCs. Therefore, we speculate that these regions, part of the input component of the pain Neuromatrix, are likely involved with unidimensional pain perception and rating. Patients with a higher pain rating had increased activity in these regions, suggesting differences in the responsiveness to SCS and the ability of imaging to prognosticate individual responses to SCS.

As ATT in the DLPC increased, there was a positive correlation in the NPRS. Similarly, there was a positive correlation between pain VAS and CBF in the PCC. Elevated cerebral blood flow to the PCC has been associated with self-referential pain catastrophizing, particularly as a component of the default mode network (DMN) (Lee et al., 2018). The DMN, which includes the PCC, DLPC, and other regions, is associated not only with passive brain processing but also has various cognitive and emotional appraisals of chronic pain. Therefore, it is possible that increased functional activation as indexed by perfusion in the PCC and DLPC in our patients may also lead to increased unidimensional rating of pain due to modulation of the DMN in cLBP patients implanted with SCS.

Finally, there was a statistically significant negative correlation between fALFF of the amygdala and CSI. The amygdala has an important function in the emotional and fear aspects of pain processing in cLBP, with it being associated with individual differences in emotional pain facilitation (Gandhi et al., 2020; Mao et al., 2022). A recent study found that reduced resting-state functional connectivity in the amygdala was correlated with mechanical pain rating in low-back-related leg pain patients (Wang et al., 2023). In our patients with SCS we observed the opposite effect, which suggests that the bottom-up effects of SCS contribute to the improvement of the emotional response in cLBP. Maladaptive anticipation and fear of pain in cLBP likely contribute to central sensitization in patients, with individual differences and the effects of SCS on the amygdala's functional state being a potential quantitative measure of the reversal of this process.

Rather than relying on subjective and qualitative measures of pain perception via the aforementioned pain scales, PCASL and rs-fMRI have the potential to be utilized as quantitative measures to evaluate treatment effectiveness. However, since we did not obtain baseline imaging of our cLBP patients before SCS, we cannot conclude whether changes in processing may be due to SCS, chronic pain experienced by patients, or both. Future studies are warranted to assess pre- vs. post-SCS changes in functional activation in cerebral pain regions, along with how these correlations change over this period.

The present study has limitations. First, we have a lack of baseline preoperative imaging of cLBP patients. This would serve as a comparator to the effects of SCS stimulation in this patient population, rather than relying strictly on comparisons to healthy controls. However, as a feasibility study, our purpose was to illustrate that rs-fMRI and PCASL from 3T MRI are effective modalities for distinguishing between healthy subjects and patients implanted with SCS. Additionally, these modalities are associated with clinical scores of pain in cLBP patients who have been implanted with SCS. Future studies are warranted to assess changes in functional imaging pre- vs. post-SCS in cLBP patients. This may even be used as a screening tool to eventually predict response to SCS by correlating the change in imaging measures with clinical response. By the same token, a comparison of MRI measures using the SCS “on” vs. “off” settings would provide an assessment of the immediate, “online” effects of SCS in cLBP patients, and future investigations are warranted for this understudied effect. Second, our limited sample size may restrict the reproducibility of results, and we did not control for age or sex. However, we utilized FDR correction, and correlation analyses were highly significant. Additionally, our patients may have other comorbidities, medications, and individual differences in their SCS parameters and timeline of obtaining SCS relative to cLBP onset. Nevertheless, functional imaging of pain regions in cLBP patients implanted with SCS is a greatly underexplored area of research that warrants further investigation, assessing differences in pain states based on SCS parameters and paradigms (e.g., paresthesia-based vs. burst vs. 10 KHz) and timeline of changes based on duration of SCS, using a larger sample size. Finally, although our patients had differing responses to SCS in regard to pain control, our correlation analyses do not provide a direct comparison of patients who obtain excellent pain control vs. those who do not. Future studies are warranted to compare these groups using a larger, predetermined group of patients, as this would shed light on the central pain mechanism as well as potentially improve the presurgical indications of SCS based on the neural activity of cLBP patients.

This pilot study explored pain states in cLBP patients implanted with SCS using rs-fMRI and PCASL, safely derived from 3T MRI. Changes in rs-fMRI and PCASL measures were observed when comparing patients to healthy control subjects, and activation of several pain regions was associated with clinical scores of pain rating and sensitization. Specifically, the frontal and limbic cortex structures, VPL thalamus, and posterior short gyrus of the insula were all associated with changes in pain rating and/or sensitization. The utilization of advanced imaging techniques in cLBP patients implanted with SCS may be eventually developed as a quantitative tool in patients and is vital for monitoring other conditions, presurgical planning, and the study of chronic pain and its neuromodulation.