Neuroimaging in Addiction: Principles of Functional Near-infrared Spectroscopy and Its Emerging Role in De-addiction: A Narrative Review

Abstract

Background

Substance use disorders (SUDs) remain a major global health challenge with high relapse rates. Functional near-infrared spectroscopy (fNIRS), a non-invasive, portable modality enabling real-time monitoring of cortical haemodynamic changes, is well suited to studying addiction neurobiology in ecologically valid settings. This review examines fNIRS principles and applications in addiction research and de-addiction therapy.

Summary

PubMed, Scopus and Web of Science were searched (2000–2025) using terms including fNIRS, addiction, substance use disorder, prefrontal cortex, neurofeedback and de-addiction. fNIRS reliably detects altered prefrontal cortex (PFC) haemodynamic responses during cue-reactivity and cognitive tasks in addiction. Findings include reduced dorsolateral PFC activation with prolonged abstinence, differential orbitofrontal cortex activation across substances, and use as a treatment-response biomarker. Emerging applications include neurofeedback and machine-learning-based classification. fNIRS offers portability and ecological validity over conventional neuroimaging. Future research should standardise protocols, validate biomarkers in multicentre trials, and integrate fNIRS with EEG and fMRI.

Key Message

fNIRS is a portable, non-invasive tool that reliably detects altered prefrontal cortex activation in substance use disorders. Cue-reactivity and cognitive paradigms identify biomarkers of craving, executive dysfunction and relapse risk. fNIRS-based neurofeedback shows early promise for improving self-regulation and reducing relapse. Wearable fNIRS with machine learning and multimodal integration enables personalised addiction monitoring.

Introduction

Substance use disorders (SUDs) constitute a major global public health burden, affecting approximately 35 million individuals worldwide and contributing to 12.4% of all deaths according to the World Health Organization. ¹ Despite advances in pharmacological and psychosocial interventions, relapse rates remain high, with an estimated 40%–60% of individuals returning to substance use within the first year of treatment. ² A fundamental barrier to improving treatment outcomes lies in the limited understanding of the dynamic neural mechanisms that drive craving, impaired executive control and relapse vulnerability in real-world settings.

The human prefrontal cortex (PFC) plays a central role in the neurobiology of addiction. It is critically involved in executive functions, including attention, decision-making, impulse control and emotion regulation—processes that are consistently disrupted in individuals with SUDs. ³ Neuroimaging studies have demonstrated that the ventrolateral and dorsolateral PFC, anterior cingulate cortex, orbitofrontal cortex (OFC), amygdala and striatum constitute a core neural circuitry whose dysfunction underlies the transition from voluntary substance use to compulsive drug-seeking behaviour. ⁴ The prefrontal cortical regions, in particular, mediate top-down regulatory control over limbic reward circuits, and their dysfunction is central to the hallmark features of addiction: heightened cue reactivity, impaired inhibitory control and diminished sensitivity to natural rewards. ⁵

Functional magnetic resonance imaging (fMRI) has been the gold standard for studying these neural processes, providing excellent spatial resolution and whole-brain coverage. However, fMRI presents significant practical limitations for addiction research, including high cost, restricted participant movement, claustrophobic scanning environments and the inability to monitor neural activity in naturalistic or clinical settings. ⁶ These constraints have driven the search for complementary neuroimaging approaches that can bridge the gap between laboratory findings and real-world clinical application.

Functional near-infrared spectroscopy (fNIRS) has emerged as a compelling alternative. First developed in the early 1990s and extensively refined over the past two decades, fNIRS is a non-invasive optical neuroimaging technique that measures changes in cortical haemoglobin concentrations by detecting the differential absorption of near-infrared (NIR) light by oxygenated haemoglobin (HbO) and deoxygenated haemoglobin (HbR).^{7, 8} Its unique combination of portability, safety (no ionising radiation), relatively low cost, tolerance to motion artefacts and capacity for real-time monitoring makes it ideally suited for longitudinal and ecological studies of brain function in addiction populations. ⁹

Several pioneering studies have demonstrated the potential of fNIRS in addiction research. Dresler et al. ¹⁰ used fNIRS to document the progressive recovery of frontal brain activity during verbal fluency tasks across different phases of alcohol withdrawal, from active dependence through abstinence to normalisation. Bunce et al. ¹¹ demonstrated that current alcohol users exhibited heightened PFC responses to alcohol-related cues compared to individuals in sustained recovery and healthy controls, suggesting that fNIRS-derived haemodynamic signatures could serve as biomarkers for addiction severity and relapse vulnerability. More recently, Gu et al. ¹² applied machine learning algorithms to fNIRS-measured PFC activation patterns to successfully classify individuals by drug type, highlighting the potential for objective, neuroimaging-based assessment of SUDs.

To verify this gap, we ran a targeted scoping search on 15 March 2026 across PubMed, Scopus and Web of Science, combining the terms ‘functional near-infrared spectroscopy’ OR ‘fNIRS’ AND ‘addiction’ OR ‘substance use disorder’ OR ‘de-addiction’ AND ‘review’ OR ‘systematic review’ OR ‘narrative review’. Of 87 records returned, 14 review articles were screened in full. Most addressed fNIRS in cognitive neuroscience or psychiatry broadly, with addiction featuring as only one of several application domains. The closest match was the narrative review by Cataldo et al., ⁹ which surveyed fNIRS findings in addiction but did not cover technical principles, neurofeedback applications or a structured comparison with fMRI in any depth. No prior work integrated all five components addressed here, which supports the rationale for the present review.

Despite this growing body of evidence, no comprehensive review has systematically examined both the technical principles and the expanding clinical applications of fNIRS specifically in the context of addiction and de-addiction. This narrative review addresses this gap by (a) elucidating the biophysical and engineering principles underlying fNIRS technology; (b) synthesising current evidence on fNIRS applications in alcohol and substance addiction research; (c) evaluating the emerging role of fNIRS-based neurofeedback (fNIRS-NF) in de-addiction therapy; (d) providing a comparative analysis of fNIRS and fMRI in addiction neuroscience and (e) identifying critical knowledge gaps and future research directions. Objectives (a)–(e) are addressed in Sections ‘Discussion: Principles and Technical Mechanisms of fNIRS’ to ‘Comparative Analysis of fNIRS and fMRI in Addiction Research’ of this review, respectively.

Discussion: Principles and Technical Mechanisms of Functional Near-infrared Spectroscopy

Biophysical Basis

fNIRS operates on the principle that biological tissue, including the scalp, skull and cerebral cortex, is relatively transparent to NIR light within the optical window of approximately 650–900 nm. Within this spectral range, the primary chromophores—HbO and HbR—exhibit distinct absorption spectra, enabling their independent quantification.^{7, 13} When neural populations become active, local metabolic demand triggers a haemodynamic response characterised by increased regional cerebral blood flow and blood volume, resulting in a rise in HbO concentration and a corresponding decrease in HbR. This neurovascular coupling mechanism forms the physiological basis for fNIRS signal interpretation and is analogous to the blood-oxygen-level-dependent (BOLD) contrast exploited by fMRI. ¹⁴

Light Source, Detection and Optical Pathway

fNIRS systems employ light-emitting diodes (LEDs) or laser diodes as light sources that emit NIR light at two or more wavelengths, typically around 760 and 850 nm, to independently measure HbO and HbR concentrations. ¹⁵ The emitted light penetrates the scalp and skull, traversing the superficial layers of the cerebral cortex along a characteristic banana-shaped (arc-like) photon trajectory before being captured by photodetectors—typically silicon photodiodes or avalanche photodiodes—positioned at a defined source–detector separation on the scalp surface. ¹⁶ The source–detector separation, typically 2.5–3.0 cm, determines the sampling depth, with the midpoint of the source–detector distance corresponding approximately to the maximum depth of light penetration, which reaches approximately 1.5–2.0 cm into cortical tissue. ⁷

Modified Beer–Lambert Law

The quantitative relationship between light attenuation and chromophore concentration is governed by the modified Beer–Lambert Law (mBLL), which accounts for both absorption and scattering of photons as they traverse biological tissue. ¹⁷ The mBLL expresses changes in optical density as a function of the extinction coefficients of HbO and HbR, the path length of light through tissue (corrected by a differential path length factor) and the source–detector separation. By measuring light attenuation at two or more wavelengths, the system of equations can be solved simultaneously to derive changes in HbO and HbR concentrations relative to baseline, providing a non-invasive, real-time index of cortical haemodynamic activity. ¹⁵

Hardware Configurations and Signal Detection

Modern fNIRS systems are available in several configurations, ranging from benchtop multichannel systems to wearable wireless devices. Continuous-wave (CW) systems, which measure changes in light intensity at constant output, are the most widely deployed due to their simplicity and cost-effectiveness.^{15, 16} Frequency-domain (FD) systems modulate light intensity at radio frequencies and measure both amplitude attenuation and phase shift, enabling absolute quantification of optical properties. Time-domain (TD) systems emit short pulses of light and measure the temporal distribution of photon arrival times, providing the most complete characterisation of tissue optical properties but at greater complexity and cost. ¹⁸ Modern multichannel systems feature arrays of source–detector pairs arranged in grids or montages that enable simultaneous monitoring of multiple cortical regions, providing topographic maps of brain activation. Recent hardware innovations have produced wearable, wireless fNIRS headsets that are lightweight and suitable for data collection in naturalistic environments, a critical advantage for addiction research involving real-world cue exposure. ¹⁹

Signal Processing and Artefact Management

Raw fNIRS optical signals require systematic preprocessing to ensure data quality. Standard processing pipelines include band-pass filtering (typically 0.01–0.2 Hz) to isolate the haemodynamic response while attenuating physiological noise from cardiac pulsations (~1 Hz), respiration (~0.2–0.4 Hz) and Mayer waves (~0.1 Hz); motion artefact correction using techniques such as wavelet filtering, spline interpolation or correlation-based signal improvement and baseline correction to account for interindividual variability in resting-state haemodynamics. ²⁰ Advanced analytical approaches, including general linear model (GLM), independent component analysis (ICA) and machine learning algorithms, are increasingly employed to extract task-related activation patterns, enhance signal-to-noise ratios and improve the interpretability of fNIRS measurements. ²¹

Temporal and Spatial Resolution

fNIRS offers excellent temporal resolution, with sampling rates of up to 10–100 Hz in modern systems, making it well-suited for tracking rapid haemodynamic changes. ⁷ However, the inherent diffuse scattering of NIR light in biological tissue limits the spatial resolution to approximately 1–3 cm, which is substantially lower than fMRI (1–3 mm). Furthermore, fNIRS is restricted to monitoring cortical regions within approximately 1.5–2.0 cm of the scalp surface and cannot image subcortical structures such as the amygdala, ventral striatum or deep limbic regions that are critically involved in addiction neurocircuitry. ²² Innovations such as time-resolved fNIRS, diffuse optical tomography (DOT) and high-density fNIRS montages are being developed to improve spatial resolution and extend depth sensitivity. ¹⁸

Limitations of Functional Near-infrared Spectroscopy

Despite its advantages, fNIRS has several important limitations that must be considered. Its spatial resolution is inferior to fMRI, and its depth sensitivity is restricted to cortical surfaces, precluding direct measurement of subcortical addiction-relevant structures. The signal is susceptible to contamination by systemic physiological changes (e.g., scalp blood flow changes) that do not reflect neural activity, a challenge partially addressed by short-separation channel regression techniques.^{22, 23} Hair density and skin pigmentation can affect signal quality, potentially introducing variability across participants. Additionally, the lack of standardised protocols for acquisition, preprocessing and analysis across laboratories limits the comparability and reproducibility of findings.

Functional Near-infrared Spectroscopy in Alcohol Addiction and De-addiction

Cue Reactivity and Neural Correlates of Craving

Cue-reactivity paradigms, in which individuals are exposed to substance-related stimuli while brain responses are measured, have been a cornerstone of fNIRS addiction research. Bunce et al. ¹¹ conducted a seminal study demonstrating that individuals with current alcohol dependence exhibited significantly heightened haemodynamic responses in the lateral PFC when exposed to alcohol-related visual cues compared to individuals in sustained recovery (90–180 days of abstinence) and healthy controls. The differential PFC activation between active users and recovering individuals suggests that fNIRS can capture the neural signature of craving intensity and may serve as an objective marker of addiction severity.

Dempsey et al. ²⁴ extended these findings by measuring PFC activation using fNIRS in individuals with alcohol addiction across varying durations of abstinence. Their results demonstrated a significant inverse relationship between the number of days of abstinence and dorsolateral and dorsomedial PFC activation in response to alcohol-related stimuli, suggesting that fNIRS can track the progressive attenuation of cue reactivity during recovery. This finding has important implications for treatment monitoring, as it provides a neurobiological index of recovery progress that complements self-report measures.

More recently, Huang et al. ²⁵ conducted a study of 123 alcohol-dependent patients and 149 healthy controls using fNIRS during a verbal fluency task. Their analysis identified specific fNIRS channels in Broca’s area and Wernicke’s area whose activation levels were significantly correlated with alcohol use parameters and could predict relapse within 3 months using logistic regression models. This study represents an important step toward the clinical translation of fNIRS as a relapse prediction tool.

Neural Recovery During Abstinence

Monitoring neural recovery during abstinence is critical for understanding the neuroplasticity of the addicted brain and for optimising treatment strategies. Dresler et al. ¹⁰ employed fNIRS to assess cortical functioning during verbal fluency tasks in abstinent alcohol-dependent patients at different phases of withdrawal. Their longitudinal findings revealed a progressive recovery of frontal brain oxygenation over time, from reduced activation during early withdrawal to normalisation during sustained abstinence. This landmark study provided direct evidence that alcohol-related neurotoxic effects on frontal cortical function are at least partially reversible, and that fNIRS possesses sufficient sensitivity to detect these temporal changes.

Schecklmann et al. ²⁶ investigated three patient groups—acute withdrawal, completed detoxification and stable moderation—during verbal fluency tasks and found generally reduced cerebral oxygenation levels in frontotemporal regions among recently withdrawn patients compared to those in later stages of recovery. These findings suggest that fNIRS-measured PFC function can serve as a longitudinal biomarker for tracking neurological recovery trajectories, informing decisions about treatment duration and intensity. These conclusions, however, rest on a narrow evidence base. Both of the studies summarised above used the same verbal fluency paradigm and recruited only individuals with alcohol use disorder. Broader evidence across different cognitive paradigms and other substance classes will be needed before fNIRS can be used to guide treatment decisions in routine clinical practice (Table 1).

Table 1. Summary of Included Studies on Functional Near-infrared Spectroscopy (fNIRS) Applications in Addiction Research.

Author/Year [Ref]	Study Design	Sample	Substance	fNIRS Method	Task/Paradigm	Key Findings
Bunce et al., 2006 11	Cross-sectional	Alcohol-dependent, recovered and healthy controls	Alcohol	CW-fNIRS; lateral PFC channels	Alcohol and natural reward cue-reactivity paradigm	Current alcohol users showed heightened lateral PFC response to alcohol cues versus recovered individuals and controls; fNIRS proposed as a biomarker for addiction severity.
Dresler et al., 2012 [10]	Cross-sectional (multi-phase)	Alcohol-dependent patients at different withdrawal phases; healthy controls	Alcohol	CW-fNIRS; prefrontal montage	Phonological and semantic verbal fluency task (VFT)	Progressive recovery of frontal brain oxygenation from active dependence through abstinence to normalisation; fNIRS detected neurotoxic effects and neural recovery.
Dempsey et al., 2015 [24]	Cross-sectional	Individuals with alcohol addiction across varying abstinence durations	Alcohol	CW-fNIRS; DLPFC and DMPFC channels	Alcohol-related image cue exposure	Increased days of abstinence associated with decreased DLPFC and DMPFC activation to alcohol stimuli; fNIRS tracks cue-reactivity attenuation during recovery.
Matsuo et al., 2003 [43]	Case-control	Alcohol-dependent patients (n = 12); healthy controls (n = 12)	Alcohol	Multichannel fNIRS; bilateral PFC	N-back working memory task	Alcohol-dependent patients showed altered haemodynamic responses in PFC during cognitive task; demonstrated PFC dysfunction in alcohol dependence.
Ieong and Yuan, 2019 [30]	Pilot; bimodal	Abstinent heroin-dependent (n = 8); healthy controls (n = 11)	Heroin/opiates	fNIRS + EEG; resting-state PFC	Resting-state functional connectivity (rsFC)	Desynchronised lower alpha rhythms, decreased PFC connectivity and asymmetric cerebrovascular injury in heroin patients; machine learning classified groups.
Huhn et al., 2019 [28]	Cross-sectional	Methadone-maintained patients	Cocaine/opioids	fNIRS; right DLPFC	Risky decision-making task (The Balloon Analogue Risk Task)	Increased right DLPFC neural activity during risky decisions associated with continued cocaine use in methadone-maintained patients.
Huhn et al., 2019 [29]	Prospective cohort	Methadone-maintained patients	Opioids/polysubstance	fNIRS; bilateral PFC	Drug cue-reactivity paradigm	PFC responses to drug cues, craving and depressive symptoms predicted treatment outcomes; fNIRS proposed for monitoring opioid substitution therapy.
Gu et al., 2021 [12]	Cross-sectional; ML classification	30 drug users (10 methamphetamine, 10 heroin, 10 mixed)	Methamphetamine, heroin, mixed	High-density fNIRS; prefrontal OFC	Resting + drug addiction induction paradigm	Machine learning (LDA, SVM) classified drug types based on OFC activation; methamphetamine users showed the highest OFC activation, heroin the lowest.
Cataldo et al., 2022 [9]	Narrative review	N/A (review of published studies)	Multiple substances and behavioural	Various fNIRS systems reviewed	Multiple paradigms reviewed	Comprehensive review of fNIRS applications in addiction; highlighted ecological validity and real-time monitoring advantages; identified research gaps.
Yang et al., 2024 [36]	Experimental (neurofeedback)	Healthy adults (proof-of-concept)	N/A (WM-related; applicable to SUD)	Real-time fNIRS-NF; left DLPFC	Working memory (n-back) + neurofeedback training	Brief fNIRS-NF training changed PFC activity and functional connectivity during subsequent WM challenge; proof-of-concept for fNIRS-NF neuromodulation.
Huang et al., 2025 [25]	Retrospective cohort	123 alcohol-dependent; 149 healthy controls	Alcohol	52-channel CW-fNIRS (Hitachi ETG-4000); 690/830 nm	Verbal fluency task (VFT)	fNIRS channels in Broca’s and Wernicke’s areas predicted 3-month relapse via logistic regression; age of drinking onset correlated with activation.
Fathi et al., 2025 [34]	Systematic review	32 studies (18 EEG-NF, 11 fMRI-NF, 3 fNIRS-NF)	Multiple substances	fNIRS-NF reviewed alongside EEG-NF and fMRI-NF	Various NF paradigms	Neurofeedback reduced drug craving; fNIRS-NF is promising but limited by small number of studies; called for standardised protocols.
Sanader Vukadinovic, 2025 [33]	Mini-review	N/A (review of recent NF studies)	Substance and behavioural addictions	fNIRS-NF reviewed	Various NF paradigms reviewed	fNIRS-NF identified as promising and scalable for addiction treatment; highlighted advantages of portability over fMRI-NF.
Liu et al., 2025 [42]	Systematic review	63 combined fMRI–fNIRS studies reviewed	Multiple (including addiction)	Combined fMRI–fNIRS integration	Multiple paradigms	fMRI–fNIRS integration provides synergistic spatiotemporal mapping; recommended for comprehensive brain activity studies including addiction.
Rahimpour Jounghani et al., 2025 [31]	Experimental	20 undergraduate students	Social media/digital addiction	Wearable CW-fNIRS; bilateral PFC	n-back, Go/No-Go + social media exposure	Social media exposure reduced EF task accuracy with decreased DLPFC/VLPFC activation; demonstrated fNIRS utility in behavioural addiction assessment.
Fazeli et al., 2020 [41]	Systematic review	N/A (review of fNIRS–SUD literature)	Multiple substances	Various fNIRS systems	Multiple paradigms	Reviewed fNIRS applications in SUD research; highlighted potential for real-world monitoring; identified need for larger sample sizes and standardised methods.
Seo et al., 2013 [27]	Prospective (fMRI)	45 treatment-seeking alcoholics	Alcohol	fMRI (comparator study)	Stress, alcohol cue, and neutral imagery	Disrupted ventromedial PFC function during stress and alcohol cues predicted subsequent relapse; provides fMRI benchmark for fNIRS validation studies.

Note: CW: Continuous wave; DLPFC: Dorsolateral prefrontal cortex; DMPFC: Dorsomedial prefrontal cortex; EEG: Electroencephalography; EF: Executive function; fMRI: Functional magnetic resonance imaging; LDA: Linear discriminant analysis; ML: Machine learning; NF: Neurofeedback; OFC: Orbitofrontal cortex; PFC: Prefrontal cortex; rsFC: Resting-state functional connectivity; SUD: Substance use disorder; SVM: Support vector machine; VFT: Verbal fluency task; VLPFC: Ventrolateral prefrontal cortex; WM: Working memory.

Relapse Prediction

The growing understanding of the PFC’s role in addiction-related executive dysfunction suggests that fNIRS-derived haemodynamic signatures may serve as predictive biomarkers for relapse. Two key themes in addiction neuroscience support this application: first, alterations in ventrolateral and dorsolateral PFC responses to drug cues versus natural reward cues reliably distinguish individuals at high relapse risk; second, failure of the PFC to respond appropriately to natural rewards predicts subsequent relapse in treated patients.^{5, 27} By monitoring these haemodynamic responses non-invasively and in real time, fNIRS offers a portable and cost-effective means to assess relapse vulnerability at the bedside or in outpatient settings, potentially enabling early intervention before relapse occurs.

Functional Near-infrared Spectroscopy in Substance Use Disorders Beyond Alcohol

Classification of Drug Types Using Functional Near-infrared Spectroscopy

Beyond alcohol addiction, fNIRS has been applied to distinguish between different SUDs based on cortical activation patterns. Gu et al. ¹² designed an fNIRS experiment with resting and drug addiction induction paradigms to collect haemodynamic data from 30 drug users (10 methamphetamine users, 10 heroin users and 10 mixed-drug users). Using machine learning algorithms, including linear discriminant analysis (LDA) and support vector machines (SVM), they achieved significant differentiation between substance types based on PFC activation patterns. Notably, methamphetamine users exhibited the highest degree of OFC activation, followed by mixed-drug users, with heroin users showing the lowest activation. This differential pattern suggests that different substances produce distinct neuroadaptive changes in PFC function, and that fNIRS can provide objective, neurobiologically based classification of SUDs—a capability that could inform personalised treatment strategies.

Opioid Use Disorders

Huhn et al. ²⁸ investigated PFC responses to drug cues in methadone-maintained patients using fNIRS and found that increased neural activity in the right dorsolateral PFC during risky decision-making tasks was associated with continued cocaine use. In a separate study, Huhn et al. ²⁹ demonstrated that fNIRS-measured PFC responses to drug cues, along with self-reported craving and current depressive symptoms, were significantly associated with treatment outcomes in methadone-maintained patients. These findings highlight the utility of fNIRS in the complex clinical context of opioid substitution therapy, where monitoring neural responses to drug cues can help identify patients at the highest risk of concurrent substance use and guide treatment intensification.

Furthermore, a bimodal neuroimaging study combining fNIRS with electroencephalography (EEG) in abstinent heroin-dependent patients revealed desynchronised alpha rhythms and decreased PFC connectivity, along with evidence of cerebrovascular injury in the PFC. ³⁰ This pilot study suggests that chronic opiate use may produce lasting structural and functional damage to prefrontal neural circuits, detectable through the combined use of fNIRS and electrophysiological measures.

Behavioural Addictions and Digital Addiction

The application of fNIRS has expanded beyond substance-related addictions to behavioural addictions, including internet gaming disorder and social media addiction. A recent naturalistic fNIRS study by Rahimpour Jounghani et al. ³¹ demonstrated that social media exposure led to reduced accuracy in executive function tasks accompanied by decreased dorsolateral and ventrolateral PFC activation, suggesting that digital addiction produces measurable prefrontal dysfunction similar to that observed in SUDs. These findings underscore the transdiagnostic utility of fNIRS as a tool for evaluating the neural consequences of both chemical and behavioural addictions.

Functional Near-infrared Spectroscopy-Neurofeedback in De-addiction

Concept and Rationale

Neurofeedback (NF) is a form of biofeedback training in which real-time information about brain activity is fed back to the individual, enabling them to learn to self-regulate specific neural circuits through operant conditioning. ³² Given that addiction is characterised by dysregulation of prefrontal cortical control over limbic reward circuits, neurofeedback targeting the PFC represents a theoretically grounded therapeutic approach. fNIRS-NF offers distinct advantages over EEG-NF and fMRI-NF: it provides better spatial specificity than EEG for targeting discrete PFC subregions, while being far more portable, affordable and tolerant of motion than fMRI—critical attributes for clinical deployment in addiction treatment settings. ³³

Current Evidence

A recent systematic review by Fathi et al. ³⁴ identified 32 neurofeedback studies in SUDs, including 3 fNIRS-NF studies. Although fNIRS-NF represents a relatively nascent approach compared to EEG-NF and fMRI-NF, preliminary results are encouraging. Clinical trials are currently underway evaluating fNIRS-NF training from the right dorsolateral PFC using closed-loop systems in patients with alcohol use disorder and prescription opioid use disorder. ³⁵ In these paradigms, patients undergo neurofeedback sessions in which they learn to modulate right dorsolateral prefrontal cortex (rDLPFC) activity while exposed to substance-related cues, with the goal of strengthening prefrontal inhibitory control and reducing cue-induced craving.

Yang et al. ³⁶ demonstrated that a brief real-time fNIRS-informed neurofeedback training protocol targeting the PFC produced measurable changes in brain activity and functional connectivity during subsequent working memory challenges. While this study was not specific to addiction, it provides proof-of-concept that fNIRS-NF can induce neuroplastic changes in PFC function—a critical finding given the central role of PFC dysfunction in addiction. Sanader Vukadinovic ³³ noted in a recent mini-review that fNIRS-NF represents a particularly promising and scalable approach for addiction treatment, given its tolerability in naturalistic settings and potential for integration into existing treatment programmes.

Integration with Cognitive Behavioural Therapy

An important emerging application is the integration of fNIRS-based monitoring with cognitive behavioural therapy (CBT) for addiction. Changes in PFC activity during CBT sessions or pharmacological interventions can be tracked in real time using fNIRS, providing objective neurobiological feedback on treatment response. ³⁷ This approach enables clinicians to identify patients who are or are not responding to specific therapeutic strategies at the neural level, potentially facilitating earlier treatment modification and personalised intervention planning.

Comparative Analysis of Functional Near-infrared Spectroscopy and Functional Magnetic Resonance Imaging in Addiction Research

Both fNIRS and fMRI measure haemodynamic correlates of neural activity, but they offer complementary strengths and limitations for addiction research. fMRI provides superior spatial resolution (1–3 mm vs 1–3 cm) and whole-brain coverage including subcortical structures such as the amygdala, ventral striatum and nucleus accumbens—regions critically implicated in reward processing and habit formation.^{4, 38} fMRI excels at mapping disrupted functional connectivity in reward and executive control networks, evaluating neural changes following pharmacological interventions (e.g., naltrexone, disulfiram) and providing detailed neural network analyses.^{39, 40}

Conversely, fNIRS offers several practical advantages: portability enabling bedside and outpatient assessment, tolerance to motion artefacts permitting naturalistic study designs, substantially lower acquisition and maintenance costs facilitating large-scale studies and the absence of claustrophobic scanning environments that may be particularly problematic for anxiety-prone addiction populations.^{6, 41} fNIRS is also compatible with longitudinal designs requiring repeated assessments, as there are no cumulative exposure risks or scanner time constraints.

A recent systematic review by Liu et al. ⁴² examining 63 studies of combined fMRI–fNIRS integration highlighted the synergistic potential of these modalities: fMRI’s high spatial resolution can be combined with fNIRS’s superior temporal resolution and ecological validity to achieve robust spatiotemporal mapping of neural activity. In addiction research, multimodal approaches combining fNIRS and fMRI, potentially supplemented by EEG, represent the frontier for achieving comprehensive characterisation of the neural dynamics underlying craving, relapse and recovery. ⁴³

Importantly, AI-driven analysis of combined fNIRS and fMRI data holds promise for enhancing relapse prediction models and developing personalised intervention strategies. ⁴⁴ The key brain regions involved in addiction—PFC, anterior cingulate cortex, amygdala and striatum—regulate impulse control, reward processing and emotional regulation and can be evaluated by both modalities, with fNIRS providing superior monitoring of cortical regions and fMRI providing access to deeper structures. ⁴

Recent Advances and Future Directions

Wearable and Wireless Functional Near-infrared Spectroscopy Systems

The development of wearable, wireless and miniaturised fNIRS devices represents a transformative advance for addiction research. ¹⁹ These systems enable data collection during real-world activities, including exposure to naturalistic substance cues in community settings, social drinking environments and during routine clinical encounters—contexts that are impossible to replicate in conventional neuroimaging laboratories. Wearable fNIRS platforms offer the unprecedented opportunity to monitor PFC function during actual moments of craving and decision-making, potentially yielding ecological insights that surpass those derived from laboratory-based paradigms.

Machine Learning and Artificial Intelligence

The integration of advanced machine learning algorithms with fNIRS data is rapidly expanding the analytical capabilities of this modality. Classification algorithms such as support vector machines, random forests and deep learning networks have been applied to fNIRS activation patterns to classify SUD types, ¹² predict treatment outcomes ²⁵ and identify neural markers of relapse vulnerability. ⁴⁴ These approaches have the potential to transform fNIRS from a research tool into a clinical decision-support system for addiction medicine.

Multimodal Integration

Combining fNIRS with other neuroimaging and neurophysiological modalities—including EEG, fMRI and transcranial direct current stimulation (tDCS)—is emerging as a powerful strategy for comprehensive characterisation of addiction neurocircuitry.^{30, 42} Concurrent fNIRS–EEG recording offers the combination of haemodynamic and electrophysiological measures from overlapping cortical regions, while sequential fNIRS–fMRI paradigms can leverage fMRI’s spatial precision for cortical and subcortical mapping alongside fNIRS’s ecological and temporal advantages.

Critical Knowledge Gaps and Research Priorities

Despite the promising evidence reviewed herein, several critical gaps must be addressed before fNIRS can achieve widespread clinical adoption in addiction medicine. First, there is a pressing need for standardised acquisition and analysis protocols to enable cross-study comparisons and meta-analyses. Second, large-scale, multicentre validation studies are required to establish the sensitivity, specificity and predictive validity of fNIRS-derived biomarkers for addiction diagnosis, treatment monitoring and relapse prediction. Third, longitudinal studies with extended follow-up periods are needed to determine whether fNIRS-measured changes in PFC function predict long-term clinical outcomes. Fourth, cost-effectiveness analyses comparing fNIRS-guided treatment strategies with standard clinical approaches are essential for informing healthcare policy. Finally, investigations into the impact of demographic variables (age, sex, ethnicity), comorbid psychiatric conditions and polysubstance use on fNIRS signals are needed to establish the generalisability and robustness of findings across diverse addiction populations.

Conclusion

fNIRS has established itself as a valuable, versatile and clinically accessible neuroimaging tool for addiction research and clinical practice. Its unique combination of non-invasiveness, portability, real-time monitoring capability and ecological validity positions it as an ideal complement to conventional neuroimaging modalities in the study of SUDs. The evidence reviewed in this article demonstrates that fNIRS reliably detects altered PFC haemodynamic responses in individuals with addiction, can track neural recovery during abstinence, shows promise as a biomarker for treatment response and relapse prediction and is emerging as a therapeutic tool through neurofeedback applications.

Both fNIRS and fMRI offer valuable but distinct insights into addiction neuroscience. While fMRI provides comprehensive neural network mapping with superior spatial resolution and whole-brain coverage, fNIRS enables portable, real-world monitoring that makes it a practical tool for longitudinal clinical assessment and intervention monitoring. Their integration through multimodal approaches, augmented by machine learning analytics, represents the most promising pathway toward personalised, neurobiologically informed addiction treatment.

As research progresses, the implementation of standardised protocols, validation in diverse populations and rigorous clinical trials will be essential for translating the potential of fNIRS from the research laboratory to routine clinical practice. The ultimate vision is one in which fNIRS-based assessments become an integral component of personalised addiction medicine—enabling objective diagnosis, real-time treatment monitoring, data-driven relapse prediction and targeted neurofeedback interventions that collectively improve recovery outcomes for individuals with SUDs.

Authors’ Contributions

MK: Conceptualisation, literature search, manuscript drafting and final approval. KM: Critical revision, intellectual content and final approval. TM: Literature search, data curation and manuscript review. JNA: Supervision, critical revision and final approval. UM: Literature search and manuscript editing. GV: Data curation and manuscript review. GA: Manuscript editing and review. SA: Literature search and manuscript review. RN: Manuscript editing and review. SS: Manuscript review. VS: Manuscript review and editing. All authors have read and approved the final version of the manuscript.

ICMJE Statement

All authors meet the International Committee of Medical Journal Editors (ICMJE) criteria for authorship. Each author has made substantial contributions to the conception or design of the work, or the acquisition, analysis or interpretation of data; has participated in drafting the work or revising it critically for important intellectual content; has approved the final version to be published and agrees to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Statement of Ethics and Informed Consent

Not applicable.