Methodological Evolution of Diffusion MRI in Chronic Pain: A Narrative Review Authors: Hairui Li, Yuqin Wang, Mingtao Li, Hongan Yang, Zhiqiang Qiu, Xiaoxue Xu Published: 2026 ABSTRACT AbstractChronic pain is a highly prevalent and clinically heterogeneous condition, and its underlying mechanisms remain incompletely understood. The lack of objective, clinically validated biomarkers has hindered diagnostic stratification, mechanistic insight, and treatment evaluation. Diffusion magnetic resonance imaging (dMRI) noninvasively quantifies water diffusion in tissue and provides indirect indices of microstructural organization and structural connectivity in the central and peripheral nervous systems, making it a widely used tool for studying pain chronification and candidate imaging markers. This narrative, methodologically oriented review traces how dMRI has been applied in chronic pain research along its technical trajectory—from diffusion tensor imaging (DTI) to multi-shell acquisition with advanced microstructural models, including diffusion kurtosis imaging (DKI), neurite orientation dispersion and density imaging (NODDI), and constrained spherical deconvolution (CSD), as well as the increasing integration with multimodal imaging. PubMed and Web of Science were searched from inception to December 2025 for studies applying dMRI in participants with chronic pain and reporting quantitative diffusion metrics. We evaluate the strengths and limitations of different dMRI approaches in the context of chronic pain and discuss their potential to improve diagnosis, biomarker discovery, and outcome prediction. Findings are appraised alongside the evidentiary limits of this literature, including small and predominantly cross-sectional samples, heterogeneous acquisition and preprocessing pipelines, and variable statistical correction. Current dMRI metrics are therefore best regarded as promising research markers rather than validated clinical tools, and standardized acquisition, multicenter replication, and prospective individual-level validation remain necessary. Introduction Chronic pain is typically defined as persistent pain lasting beyond three months, or present on at least half the days over a six-month period. This duration exceeds the expected timeframe for normal tissue healing, and the pain no longer serves the adaptive, protective function characteristic of acute nociception.1–4 Chronic pain poses a significant burden on individuals and society. Taking back pain as an illustrative example, in 2020 the global prevalence of low back pain exceeded 500 million cases, accounting for 7.7% of all years lived with disability (YLDs). Projections indicate that the total number of cases will increase by 36.4% by 2050, underscoring its status as a major public health concern. Affected individuals often report severely diminished quality of life and persistently unmet treatment needs, while healthcare systems face mounting pressure from this growing patient population. Concurrently, associated healthcare expenditures and productivity losses due to work absenteeism continue to rise.5–8 The pathophysiology of chronic pain is highly complex. A substantial body of evidence from both human and animal studies indicates that chronic pain is closely associated with reorganization of the peripheral and central nervous systems, accompanied by a spectrum of neuronal and glial alterations that are linked to the persistence of pain.9 Diffusion magnetic resonance imaging (dMRI) offers a non-invasive method to probe such microstructural changes in vivo. It has thus been widely used for investigating white matter microstructure and structural connectivity in the human brain.10–12 Despite this promise, dMRI-derived measures are not yet established biomarkers for chronic pain. Diffusion metrics are sensitive to acquisition design, including b-value sampling, the number and distribution of diffusion-encoding directions, spatial resolution, and scanner field strength. They are also affected by susceptibility-induced echo-planar distortion, head motion, and choices made during denoising, distortion correction, registration, and model fitting. Consequently, nominally equivalent estimates can differ appreciably across scanners, protocols, and analysis pipelines, limiting cross-study comparability and the pooling of multi-site data.13 In chronic pain research, acquisition and preprocessing practices remain heterogeneous, and the available evidence is based largely on small cohorts reporting group-level differences or group–clinical associations. Clinical translation would additionally require demonstrated test–retest reliability in relevant patient populations, transportability across sites and populations, and diagnostic or prognostic performance at the individual-patient level.4 Advanced diffusion models also involve important practical trade-offs. Multi-shell acquisitions and high-angular-resolution schemes can improve microstructural specificity and the characterization of complex fiber architecture.14 However, they generally require longer scan times, which may be difficult for patients with ongoing pain to tolerate. Such acquisitions also place greater demands on signal quality and gradient performance. Their analysis often relies on computationally demanding model-fitting and tractography pipelines, in which analytical assumptions and parameter settings can influence the resulting estimates.11,15 Accordingly, these protocols have not yet been widely adopted in routine clinical imaging. Conventional single-shell DTI therefore remains valuable when acquisition efficiency, accessibility, and scalability are priorities, particularly in large cohorts and prognostic studies. The methodological generations reviewed here should be regarded as complementary tools selected according to the clinical question and acquisition constraints, rather than as a linear hierarchy in which newer models invariably replace earlier approaches. Existing reviews of dMRI in chronic pain have largely been organized by pain diagnosis or by anatomical finding, and have focused on cerebral white matter in chronic primary pain.16 By contrast, our review is organized around diffusion methodology rather than pain diagnosis. Unlike previous reviews that focused predominantly on cerebral white matter, this review examines dMRI across the brain, brainstem, spinal cord, and peripheral and cranial nerves. It also covers all seven categories defined in the ICD-11 classification of chronic pain.1,5 This framework traces the methodological evolution of dMRI from conventional single-shell DTI to multi-shell acquisitions, advanced microstructural models, tractography and structural connectomics, and multimodal integration. We synthesize findings across diffusion approaches, compare their strengths and limitations, and critically appraise the evidence supporting their proposed roles in pain diagnosis, biomarker development, and treatment-response prediction. This narrative review synthesizes the literature on diffusion MRI in chronic pain. PubMed and Web of Science were searched from inception to December 31, 2025, combining controlled vocabulary with free-text terms for chronic pain and for diffusion MRI and its analysis frameworks. Free-text terms were necessary because newer models, including NODDI and fixel-based analysis, lack dedicated MeSH descriptors. The database search was supplemented by manual screening of the reference lists of relevant articles. Title and abstract screening, full-text assessment, and study inclusion were conducted by one author and reviewed by a coauthor. Eligible studies were peer-reviewed original reports in English that enrolled participants with pain persisting or recurring for at least three months and reported quantitative diffusion metrics. Case reports, abstracts without full text, and studies of acute or experimentally induced pain were excluded. The review is organized primarily by diffusion methodology rather than by pain diagnosis; therefore, studies were selected to illustrate the capabilities and limitations of successive generations of diffusion models. Within each ICD-11 category, longitudinal studies and those applying advanced models were preferred, but for most categories neither was available—an unevenness that reflects the distribution of the existing literature rather than selective citation. Exhaustive coverage of every chronic pain diagnosis is therefore not claimed. Applications and Findings of Single-Shell Diffusion MRI in Chronic Pain Fundamental Principles of Diffusion Tensor Imaging Diffusion tensor imaging (DTI), derived from diffusion-weighted MRI, is the earliest and most widely applied diffusion model in chronic pain research.16 The core principle of DTI is its extension of the conventional apparent diffusion coefficient (ADC) to a diffusion tensor, which allows full characterization of the anisotropic diffusion of water molecules in three-dimensional space.12 DTI employs a single-shell acquisition scheme (typically at b= 1000s/mm2) and assumes Gaussian diffusion, providing core metrics including fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD). FA, the most widely used metric, reflects the directionality of water diffusion and serves as a common proxy for white matter integrity. MD represents the overall magnitude of diffusion; an elevation in MD often suggests underlying tissue damage or edema.17 Applications of Diffusion Tensor Imaging in Chronic Pain Alterations in White Matter Microstructure DTI allows the non-invasive, in vivo characterization of neural fiber tracts. Its application to investigate the effects of chronic pain on cerebral white matter has thus become a central focus in the field. Accumulating evidence indicates that patients with chronic pain exhibit significant microstructural alterations in white matter, which correlate with clinical symptom severity.18–20 Using tract-based spatial statistics (TBSS), Filimonova et al reported widespread reductions in fractional anisotropy (FA) in patients with primary trigeminal neuralgia (PTN), affecting regions including the bilateral corona radiata, internal capsule, thalamic white matter, and cerebellar peduncles.21 Huang et al utilized DTI-based structural connectomics with graph theory analysis to identify altered nodal efficiency within the frontoparietal control network in patients with chronic prostatitis/chronic pelvic pain syndrome (CP/CPPS); these network alterations significantly correlated with clinical symptom severity.22 In fibromyalgia, Kim et al used tract-based spatial statistics and identified lower FA in the left body of the corpus callosum. This alteration was associated with greater sensory pain severity.23 Tu et al subsequently applied probabilistic diffusion tractography and graph-theoretical analysis to examine whole-brain structural networks. Patients showed reduced network segregation and integration, with disrupted structural connectivity involving the prefrontal cortex, basal ganglia, and thalamus.24 However, the network measures were not significantly associated with clinical variables after correction for multiple comparisons. Painful chronic pancreatitis provides an example of chronic secondary visceral pain. Frøkjær et al reported altered diffusivity and anisotropy in the amygdala, cingulate cortex, insula, prefrontal cortex, and secondary somatosensory cortex. Selected cingulate and prefrontal measures were also associated with clinical pain scores.25 In a seven-year follow-up study, Muthulingam et al found widespread reductions in FA at baseline.26 However, subsequent whole-brain FA decline was similar in patients and healthy controls. Baseline abnormalities were also associated with smoking exposure and disease duration. These findings illustrate the value of longitudinal diffusion MRI for distinguishing pain-related neuroplasticity from disease-related and systemic confounding effects. For chronic pain following mild traumatic brain injury, Kim et al used diffusion tensor tractography to examine the spinothalamocortical tract, finding that patients with central pain showed lower fractional anisotropy and tract volume.27 Evidence in cancer-related pain remains limited. In patients with lung cancer, Li et al used automated fiber quantification to compare patients with pain, painless cancer patients, and healthy controls. The pain group showed whole-tract and node-specific diffusion abnormalities, particularly in the right hippocampal cingulum and inferior longitudinal fasciculus. Selected diffusion measures were also associated with pain intensity or duration.28 Collectively, these findings indicate that white matter abnormalities in chronic pain extend beyond canonical nociceptive pathways, reflecting a broader, brain-wide reorganization of structural networks. In addition to delineating associations between white matter microstructure and pain severity, DTI has shown promise as a biomarker for predicting the clinical trajectory and treatment response in chronic pain. A longitudinal study by Mansour et al revealed that baseline FA differences in the subacute phase of low back pain could predict pain outcomes at one year. Patients whose pain persisted exhibited significantly lower FA in tracts such as the left superior longitudinal fasciculus and internal capsule compared to those who recovered.29 These findings were further corroborated by a multicenter study conducted by Mišić et al, which underscored the robustness of white matter microstructure as a prognostic marker. The authors identified higher FA values in the right superior longitudinal fasciculus (SLF) as a candidate prognostic marker of recovery in patients with subacute low back pain, a result that was consistently replicated across three independent cohorts.30 Collectively, these studies suggest that DTI-derived white matter measures may provide complementary prognostic information, but their reliability, incremental value beyond conventional clinical assessments, and utility for individual-level decision-making remain to be established. Microstructural Alterations in Peripheral or Cranial Nerves Beyond central white matter, studies have also examined microstructural changes in peripheral and cranial nerves. In trigeminal neuralgia (TN), for example, Hung et al reported that preoperative AD of the affected trigeminal nerve at the pontine segment, along with threshold-based metrics derived from AD, MD, and FA, could effectively discriminate patients who were nonresponsive to surgical intervention.31 Furthermore, Willsey et al extracted RD from the pontine segment of the trigeminal nerve and normalized these values to background pontine RD. By integrating these imaging features with preoperative symptom duration, the authors developed a machine-learning classifier that successfully differentiated patients with postoperative pain relief from those with pain recurrence, reinforcing that DTI encodes clinically relevant prognostic information.32 Similarly, Zhang et al used DTI tractography to automatically extract diffusion metrics from the spinal trigeminal tract (SpTV) and performed multivariate analyses incorporating neurovascular contact (NVC) grading, further substantiating the value of DTI parameters as predictive biomarkers.33 Taken together, these studies indicate that incorporating quantitative DTI metrics from key segments of the trigeminal nerve into preoperative assessment may help inform surgical management in trigeminal neuralgia. Furthermore, this work provides a methodological framework for identifying predictive biomarkers in other peripheral neuropathic pain conditions. These prognostic models were developed in small single-center cohorts without testing in independent samples, so their discriminative performance requires external validation. In a study of patients with cervicogenic headache (CGH), Wang et al found that FA values of the greater occipital nerve (GON) and the C2/C3 dorsal root ganglia (DRG) on the symptomatic side were significantly lower than those on the asymptomatic side, whereas apparent diffusion coefficient (ADC) values were markedly increased.34 These observations further support that chronic pain is linked to microstructural changes in peripheral nerves. Collectively, DTI research indicates that various chronic pain conditions are associated with alterations in neural microstructure. Notably, these alterations may correlate with clinical symptoms, neural function, and outcomes. Several limitations qualify the studies reviewed above. Most were cross-sectional and conducted at single centers, and the acquisition protocols, analysis frameworks, and approaches to multiple-comparison correction differed substantially between them, which constrains direct comparison. Few examined whether group-level differences were reproducible at the individual level. FA and MD are also biologically non-specific: comparable changes may arise from axonal loss, altered myelination, increased extracellular water, or partial-volume and crossing-fiber effects. Interpretation is further complicated in populations where systemic disease or its treatment may affect white matter independently of pain. These limitations have prompted researchers to seek more advanced diffusion imaging techniques.35–37 Applications and Findings of Multi-Shell Acquisition and Advanced Diffusion Modeling DTI assumes water diffusion follows a Gaussian distribution. In biological tissues, however, microstructural barriers—including cell membranes and myelin sheaths—restrict water motion, leading to intrinsically non-Gaussian diffusion. As a consequence, DTI lacks accuracy in characterizing diffusion properties within complex microenvironments, such as regions with crossing fibers, and exhibits limited pathological specificity.38,39 To better characterize white matter microstructure in regions of fiber crossing, advanced diffusion imaging techniques have been developed, most notably high angular resolution diffusion imaging (HARDI). HARDI is characterized by high angular resolution, which is achieved through the acquisition of a large number of diffusion-encoding directions and was initially implemented using single high b-value shell acquisitions.40 With advances in acquisition methods, contemporary studies increasingly combine HARDI with multi-shell protocols, enabling the use of more sophisticated multi-compartment models.14,41 Building on this progress, several diffusion models and tract-based analysis techniques have been introduced to examine white matter microstructure from complementary perspectives. Among the most widely applied are neurite orientation dispersion and density imaging (NODDI), diffusion kurtosis imaging (DKI), and constrained spherical deconvolution (CSD). A comparative overview of their acquisition strategies and analytical features is presented in Table 1.Table 1Comparison of Key Features of Different Diffusion MRI Models and Analysis Frameworks (Applicable to Chronic Pain Research)Model/MethodMinimum Acquisition/Analysis Requirements (Recommended)Key Metrics & Interpretive DimensionsUse Cases & Limitations (Highlights)DTI (Single-Shell)b=800–1000 s/mm2, ≥30 directions (recommended ≥45); b=0 s/mm2, optional reverse phase-encoding for EPI distortion correctionFA, MD, AD, RD: anisotropy and overall diffusivity (approximate WM integrity)Advantages: clinically feasible; short scan time; suitable for large cohorts and prognostic/correlation analyses.Limitation: non-specific in crossing fibers, partial volume and free-water effects; sensitive to motion/distortion/SNR; limited cross-site comparability.DKI (Multi-Shell)≥2 non-zero b-values with sufficient directions; higher SNR requirement than DTI; longer scan timeMK, K∥, K⊥: non-Gaussian diffusion; microstructural complexity/heterogeneityAdvantages: more sensitive than DTI to subtle changes; potential for gray matter and peripheral nerves.Limitation: fitting sensitive to noise and acquisition design; parameter stability requires evaluation; biological interpretation needs convergent evidence.NODDI (Multi-Shell, Three-Compartment)Multi-shell acquisition (≥2 non-zero b-values; ≥30 directions per shell recommended) with strict preprocessingNDI, ODI, ISOVF/FISO: neurite density, orientation dispersion, free-water fraction (greater biophysical interpretability)Advantages: partially separates neurite alterations from increased free water, facilitating mechanistic hypotheses (eg, development, inflammation).Limitation: simplified model assumptions and possible parameter degeneracy; protocol/software sensitive; individual-level stability and thresholds require external validation.CSD (HARDI)High angular resolution (single-shell high-b or multi-shell; ≥45–60 directions recommended)FOD; FD, FC, FDC: resolves crossing fibers; quantifies fiber density, bundle cross-section and their combinationAdvantages: resolves crossing fibers; suitable for fine tract quantification, connectomics and along-tract analyses; sensitive to network remodeling.Limitation: complex pipeline; sensitive to acquisition and preprocessing; tractography and filtering choices affect results; cross-site standardization and reproducibility remain challenging.Abbreviations: DTI, diffusion tensor imaging; FA, fractional anisotropy; MD, mean diffusivity; AD, axial diffusivity; RD, radial diffusivity; DKI, diffusion kurtosis imaging; MK, mean kurtosis; K∥, axial kurtosis; K⊥, radial kurtosis; SNR, signal-to-noise ratio; HARDI, high angular resolution diffusion imaging; WM, white matter; NODDI, neurite orientation dispersion and density imaging; NDI, neurite density index; ODI, orientation dispersion index; ISOVF/FISO, isotropic volume fraction/fraction of isotropic volume; CSD, constrained spherical deconvolution; EPI, echo-planar imaging; FD, fiber density; FC, fiber cross-section; FDC, fiber density and cross-section; FOD, fiber orientation distribution; ≥, greater than or equal to. Table 1Comparison of Key Features of Different Diffusion MRI Models and Analysis Frameworks (Applicable to Chronic Pain Research)Model/MethodMinimum Acquisition/Analysis Requirements (Recommended)Key Metrics & Interpretive DimensionsUse Cases & Limitations (Highlights)DTI (Single-Shell)b=800–1000 s/mm2, ≥30 directions (recommended ≥45); b=0 s/mm2, optional reverse phase-encoding for EPI distortion correctionFA, MD, AD, RD: anisotropy and overall diffusivity (approximate WM integrity)Advantages: clinically feasible; short scan time; suitable for large cohorts and prognostic/correlation analyses.Limitation: non-specific in crossing fibers, partial volume and free-water effects; sensitive to motion/distortion/SNR; limited cross-site comparability.DKI (Multi-Shell)≥2 non-zero b-values with sufficient directions; higher SNR requirement than DTI; longer scan timeMK, K∥, K⊥: non-Gaussian diffusion; microstructural complexity/heterogeneityAdvantages: more sensitive than DTI to subtle changes; potential for gray matter and peripheral nerves.Limitation: fitting sensitive to noise and acquisition design; parameter stability requires evaluation; biological interpretation needs convergent evidence.NODDI (Multi-Shell, Three-Compartment)Multi-shell acquisition (≥2 non-zero b-values; ≥30 directions per shell recommended) with strict preprocessingNDI, ODI, ISOVF/FISO: neurite density, orientation dispersion, free-water fraction (greater biophysical interpretability)Advantages: partially separates neurite alterations from increased free water, facilitating mechanistic hypotheses (eg, development, inflammation).Limitation: simplified model assumptions and possible parameter degeneracy; protocol/software sensitive; individual-level stability and thresholds require external validation.CSD (HARDI)High angular resolution (single-shell high-b or multi-shell; ≥45–60 directions recommended)FOD; FD, FC, FDC: resolves crossing fibers; quantifies fiber density, bundle cross-section and their combinationAdvantages: resolves crossing fibers; suitable for fine tract quantification, connectomics and along-tract analyses; sensitive to network remodeling.Limitation: complex pipeline; sensitive to acquisition and preprocessing; tractography and filtering choices affect results; cross-site standardization and reproducibility remain challenging.Abbreviations: DTI, diffusion tensor imaging; FA, fractional anisotropy; MD, mean diffusivity; AD, axial diffusivity; RD, radial diffusivity; DKI, diffusion kurtosis imaging; MK, mean kurtosis; K∥, axial kurtosis; K⊥, radial kurtosis; SNR, signal-to-noise ratio; HARDI, high angular resolution diffusion imaging; WM, white matter; NODDI, neurite orientation dispersion and density imaging; NDI, neurite density index; ODI, orientation dispersion index; ISOVF/FISO, isotropic volume fraction/fraction of isotropic volume; CSD, constrained spherical deconvolution; EPI, echo-planar imaging; FD, fiber density; FC, fiber cross-section; FDC, fiber density and cross-section; FOD, fiber orientation distribution; ≥, greater than or equal to. Comparison of Key Features of Different Diffusion MRI Models and Analysis Frameworks (Applicable to Chronic Pain Research) Abbreviations: DTI, diffusion tensor imaging; FA, fractional anisotropy; MD, mean diffusivity; AD, axial diffusivity; RD, radial diffusivity; DKI, diffusion kurtosis imaging; MK, mean kurtosis; K∥, axial kurtosis; K⊥, radial kurtosis; SNR, signal-to-noise ratio; HARDI, high angular resolution diffusion imaging; WM, white matter; NODDI, neurite orientation dispersion and density imaging; NDI, neurite density index; ODI, orientation dispersion index; ISOVF/FISO, isotropic volume fraction/fraction of isotropic volume; CSD, constrained spherical deconvolution; EPI, echo-planar imaging; FD, fiber density; FC, fiber cross-section; FDC, fiber density and cross-section; FOD, fiber orientation distribution; ≥, greater than or equal to. Applications and Findings of Neurite Orientation Dispersion and Density Imaging (NODDI) in Chronic Pain Neurite orientation dispersion and density imaging (NODDI) is a multi-compartment diffusion MRI model that separates the voxel-wise diffusion signal into three tissue compartments: intra-neurite, extra-neurite, and isotropic free water. This framework provides estimates of biologically relevant microstructural parameters, including the neurite density index (NDI), orientation dispersion index (ODI), and the isotropic volume fraction/fraction of isotropic volume (ISOVF/FISO). The intra-neurite compartment represents water restricted within axons and dendrites. The extra-neurite compartment corresponds to water in the extracellular space around neurites, where diffusion is hindered by the geometry and orientation of surrounding neurites. The isotropic free water compartment reflects freely diffusing water, such as that in cerebrospinal fluid and other fluid-filled spaces.14 Compared with DTI, neurite orientation dispersion and density imaging offers a more precise characterization of microstructural alterations—an advantage validated by histopathological studies in conditions such as multiple sclerosis.42,43 By quantifying white matter microstructural changes in a manner that affords greater biological interpretability, NODDI offers a novel imaging framework for elucidating neurite-level pathological mechanisms underlying chronic pain. Consequently, an increasing number of studies have begun to apply NODDI to the investigation of chronic pain conditions. NODDI has been applied to study microstructural changes across various chronic pain conditions. In one study of new daily persistent headache (NDPH)—a rare, treatment-refractory primary headache—Li et al found significantly lower NDI in patients compared with healthy controls. Reductions were observed in multiple white matter tracts, including the inferior fronto-occipital fasciculus, inferior longitudinal fasciculus, and superior longitudinal fasciculus, as well as in cortical gray matter regions such as the superior frontal gyrus, orbitofrontal cortex, and insula. These findings indicate widespread neurite density loss, with the affected regions largely implicated in pain processing, emotional regulation, and cognitive function.44 Xiong et al applied NODDI to patients with unilateral trigeminal neuralgia (TN), acquiring diffusion MRI data and performing both tract-based spatial statistics (TBSS) and atlas-based region-of-interest analyses to compare whole-brain white matter and selected gray matter nuclei between groups. Their results demonstrated that patients with either left- or right-sided TN exhibited bilateral reductions in neurite density, accompanied by increased free water fractions across both cerebral hemispheres. These alterations were interpreted by the authors as compatible with neuroinflammatory edema, although the isotropic free-water fraction is not specific to inflammation and this interpretation remains hypothetical.45 In pediatric and adolescent populations, NODDI applications in chronic pain remain limited. In a recent study of youths aged 10–24 years with non-disease-related chronic pain, Timmers et al performed the first NODDI investigation in this population. Using tract-based spatial statistics (TBSS), they found widespread increases in ODI across multiple white matter tracts—including the bilateral corona radiata, corticospinal tract, superior longitudinal fasciculus, left uncinate fasciculus, and anterior corpus callosum—suggesting more dispersed fiber orientations. Complementary automated fiber quantification (AFQ) analyses further revealed elevated NDI in specific tracts such as the right cingulum bundle and left corticospinal tract. Notably, microstructural changes across several white matter tracts were significantly correlated with pain catastrophizing. Stronger associations were observed in tracts connecting cortical and limbic regions, including the uncinate fasciculus, cingulum, and anterior thalamic radiation. Together, these findings indicate that white matter microstructural reorganization—particularly increased orientation dispersion—may be a key neurobiological feature of chronic pain in adolescence. The association between these alterations and pain catastrophizing further implicates their role in pain chronification as well as in emotional and cognitive processes during neurodevelopment.46 Taken together, these studies illustrate how NODDI can be applied across various chronic pain populations, offering a more sensitive and biologically specific approach to characterizing their microstructural underpinnings. By detecting alterations in neural microstructure, NODDI provides a new lens for examining the neurobiological heterogeneity of pain chronification and advances our mechanistic understanding across different pain conditions. At the same time, several constraints apply to the NODDI studies reviewed above. They remain few in number, are confined to a narrow range of pain conditions, and are predominantly cross-sectional, so the consistency of reported effects across pain phenotypes and their temporal stability cannot yet be assessed. NODDI also rests on fixed assumptions about intrinsic diffusivity and compartment geometry; where these assumptions are violated, parameter estimates may be biased, and NDI and ODI can be interdependent rather than independently interpretable Estimates are further sensitive to the number of shells, b-value selection, and the fitting software used, and normative ranges and test–retest reliability in patient populations remain to be established. Because the intra-neurite compartment encompasses both axons and dendrites, a reduction in NDI does not by itself localize the change to a particular cellular element, and the cross-sectional designs used to date leave open whether such differences precede or follow the onset of pain. Applications and Findings of Diffusion Kurtosis Imaging (DKI) in Chronic Pain Diffusion kurtosis imaging (DKI) is a higher-order extension of diffusion-weighted imaging (DWI) designed to characterize the diffusion properties of water molecules in biological tissues. Unlike diffusion tensor imaging (DTI), which assumes that water diffusion follows a Gaussian distribution, DKI explicitly accounts for the non-Gaussian nature of diffusion in complex tissue environments. Building upon the second-order three-dimensional diffusion tensor employed in conventional DTI, DKI incorporates an additional fourth-order three-dimensional kurtosis tensor, thereby enabling a higher-order in vivo description of water diffusion. This advanced modeling framework provides enhanced sensitivity for detecting microstructural alterations in neural tissue.47–49 By introducing an additional kurtosis term, DKI is able to capture diffusion abnormalities arising from hindrance or restriction imposed by complex cellular architecture. Previous studies have demonstrated that DKI-derived metrics exhibit greater stability and accuracy than those obtained with DTI, including DTI fitted using dual b-value acquisitions. Importantly, DKI requires only a modest increase in scan time and magnetic resonance hardware demands, yet provides additional, valuable information regarding tissue microstructural complexity and directional anisotropy.50,51 Through this higher-order modeling, DKI yields diffusion metrics that are more sensitive to neural tissue microstructure than those from conventional DTI.52,53 The principal parameters derived from DKI include the apparent diffusion coefficient (ADC), the apparent kurtosis coefficient (AKC), mean kurtosis (MK), axial kurtosis (Ka or K∥), and radial kurtosis (Kr or K⊥). Among these measures, mean kurtosis (MK) is the most widely used parameter and is defined as the average kurtosis computed across all diffusion directions. MK reflects the overall degree of non-Gaussian diffusion in tissue. Higher MK values indicate greater restriction of water diffusion, typically corresponding to increased microstructural complexity or heterogeneity.54 Central Nervous System Microstructural Alterations In patients with left-sided classical trigeminal neuralgia, Xiong et al used DKI to show that, alongside widespread reductions in FA and increases in MD in white matter, significant decreases in MK also occurred in gray matter nuclei, including the bilateral thalamus and caudate nucleus. These MK reductions were more pronounced in the left thalamus, showing a clear left-lateralized asymmetry. Moreover, both disease duration and pain intensity were negatively correlated with MK in the thalamus, internal capsule, and related regions, linking microstructural disruption to clinical pain features.55 In patients with postherpetic neuralgia (PHN), Zhang et al applied DKI and found significant reductions in Ka, MK, and Kr across multiple brain regions, including the bilateral insula, superior temporal gyrus, right thalamus, and caudate nucleus. Notably, Ka in the bilateral insula and MK in the right insula were negatively correlated with visual analog scale (VAS) scores, indicating that microstructural changes in these pain-related regions reflect pain severity.56 DKI has also been applied to fibromyalgia as a representative chronic primary pain condition. Lo et al reported lower kurtosis-related measures in the bilateral dorsolateral prefrontal and orbitofrontal cortices. Amygdala axial kurtosis was also associated with clinical pain severity. The authors interpreted these findings, together with circulating neuronal biomarkers, as being compatible with neuroinflammatory processes.57 However, DKI does not directly measure neuroinflammation, and the study included only 14 patients and 14 healthy controls. These results therefore support a microstructural inflammatory hypothesis rather than establishing a specific pathological mechanism. Peripheral Nerve Microstructural Alterations Qi et al were the first to combine DTI and DKI to assess the microstructural integrity of the trigeminal nerve itself. Their analysis revealed significantly reduced FA, MK, and Kr, alongside an increased apparent diffusion coefficient (ADC), on the affected side. Receiver operating characteristic (ROC) analysis using the difference ratio (DR) identified FA and DR as the top diagnostic markers, with an area under the curve (AUC) of 0.974.58 These results indicate that DTI and DKI can detect microstructural damage in the trigeminal nerve, with FA and DR showing the highest discriminative performance in this cohort. Collectively, these observations suggest that DKI is highly sensitive to microstructural abnormalities in both peripheral nerves and central white matter and gray matter in patients with chronic pain, although what these abnormalities represent at the tissue level remains uncertain. Moreover, these findings suggest that DKI-derived metrics may be more sensitive than conventional DTI parameters in detecting microstructural changes associated with clinical manifestations. However, these studies were limited by small samples, single-center designs, and the absence of external validation. Diffusion kurtosis imaging also places greater demands on data acquisition than conventional DTI, typically requiring higher signal-to-noise ratios and denser diffusion sampling. Kurtosis estimates are sensitive to the selected b-value range, fitting algorithm, and residual motion, which complicates comparisons across scanners and sites. Although increased kurtosis is often interpreted as reflecting greater microstructural complexity, similar changes may arise from variations in cellularity, membrane density, or free-water content. Kurtosis therefore lacks a unique biological interpretation. Applications and Findings of Constrained Spherical Deconvolution (CSD) in Chronic Pain Constrained spherical deconvolution (CSD) is a white matter modeling technique derived from high angular resolution diffusion imaging (HARDI). Its principle involves representing the voxel-wise diffusion signal as the spherical convolution of a single-fiber response function with a fiber orientation distribution (FOD). By applying a non-negativity constraint during deconvolution, CSD estimates the orientations and relative densities of multiple fiber populations within individual voxels.59,60 Compared with DTI—which assumes a single dominant diffusion direction per voxel—CSD resolves multiple crossing, branching, or fanning fiber bundles within individual voxels, greatly improving the characterization of complex white matter geometry. The amplitude of the FOD is considered to reflect, in part, the axonal volume fraction or fiber density along a given orientation. From FOD-based measures, quantitative metrics such as fiber density (FD), fiber cross-section (FC), and their combined index (FDC) can be derived, allowing integrated assessment of axonal microstructural integrity and macrostructural bundle morphology. Due to its improved anatomical accuracy and biological interpretability in regions of fiber crossing, CSD has become widely used for diffusion MRI–based tractography and quantitative white matter pathway analysis. It is widely used in multiple sclerosis and neurodegenerative disease research,61,62 offering key insights into white matter network reorganization and axonal pathology. Nevertheless, its application in chronic pain remains relatively limited. Robertson et al applied fixel-based analysis (FBA) derived from CSD to investigate white matter microstructural alterations in patients with chronic back pain (CBP). They reported significantly reduced FD and FDC in multiple white matter tracts, including the right anterior thalamic radiation, the bilateral spinothalamic tracts, and the body and splenium of the corpus callosum, whereas FD was significantly increased in the genu of the corpus callosum. Importantly, these microstructural metrics were significantly negatively correlated with scores on the Pain Catastrophizing Scale (PCS), indicating a close association between white matter abnormalities and clinical symptom severity.63 Likewise, using a CSD-based fixel-based analysis (FBA) framework, Mills et al studied white matter in neuropathic pain. They observed reduced FD and FDC in the corona radiata and thalamus–primary somatosensory cortex pathways, regions critical for ascending nociception. Notably, these changes were primarily seen in female patients. Women also showed decreased FD in the cingulum bundle (linked to the default mode network) alongside increased FDC in the anterior insula–temporoparietal junction and thalamus–posterior insula pathways. These findings suggest more pronounced white matter alterations in sensory–emotional–cognitive networks in women with neuropathic pain, pointing to sex-specific differences in central pain processing and vulnerability.64 Kurokawa et al employed a multi-shell, multi-tissue constrained spherical deconvolution (MSMT-CSD) framework, in combination with anatomically constrained tractography (ACT) and spherical deconvolution informed filtering of tractograms (SIFT), to investigate the structural connectome of patients with burning mouth syndrome (BMS). Their results demonstrated significantly increased betweenness centrality among key regions of the pain matrix, including the insula, amygdala, and orbitofrontal cortex, alongside a reduction in betweenness centrality in the right inferior temporal cortex. These network-level alterations were interpreted as reflecting a reorganization of emotion–affective pain pathways.65 By employing MSMT-CSD, this work improved the biological plausibility of connectome reconstruction and identified structural differences within central affective-regulatory networks in BMS. Overall, studies employing CSD and its advanced variants have elucidated white matter microstructural plasticity across perceptual, emotional, and cognitive networks in patients with chronic pain from multiple perspectives. By improving the representation of crossing-fiber architecture, CSD-based models provide complementary, model-dependent measures for characterizing white matter alterations associated with chronic pain. These approaches represent a shift from single-tensor modeling toward higher-angular-resolution and, increasingly, multi-shell analytical frameworks. Several limitations should be considered when interpreting the CSD studies reviewed above. The evidence remains limited to a small number of single-center studies with modest sample sizes and no independent external validation. Fixel-based analysis involves a multi-stage pipeline—including response-function estimation, population-template construction, fixel correspondence, and connectivity-based smoothing—and analytical decisions at each stage may propagate to the resulting metrics. Cross-site reproducibility has yet to be established in pain cohorts. FD and FC are also acquisition- and processing-dependent: FD is influenced by the b-value scheme and angular resolution, whereas FC is estimated from nonlinear spatial transformations and is therefore sensitive to template construction and registration quality. Graph-theoretical measures, including betweenness centrality, vary with node parcellation, edge definition, and thresholding, potentially altering the inferred network topology. These measures should therefore be interpreted as model- and pipeline-dependent properties of reconstructed tractograms rather than as direct measures of axonal architecture. Applications and Findings of Multimodal Magnetic Resonance Imaging Chronic pain is a complex condition that arises from dynamic interactions among sensory, affective, cognitive, and modulatory networks.66–68 Given this complexity, no single neuroimaging modality is sufficient to comprehensively capture the multifaceted mechanisms underlying chronic pain. A central advantage of multimodal neuroimaging lies in the integration of complementary physiological sensitivities and spatiotemporal resolutions afforded by different imaging techniques.69 Consequently, approaches that combine diffusion MRI with other modalities—such as structural MRI, resting-state or task-based fMRI, and even genetics—are increasingly used to investigate the system-level mechanisms of chronic pain. The term “multimodal imaging” is often used in a narrow sense to describe the integration of data acquired using different imaging instruments. To enable truly simultaneous data acquisition, dedicated hardware configurations must be developed to ensure that signals from one modality can be obtained with minimal interference from, or with interference that can be effectively mitigated by, the other modality.69 For example, Vachon-Presseau et al conducted a 3-year longitudinal study in patients with subacute back pain (SBP), integrating multimodal neuroimaging—including structural MRI (T1-weighted imaging), diffusion tensor imaging (DTI), and functional MRI (fMRI)—together with genetic and behavioral data. Using path analysis, they demonstrated that greater white matter connectivity density within the dorsomedial prefrontal cortex–amygdala–nucleus accumbens circuit, along with smaller amygdala volume, constituted independent risk factors for pain chronification, collectively accounting for approximately 60% of the variance in pain persistence. Furthermore, they identified OPRD1 gene polymorphisms as a candidate genetic mediator of corticolimbic-based vulnerability to chronic pain.70 By integrating neuroimaging, genetic, and behavioral measures, this study suggested that intrinsic neuroanatomical and genetic characteristics of the corticolimbic system contribute to the transition from acute to chronic pain, beyond classical nociceptive processing pathways or initial pain intensity. These findings highlight novel neurobiological targets for the early prediction and prevention of pain chronification. In a wider sense, multimodal imaging also includes the combination of non-redundant data (ie, contrasts) acquired with the same instrument.69 For example, Li et al performed a multimodal study in patients with migraine by combining NODDI and DTI with structural (T1-weighted) MRI. They examined white matter microstructure using tract-based spatial statistics (TBSS) and cortical gray matter using surface-based analysis (SBA). Using the JuSpace toolbox, the authors then conducted a cross-modal spatial correlation analysis between cortical neurite damage patterns in patients and neurotransmitter distribution maps derived from an independent PET database of healthy individuals. Results showed widespread neurite damage in migraine, where both axonal injury and demyelination worsened with disease chronicity. Furthermore, reductions in cortical neurite density were significantly correlated with the spatial distributions of specific neurotransmitter systems, including dopamine D2 receptors and serotonin transporters.71 Collectively, this study demonstrates that patterns of neurite damage in migraine are spatially correlated with the distributions of specific neurotransmitter systems, which the authors interpreted as evidence of structure–chemistry correspondence. Bishop et al employed a multimodal MRI approach combining structural MRI (T1-weighted imaging) with multiple diffusion models, including the single-tensor diffusion tensor imaging (DTI) model and the ball-and-sticks (BAS) model, to investigate chronic musculoskeletal pain (CMP). Their analysis revealed that patients showed not only gray matter alterations, with increased subcortical nucleus volumes and reduced sensorimotor and prefrontal cortical volumes, but also decreased white matter complexity and fiber density (FD). Notably, the authors observed a dissociation in which FD was reduced while fiber cross-section (FC) remained unchanged. This pattern suggests that non-neuronal mechanisms such as gliosis may contribute to white matter changes, whereas altered white matter complexity reflects neuronal-level structural remodeling.72 Collectively, this multiparametric approach yields more comprehensive pathophysiological insights than reliance on a single metric such as FA. Moreover, most studies suggest that patients with chronic pain exhibit both functional alterations and accompanying structural changes in the brain.73,74 Integrating diffusion with functional neuroimaging therefore offers a powerful framework for probing structure–function relationships, enabling a more comprehensive understanding of pain-related functional deficits and the pathophysiology of chronic pain. In a rat model of chronic neuropathic pain, Zhou et al combined DTI and resting-state fMRI (rs-fMRI) with behavioral tests and immunofluorescence to validate imaging findings. The results revealed clear sex differences. Males showed microstructural damage in hippocampal CA1/CA2 subfields, reflected by increased mean diffusivity (MD) and radial diffusivity (RD). In contrast, females displayed enhanced functional activity in the dentate gyrus (DG), along with decreased small-world properties and increased global efficiency of hippocampal functional networks. By integrating DTI and rs-fMRI metrics, the study uncovered a sexually dimorphic pattern of hippocampal neuroplasticity in chronic pain, indicating that the hippocampus may contribute differently to pain processing in males and females.75 Multimodal connectomic analyses have also been used to inform target selection for neuromodulation. In healthy participants, Kong et al combined functional MRI and DTI to identify cortical regions with convergent functional and structural connectivity to the hippocampus, amygdala, and nucleus accumbens, and found that the medial prefrontal cortex displays bimodal connectivity with all three. The authors proposed such regions as candidate targets for non-invasive brain stimulation aimed at indirectly modulating deep brain structures.76 This approach has not yet been tested in patients with chronic pain, and its therapeutic relevance therefore remains to be established. Collectively, these multimodal studies are consistent with a distributed-network account of chronic pain and link observations across several levels of analysis, from genetic variation and brain microstructure to functional organization and behavior. However, the evidence base remains small, methodologically heterogeneous, and predominantly cross-sectional. Among the studies reviewed above, only one used a longitudinal design. Its path model explained approximately 60% of the variance in pain persistence within the original sample, but the model was not evaluated in an independent cohort; the reported explanatory performance may therefore be optimistic. Multimodal designs also generate many measures relative to the number of participants, increasing the susceptibility of path-analytic and predictive models to overfitting in the absence of external validation. In addition, modality-specific preprocessing choices may compound across analytical pipelines. Cross-modal spatial correlations with independent reference maps demonstrate spatial correspondence but do not establish mechanistic coupling. Finally, one study was conducted in a rodent model, and its findings cannot be assumed to generalize directly to patients. Summary and Future Perspectives This review outlines advances in diffusion MRI (dMRI) for chronic pain research. The field has evolved from early single-shell models like DTI to advanced multi-shell models—such as DKI, NODDI, and CSD—and further to their integration with multimodal neuroimaging. These developments have broadened the characterization of microstructural and network-level alterations associated with chronic pain. To summarize key dMRI findings across clinical subtypes, a synthesis is provided in Table 2.Table 2Representative Diffusion MRI Findings Across Pain TypesPain Type (Mechanistic Tendency)Primary Affected Structures/Networks (Examples)Diffusion Models and Clinical Associations (Examples)Key Limitations (Examples)Chronic back pain (pain plasticity/ affective dimension)Anterior thalamic radiation, spinothalamic tract, corpus callosum, among others (not restricted to a single pathway)FBA: widespread reductions in fiber density (FD) and fiber density and cross-section (FDC), negatively correlated with pain catastrophizing scale (PCS)Predominantly cross-sectional; need for standardized preprocessing; limited cross-cohort reproducibilitySubacute back pain (susceptibility/ resilience)Superior longitudinal fasciculus, internal capsule, and other pathways related to sensory–cognitive regulationDTI: baseline FA differences predict pain persistence at 1-year follow-up; multicenter studies link right SLF FA to recovery/resilienceRelatively stronger longitudinal and multicenter evidence; individual-level stability and incremental clinical value remain to be establishedClassical / idiopathic trigeminal neuralgia (neuropathic; peripheral–central coexistence)Corona radiata, internal capsule, thalamic white matter, gray nuclei (eg, thalamus); peripheral/brainstem segments (pontine trigeminal nerve, SpTV)DTI/TBSS show widespread FA reductions; peripheral nerve DTI combined with clinical variables predicts surgical response/recurrence; DKI and NODDI reveal non-Gaussian diffusion and neurite abnormalitiesHigh methodological heterogeneity; peripheral nerve segmentation, EPI distortion, and small-ROI stability remain major technical challengesPostherpetic neuralgia (neuropathic)Insula, thalamus, caudate nucleus, and other pain processing/modulatory regionsDKI: reductions in K∥, K⊥, and MK; insular metrics correlate with VAS scoresSmall sample sizes; predominantly cross-sectional; longitudinal validation of treatment sensitivity is neededPainful chronic pancreatitis (chronic secondary visceral pain; central sensitization)Amygdala, cingulate cortex, insula, prefrontal cortex, secondary somatosensory cortex, corpus callosum, thalamic radiations, and corona radiataDTI: altered ADC and FA in visceral pain-related regions, with selected measures associated with pain scores.Small baseline samples and substantial longitudinal attrition; smoking, disease duration, and other systemic factors may confound diffusion changes; abnormalities cannot be attributed to pain aloneCancer-related pain in lung cancer (Chronic cancer-related pain)Right hippocampal cingulum, inferior longitudinal fasciculus, and other pain-related white matter tractsDTI: whole-tract and node-specific diffusion abnormalities; selected diffusion measures associated with pain intensity or durationSingle-center cross-sectional evidence; heterogeneity in cancer stage, metastasis, treatment, systemic inflammation, and medication may confound diffusion changes; no independent or longitudinal validationCervicogenic headache (peripheral nerve–related)Greater occipital nerve and C2/C3 dorsal root ganglia (DRG)Peripheral nerve DTI: reduced FA and increased ADC on the symptomatic sideHigh demands on spatial resolution, coils, and motion control; standardized protocols and larger samples requiredChronic prostatitis/ chronic pelvic pain syndrome (pain plasticity and cognitive/affective regulation)Abnormal structural connectivity efficiency within the frontoparietal control networkDTI connectomics with graph theory: network topological alterations correlate with symptom severityCautious interpretation of network metrics is warranted; independent replication and transparent reporting of fusion pipelines are neededNew daily persistent headache (primary headache spectrum)Long-range fiber tracts (eg, superior/inferior longitudinal fasciculi, inferior fronto-occipital fasciculus) and frontal/orbitofrontal/insular gray matterNODDI: reduced neurite density index (NDI) in white matter and cortex, involving pain–affective–cognitive networksRare condition, mostly single-center studies; stratified control of disease duration, medication, and comorbidities is requiredChronic post-traumatic central pain after mild traumatic brain injurySpinothalamocortical tractDTT: lower FA and tract volume and higher MD of the spinothalamocortical tract in patients with central pain than in painless mTBI patients and healthy controls; no associations with clinical measures were examined.Retrospective design; tract abnormalities do not establish causal injury; independent longitudinal validation is requiredAdolescent chronic non-disease-related pain (developmental stage; prominent catastrophizing)Corona radiata, CST, superior longitudinal fasciculus, uncinate fasciculus, corpus callosum; along-tract analyses reveal segment-specific effectsNODDI: globally increased ODI; elevated NDI in selected tract segments; multiple metrics correlate with pain catastrophizingNeed to disentangle developmental versus pain-related effects; longitudinal follow-up required to test vulnerability markersNotes: This table is an exemplary summary intended to illustrate the application of different diffusion models across various chronic pain types, rather than a systematic search or meta-analysis.Abbreviations: DTI, diffusion tensor imaging; FA, fractional anisotropy; DKI, diffusion kurtosis imaging; MK, mean kurtosis; K∥, axial kurtosis; K⊥, radial kurtosis; NODDI, neurite orientation dispersion and density imaging; NDI, neurite density index; ODI, orientation dispersion index; ADC, apparent diffusion coefficient; EPI, echo-planar imaging; FD, fiber density; FDC, fiber density and cross-section; SpTV, spinal trigeminal tract; FBA, fixel-based analysis; SLF, superior longitudinal fasciculus; TBSS, tract-based spatial statistics; CST, corticospinal tract; PCS, pain catastrophizing scale; ROI, region of interest; VAS, visual analog scale; DRG, dorsal root ganglia; DTT, diffusion tensor tractography; MD, mean diffusivity; mTBI, mild traumatic brain injury. Table 2Representative Diffusion MRI Findings Across Pain TypesPain Type (Mechanistic Tendency)Primary Affected Structures/Networks (Examples)Diffusion Models and Clinical Associations (Examples)Key Limitations (Examples)Chronic back pain (pain plasticity/ affective dimension)Anterior thalamic radiation, spinothalamic tract, corpus callosum, among others (not restricted to a single pathway)FBA: widespread reductions in fiber density (FD) and fiber density and cross-section (FDC), negatively correlated with pain catastrophizing scale (PCS)Predominantly cross-sectional; need for standardized preprocessing; limited cross-cohort reproducibilitySubacute back pain (susceptibility/ resilience)Superior longitudinal fasciculus, internal capsule, and other pathways related to sensory–cognitive regulationDTI: baseline FA differences predict pain persistence at 1-year follow-up; multicenter studies link right SLF FA to recovery/resilienceRelatively stronger longitudinal and multicenter evidence; individual-level stability and incremental clinical value remain to be establishedClassical / idiopathic trigeminal neuralgia (neuropathic; peripheral–central coexistence)Corona radiata, internal capsule, thalamic white matter, gray nuclei (eg, thalamus); peripheral/brainstem segments (pontine trigeminal nerve, SpTV)DTI/TBSS show widespread FA reductions; peripheral nerve DTI combined with clinical variables predicts surgical response/recurrence; DKI and NODDI reveal non-Gaussian diffusion and neurite abnormalitiesHigh methodological heterogeneity; peripheral nerve segmentation, EPI distortion, and small-ROI stability remain major technical challengesPostherpetic neuralgia (neuropathic)Insula, thalamus, caudate nucleus, and other pain processing/modulatory regionsDKI: reductions in K∥, K⊥, and MK; insular metrics correlate with VAS scoresSmall sample sizes; predominantly cross-sectional; longitudinal validation of treatment sensitivity is neededPainful chronic pancreatitis (chronic secondary visceral pain; central sensitization)Amygdala, cingulate cortex, insula, prefrontal cortex, secondary somatosensory cortex, corpus callosum, thalamic radiations, and corona radiataDTI: altered ADC and FA in visceral pain-related regions, with selected measures associated with pain scores.Small baseline samples and substantial longitudinal attrition; smoking, disease duration, and other systemic factors may confound diffusion changes; abnormalities cannot be attributed to pain aloneCancer-related pain in lung cancer (Chronic cancer-related pain)Right hippocampal cingulum, inferior longitudinal fasciculus, and other pain-related white matter tractsDTI: whole-tract and node-specific diffusion abnormalities; selected diffusion measures associated with pain intensity or durationSingle-center cross-sectional evidence; heterogeneity in cancer stage, metastasis, treatment, systemic inflammation, and medication may confound diffusion changes; no independent or longitudinal validationCervicogenic headache (peripheral nerve–related)Greater occipital nerve and C2/C3 dorsal root ganglia (DRG)Peripheral nerve DTI: reduced FA and increased ADC on the symptomatic sideHigh demands on spatial resolution, coils, and motion control; standardized protocols and larger samples requiredChronic prostatitis/ chronic pelvic pain syndrome (pain plasticity and cognitive/affective regulation)Abnormal structural connectivity efficiency within the frontoparietal control networkDTI connectomics with graph theory: network topological alterations correlate with symptom severityCautious interpretation of network metrics is warranted; independent replication and transparent reporting of fusion pipelines are neededNew daily persistent headache (primary headache spectrum)Long-range fiber tracts (eg, superior/inferior longitudinal fasciculi, inferior fronto-occipital fasciculus) and frontal/orbitofrontal/insular gray matterNODDI: reduced neurite density index (NDI) in white matter and cortex, involving pain–affective–cognitive networksRare condition, mostly single-center studies; stratified control of disease duration, medication, and comorbidities is requiredChronic post-traumatic central pain after mild traumatic brain injurySpinothalamocortical tractDTT: lower FA and tract volume and higher MD of the spinothalamocortical tract in patients with central pain than in painless mTBI patients and healthy controls; no associations with clinical measures were examined.Retrospective design; tract abnormalities do not establish causal injury; independent longitudinal validation is requiredAdolescent chronic non-disease-related pain (developmental stage; prominent catastrophizing)Corona radiata, CST, superior longitudinal fasciculus, uncinate fasciculus, corpus callosum; along-tract analyses reveal segment-specific effectsNODDI: globally increased ODI; elevated NDI in selected tract segments; multiple metrics correlate with pain catastrophizingNeed to disentangle developmental versus pain-related effects; longitudinal follow-up required to test vulnerability markersNotes: This table is an exemplary summary intended to illustrate the application of different diffusion models across various chronic pain types, rather than a systematic search or meta-analysis.Abbreviations: DTI, diffusion tensor imaging; FA, fractional anisotropy; DKI, diffusion kurtosis imaging; MK, mean kurtosis; K∥, axial kurtosis; K⊥, radial kurtosis; NODDI, neurite orientation dispersion and density imaging; NDI, neurite density index; ODI, orientation dispersion index; ADC, apparent diffusion coefficient; EPI, echo-planar imaging; FD, fiber density; FDC, fiber density and cross-section; SpTV, spinal trigeminal tract; FBA, fixel-based analysis; SLF, superior longitudinal fasciculus; TBSS, tract-based spatial statistics; CST, corticospinal tract; PCS, pain catastrophizing scale; ROI, region of interest; VAS, visual analog scale; DRG, dorsal root ganglia; DTT, diffusion tensor tractography; MD, mean diffusivity; mTBI, mild traumatic brain injury. Representative Diffusion MRI Findings Across Pain Types Notes: This table is an exemplary summary intended to illustrate the application of different diffusion models across various chronic pain types, rather than a systematic search or meta-analysis. Abbreviations: DTI, diffusion tensor imaging; FA, fractional anisotropy; DKI, diffusion kurtosis imaging; MK, mean kurtosis; K∥, axial kurtosis; K⊥, radial kurtosis; NODDI, neurite orientation dispersion and density imaging; NDI, neurite density index; ODI, orientation dispersion index; ADC, apparent diffusion coefficient; EPI, echo-planar imaging; FD, fiber density; FDC, fiber density and cross-section; SpTV, spinal trigeminal tract; FBA, fixel-based analysis; SLF, superior longitudinal fasciculus; TBSS, tract-based spatial statistics; CST, corticospinal tract; PCS, pain catastrophizing scale; ROI, region of interest; VAS, visual analog scale; DRG, dorsal root ganglia; DTT, diffusion tensor tractography; MD, mean diffusivity; mTBI, mild traumatic brain injury. Summary First, accumulating evidence indicates that different kinds of chronic pain are associated with alterations in white matter microstructure and gray matter across multiple brain networks. These changes involve not only ascending nociceptive pathways, such as the spinothalamic tract and thalamic radiations, but also regulatory circuits including the cortico–limbic system, the default mode network, the salience network, and executive control networks. Collectively, these findings suggest that chronic pain is more accurately characterized as a whole-brain network disorder, reflecting a system-level phenotype rather than a focal abnormality confined to a single pathway. Second, diffusion MRI–derived measures, including FA, RD, MK, and NDI, have been repeatedly reported to correlate with pain intensity, pain catastrophizing, affective and cognitive comorbidities, and treatment response. A smaller number of longitudinal studies further suggest that baseline white matter microstructural characteristics may predict pain chronification and subsequent clinical outcomes. Together, these findings indicate that diffusion MRI metrics may reflect both vulnerability to chronic pain and disease-related changes, supporting their potential utility as neuroimaging biomarkers. However, their reliability and generalizability at the individual level remain to be established through more rigorous external validation. Third, from a methodological perspective, DTI provided early studies with reproducible estimates of white matter integrity. However, its dependence on Gaussian diffusion assumptions, along with sensitivity to fiber crossing and partial volume effects, limits its specificity for underlying pathology. By contrast, multi-shell acquisitions and advanced diffusion models enable the characterization of non-Gaussian diffusion, neurite density and orientation dispersion, and fiber morphometric properties, thereby offering a more detailed representation of microstructural complexity. These approaches provide potentially more interpretable quantitative indices that may help distinguish changes related to axonal or myelin integrity from those associated with increased free-water content. Finally, multimodal integration studies have promoted a conceptual shift in the field, moving from isolated descriptions of structural or functional abnormalities toward integrative frameworks that link genes, structural connectivity, functional networks, and behavior or symptoms. Within this framework, cortico–limbic circuits may play a central role in pain chronification and affective dysregulation. A promising direction is the development of testable mechanistic models in which structural connectivity constrains the emergence of functional network alterations, and imaging phenotypes are linked to actionable intervention targets, thereby improving causal inference and clinical relevance. Current Challenges and Future Directions Firstly, most existing studies rely on cross-sectional designs, which limits causal inference regarding the direction of microstructural abnormalities. Future work should prioritize large-scale, prospective longitudinal cohorts with repeated measurements, combined with causal modeling approaches (eg, cross-lagged models) and pre–post intervention imaging (eg, pharmacological, surgical, or neuromodulatory treatments). Such designs are essential to distinguish vulnerability markers from reversible plastic changes and to clarify which diffusion metrics are suitable for risk stratification versus treatment monitoring. Secondly, diffusion MRI metrics are sensitive to acquisition parameters, preprocessing pipelines, head motion, and cross-site variability. Additionally, conventional DTI remains limited in regions with complex fiber architecture. Where feasible, future studies should adopt standardized multi-shell acquisition protocols with rigorous quality control, increasingly employ higher-order models such as NODDI and DKI, and extend imaging beyond intracranial white matter to encompass the brainstem, spinal cord, and peripheral nervous system. Investigating the entire pain neuroaxis as an integrated unit is crucial for deriving reproducible and comparable structural phenotypes. Thirdly, given the marked heterogeneity of chronic pain, single diagnostic categories rarely correspond to unique imaging patterns. Integrating mechanism-based pain classifications, quantitative sensory testing, psychosocial and behavioral factors, and medication profiles within high-dimensional frameworks may enable the identification of shared transdiagnostic dimensions and disease-specific connectomic fingerprints—unique whole-brain connectivity patterns characteristic of individual patients or subgroups. This approach can advance objective, mechanism-oriented stratification. While multimodal integration improves predictive potential, it also introduces risks of overfitting, bias, and reduced interpretability. Therefore, robust external validation, feature stability assessment, and explainable modeling approaches are critical. Integrative frameworks that jointly model structural connectivity, functional networks, and clinical measures may further support cross-level inferences, moving beyond correlation toward testable mechanistic hypotheses. Finally, clinical translation depends on the development of actionable workflows, including clearly defined acquisition requirements, reliability metrics, sensitivity to treatment-related change, and demonstrable clinical utility beyond existing scales. Individualized prediction models informed by structural connectivity may ultimately support risk stratification, treatment selection, and target identification, with validation in real-world cohorts and clinical trials. In summary, diffusion MRI is moving from descriptive analyses toward quantitative frameworks that can be linked to mechanism. One direction worth exploring, though at present speculative, is whether diffusion measures might be organized along three complementary axes—vulnerability, current disease state, and treatment responsiveness—each requiring its own validation strategy: prospective cohorts sampled before pain onset for vulnerability, cross-sectional and test–retest designs for disease state, and interventional studies with pre–post imaging for treatment responsiveness. Whether such a framework proves tractable will depend on standardized multi-shell acquisition, replication across centers, and demonstration of incremental value over existing clinical assessment. At present, diffusion MRI remains a research tool in chronic pain rather than an established component of clinical care.