Frontiers in Neuroscience

Glaucoma is increasingly recognized as a neurodegenerative condition involving both retinal and central nervous system structures. Here, we present an integrated framework that combines MK-Curve-corrected diffusion kurtosis imaging (DKI), tractometry, and deep autoencoder-based normative modeling to detect localized white matter abnormalities associated with glaucoma. Using UK Biobank diffusion MRI data, we show that MK-Curve approach corrects anatomically implausible values and improves the reliability of DKI metrics - particularly mean (MK), radial (RK), and axial kurtosis (AK) - in regions of complex fiber architecture. Tractometry revealed reduced MK in glaucoma patients along the optic radiation, inferior longitudinal fasciculus, and inferior fronto-occipital fasciculus, but not in a non-visual control tract, supporting disease specificity. These abnormalities were spatially localized, with significant changes observed at multiple points along the tracts. MK demonstrated greater sensitivity than MD and exhibited altered distributional features, reflecting microstructural heterogeneity not captured by standard metrics. Node-wise MK values in the right optic radiation showed weak but significant correlations with retinal OCT measures (ganglion cell layer and retinal nerve fiber layer thickness), reinforcing the biological relevance of these findings. Deep autoencoder-based modeling further enabled subject-level anomaly detection that aligned spatially with group-level changes and outperformed traditional approaches. Together, our results highlight the potential of advanced diffusion modeling and deep learning for sensitive, individualized detection of glaucomatous neurodegeneration and support their integration into future multimodal imaging pipelines in neuro-ophthalmology.

Acoustic simulations of cochlear implants (CIs) allow for studies of perceptual performance with minimized effects of large CI individual variability. Different from conventional simulations using continuous sinusoidal or noise carriers, the present study employs pulsatile Gaussian-enveloped tones (GETs) to simulate several key features in modern CIs. Subject to the time-frequency uncertainty principle, the GET has a well-defined tradeoff between its duration and bandwidth. Two types of GET vocoders were implemented and evaluated in normal-hearing listeners. In the first implementation, constant 100-Hz GETs were used to minimize within-channel temporal overlap while different GET durations were used to simulate electric channel interaction. This GET vocoder could produce vowel and consonant recognition similar to actual CI performance. In the second implementation, 900-Hz/channel pulse trains were directly mapped to 900-Hz GET trains to simulate the maxima selection and amplitude compression of a widely-used n-of-m processing strategy, or the Advanced Combination Encoder. The simulated and actual implant performance of speech-in-noise recognition was similar in terms of the overall trend, absolute mean scores, and standard deviations. The present results suggest that the pulsatile GET vocoders can be used as alternative vocoders to simultaneously simulate several key CI processing features and result in similar speech perception performance to that with modern CIs.

This study examines cortical responses to chromatic and luminance stimuli in individuals with normal trichromatic vision, Daltonism, and achromatopsia. Functional magnetic resonance imaging (fMRI) data were collected using stimuli modeled after Wade et al. (2008) to evaluate the differential activation of visual cortical areas. In normal trichromats, hV4 demonstrated the highest chromatic sensitivity, while ventral areas showed stronger responses to color compared to dorsal regions. In Daltonic and achromatopsia participants, cortical activation was observed under combined chromatic and luminance conditions; however, no significant color-specific activity was detected, even in hV4. This work establishes a baseline for understanding cortical responses in color vision deficiencies and preceded gene therapy studies in the same achromatopsia patients (Fischer et al., 2020; Seitz et al., 2022). These findings contribute to ongoing research into neural plasticity and targeted therapeutic interventions.

Respiration is strongly linked to internal states such as arousal, emotion, and even cognitive processes and provides objective biological information to estimate these states in humans and animals. However, the measurement of respiration has not been established in macaque monkeys that have been widely used as model animals for understanding various higher brain functions. In the present study, we developed a method to monitor the respiration of behaving monkeys. We first measured the temperature of their nasal breathing, which changes between inspiration and expiration phases, in an anesthetized condition and estimated the respiration pattern. We compared the estimated pattern with that obtained by a conventional chest band method that has been used in humans and applies to anesthetized, but not behaving, monkeys. These respiration patterns matched well, suggesting that the measurement of nasal air temperature can be used to monitor the respiration of monkeys. Furthermore, we confirmed that the respiration frequency in behaving monkeys monitored by the measurement of nasal air temperature was not affected by the orofacial movement of licking to obtain the liquid reward. We next examined the frequency of respiration when they listened to music or white noise. The respiratory frequency was higher when the monkeys listened to music than the noise. This result is consistent with a phenomenon in humans and indicates the accuracy of our monitoring method. These data suggest that the measurement of nasal air temperature enables us to monitor the respiration of behaving monkeys and thereby estimate their internal states. Significance StatementWhile respiration is linked with internal processing, such as emotional and cognitive states, methods have not been established for physiological research on monkeys. We developed a novel method that obtained respiration signals by measuring the nasal air temperatures of behaving monkeys. Our method was able to continuously track the respiration pattern without distortions evoked by orofacial movements to lick the liquid reward. The respiratory frequency increased while listening to music than when listening to white noise for all monkeys. These results demonstrate that nasal air temperature measurements can be used to monitor the respiration patterns of aroused monkeys, allowing us to understand their internal state. This will be useful for investigating the underlying neuronal mechanism of neuropsychiatric disorders using monkeys.

PurposeThe auditory brainstem response (ABR) is a widely used objective electrophysiology measure for non-invasively assessing auditory function and neural activity in the auditory brainstem, but its ability to reflect detailed neuronal processing is limited due to the averaging nature of the electroencephalogram-type recordings. MethodThis study addresses this limitation by developing a computational model of the auditory brainstem which is capable of synthesizing ABR traces based on a large, population scale neural extrapolation of a spiking neuronal network of auditory brainstem circuitry. The model was able to recapitulate alterations in ABR waveform morphology that have been shown to be present in two medical conditions: animal models of autism and aging. Moreover, in both of these conditions, these ABR alterations are caused by known distinct changes in auditory brainstem physiology, and the model could recapitulate these changes. ResultsIn the autism model, the simulation revealed myelin deficits and hyperexcitability, which caused a decreased wave III amplitude and a prolonged wave III-V interval, consistent with experimentally recorded ABRs in Fmr1-KO mice. For the aging condition, the model recapitulated ABRs recorded in aged gerbils and indicated a reduction in activity in the medial nucleus of the trapezoid body (MNTB), a finding validated by confocal imaging data. ConclusionThese results demonstrate not only the models accuracy but also its capability of linking features of ABR morphology to underlying neuronal properties and suggesting follow-up physiological experiments.

Background and objectivesBrain-computer interfaces (BCIs) hold promise as augmentative and alternative communication technology for people with severe motor and speech impairment (locked-in syndrome) due to neural disease or injury. Although such BCIs should be available 24/7, to enable communication at all times, feasibility of nocturnal BCI use has not been investigated. Here, we addressed this question using data from an individual with amyotrophic lateral sclerosis (ALS) who was implanted with an electrocorticography-based BCI that enabled the generation of click-commands for spelling words and call-caregiver signals. MethodsWe investigated nocturnal dynamics of neural signal features used for BCI control, namely low (LFB: 10-30Hz) and high frequency band power (HFB: 65-95Hz). Additionally, we assessed the nocturnal performance of a BCI decoder that was trained on daytime data by quantifying the number of unintentional BCI activations at night. Finally, we developed and implemented a nightmode decoder that allowed the participant to call a caregiver at night, and assessed its performance. ResultsPower and variance in HFB and LFB were significantly higher at night than during the day in the majority of the nights, with HFB variance being higher in 88% of nights. Daytime decoders caused 245 unintended selection-clicks and 13 unintended caregiver-calls per hour when applied to night data. The developed nightmode decoder functioned error-free in 79% of nights over a period of {+/-}1.5 years, allowing the user to reliably call the caregiver, with unintended activations occurring only once every 12 nights. DiscussionReliable nighttime use of a BCI requires decoders that are adjusted to sleep-related signal changes. This demonstration of a reliable BCI nightmode and its long-term use by an individual with advanced ALS underscores the importance of 24/7 BCI reliability. Trial registrationThis trial is registered in clinicaltrials.gov under number NCT02224469 (https://clinicaltrials.gov/study/NCT02224469?term=NCT02224469&rank=1). Date of submission to registry: August 21, 2014. Enrollment of first participant: September 7, 2015.

Subjective tinnitus is the perception of sound in the absence of an external source, with a neurobiological basis that remains poorly understood. Importantly, tinni-tus prevalence increases with aging, reaching up to 25 % in adults aged above 65 years. This study examines white matter tract alterations in older adults with tinnitus using diffusion magnetic resonance imaging. The research involved 96 individuals from the Chilean ANDES cohort, including 56 patients with tinnitus and 40 controls. Thirty-six deep white matter (DWM) and 84 superficial white matter (SWM) bundles were segmented. For each bundle, we extracted four diffu-sion tensor imaging metrics: axial diffusivity (AD), mean diffusivity (MD), radial diffusivity (RD), and fractional anisotropy (FA). With these features, we trained an Extreme Gradient Boosting classifier to predict tinnitus, achieving an AUC of 0.93, highlighting the relevance of AD. Key DWM bundles included the ante-rior arcuate fasciculus, inferior longitudinal fasciculus, and cingulum short fibers (right hemisphere), and the superior motor thalamic radiation (left hemisphere). Significant SWM bundles in the left hemisphere included the superior parietal, cuneus, lingual, superior temporal, caudal middle frontal, precentral, fusiform, postcentral, supramarginal, rostral middle frontal, and superior frontal regions. Tinnitus patients showed decreased AD, MD, and RD, and increased FA, sug-gesting microstructural reorganization. These changes may reflect adaptive or maladaptive plasticity. Increased FA could signal compensatory responses, while decreased AD might indicate axonal damage. Meanwhile, the decrease in MD and RD could indicate an increase in myelin integrity. This is the first study to investigate SWM bundle alterations in tinnitus patients.

Here, we present a novel analysis pipeline for scalp EEG signals. With our approach, we focus on the interchannel relationships using standard correlation techniques in the spatiotemporal and frequency domains as well as a recently developed triple correlation-based analysis. Since we compute correlations in multichannel EEG data, we employ (non-standard) bipolar montages with non-overlapping electrode positions to avoid montage-induced correlations in space-, time- and frequency domains. In addition to the correlation studies, we determine the EEG frequency components, phase-amplitude coupling and a burst index. The procedures are tested with simulated signals and illustrated with results obtained from a small patient group.

Functional near-infrared spectroscopy (NIRS) is an increasingly popular method in hearing research. However, few studies have considered efficient stimulation parameters for fNIRS auditory experimental design. The objectives of our study are (1) to characterize the auditory hemodynamic responses to trains of white noise with increasing stimulation durations (8s, 10s, 15s, 20s) in terms of amplitude and response linearity; (2) to identify the most-efficient stimulation duration using fNIRS; and (3) to generalize results to more ecological environmental stimuli. We found that cortical activity is augmented following the increments in stimulation durations and reaches a plateau after about 15s of stimulation. The linearity analysis showed that this augmentation due to stimulation duration is not linear in the auditory cortex, the non-linearity being more pronounced for longer durations (15s and 20s). The 15s block duration that we propose as optimal precludes signal saturation, is associated with a high response amplitude and a relatively short total experimental duration. Moreover, the 15s duration remains optimal independently of the nature of presented sounds. The sum of these findings suggests that 15s stimulation duration used in the appropriate experimental setup allows researchers to acquire optimal fNIRS signal quality.

In the last decades, planar multi-electrode arrays (MEAs) have been widely used to record activity from in vitro neuronal cell cultures and tissue slices. Though successful, this technique bears some limitations, particularly relevant when applied to three-dimensional (3D) tissue, such as brain slices, spheroids or organoids. For example, planar MEAs signals are informative on just one side of a 3D-organized structure. This limits the interpretation of the results in terms of network functions in a complex structured and hyperconnected brain tissue. Moreover, the side in contact with the MEAs often shows lower oxygenation rates and related vitality issues. To overcome these problems, we empowered a CMOS high-density multi-electrode array (HD-MEA) with thousands of microneedles (needles) of 65-90 m height, able to penetrate and record in-tissue signals, providing for the first time a 3D HD-MEA chip. We propose a CMOS-compatible fabrication process to produce arrays of needles of different widths mounted on large pedestals to create microchannels underneath the tissue. By using cerebellar and cortico-hippocampal slices as a model, we show that the needles efficiently penetrate the 3D tissue while the microchannels allow the flowing of maintenance solutions to increase tissue vitality in the recording sites. These improvements are reflected by the increase in electrodes sensing capabilities, the number of sampled neuronal units (compared to matched planar technology), and the efficiency of compound effects. Importantly, each electrode can also be used to stimulate the tissue with optimal efficiency due to the 3D structure. Furthermore, we demonstrate how the 3D HD-MEA can efficiently penetrate and get outstanding signals from in vitro 3D cellular models as brain spheroids. In conclusion, we describe a new recording device characterized by the highest spatio-temporal resolution reported for a 3D MEA and significant improvements in the quality of recordings, with a high signal-to-noise ratio and improved tissue vitality. The applications of this game-changing technique are countless, opening unprecedented possibilities in the neuroscience field and beyond.

PurposeTo evaluate the feasibility of detecting single-trial cortical responses to individual microperimetry (MP) stimuli using electroencephalography (EEG) under non-ideal synchronization conditions, and to explore EEG-based stimulus registration independent of patient responses. MethodsThis proof-of-concept study acquired EEG data from two healthy participants (12 trials) during MP testing using low- and high-intensity single-flash stimuli. Occipital EEG signals were recorded with an 8-channel portable system, band-pass filtered (4-49 Hz), normalized, and segmented into 600-ms epochs time-locked to MP stimuli using offline synchronization, resulting in an estimated temporal uncertainty of [~]250 ms. A bidirectional long short-term memory (BiLSTM) deep learning model classified stimulus-present versus stimulus-absent EEG segments. Performance was assessed using accuracy, sensitivity, specificity, and F1-score, with emphasis on feasibility rather than generalizability. ResultsAcross trials, EEG-based classification performance exceeded chance levels. For high-intensity stimuli, detection accuracy approached 80%, while low-intensity stimuli demonstrated greater inter-trial variability. Occipital electrode configurations consistently outperformed parietal or combined montages, consistent with visual cortex neuroanatomy. Despite the absence of hardware-level synchronization and single-trial analysis, detectable neural signatures of isolated MP flashes were observed. ConclusionThese findings demonstrate the feasibility of detecting single-flash MP stimuli from occipital EEG using deep learning, even under constrained acquisition conditions. While not intended for immediate clinical deployment, this work motivates future studies incorporating precise synchronization, optimized stimulus designs, and larger cohorts to evaluate EEG-augmented microperimetry as a potential objective adjunct to subjective functional testing. HighlightsO_LIEEG detects single-flash microperimetry stimuli without hardware-level synchronization C_LIO_LIOccipital EEG channels enable stimulus detection independent of patient responses C_LIO_LIBiLSTM deep learning decodes non-repetitive, long-duration microperimetry flashes C_LIO_LIDetection accuracy approaches 80% for high-intensity microperimetry stimuli C_LI

On-line classification of neural recordings can be extremely useful in brain-machine interface, prosthetic applications or therapeutic intervention. In this work we present a feasibility study for developing compact low-power VLSI systems able to classify neural recordings in real-time, using spike-based neuromorphic circuits. We developed a framework for classifying extra-cellular recordings made in rat auditory cortex in response to different auditory stimuli and porting the classification algorithm onto a spiking multi-neuron VLSI chip with programmable synaptic weights. We present recording methods and software classification algorithms; we demonstrate real-time classification in hardware and quantify the system performance; finally, we identify the potential sources of problems in developing such types of systems and propose strategies for overcoming them.

Novel in vitro platforms based on human neurons are needed to improve early drug testing and address the stalling drug discovery in neurological disorders. Topologically controlled circuits of human induced pluripotent stem cell (iPSC)-derived neurons have the potential to become such a testing system. In this work, we build in vitro co-cultured circuits of human iPSC-derived neurons and rat primary glial cells using microfabricated polydimethylsiloxane (PDMS) structures on microelectrode arrays (MEAs). Such circuits are created by seeding either dissociated cells or pre-aggregated spheroids at different neuron-to-glia ratios. Furthermore, an antifouling coating is developed to prevent axonal overgrowth in undesired locations of the microstructure. We assess the electrophysiological properties of different types of circuits over more than 50 days, including their stimulation-induced neural activity. Finally, we demonstrate the effect of magnesium chloride on the electrical activity of our iPSC circuits as a proof-of-concept for screening of neuroactive compounds.

Modern AI has flourished without molecular detail. Here I examine neural maturation in computational and molecular-physiological contexts. Immature transmission supports learning but collapses beyond task/load thresholds. A developmental screen identified KCNH7, upregulated in mouse cortex, incorporating its channel kinetics elevated spike threshold and stabilized recall, linking maturation to robustness without mechanistic elaboration. In unsupervised Restricted Boltzmann Machines, greater dynamical instability improved performance on complex tasks, highlighting a complementary regime. Developmentally plausible changes in KCNH7 conductance re-tuned reservoir dynamics and chaos, with appropriate levels improving memory and generalization, whereas excessive conductance degraded stability. Brief GPCR-like K+ modulation balanced responsiveness and accuracy, suggesting neuromodulatory K+ control as a lever for high-dimensional learning.

A current challenge of neurotechnologies is the development of speech brain-computer interfaces to restore communication in people unable to speak. To achieve a proof of concept of such system, neural activity of patients implanted for clinical reasons can be recorded while they speak. Using such simultaneously recorded audio and neural data, decoders can be built to predict speech features using features extracted from brain signals. A typical neural feature is the spectral power of field potentials in the high-gamma frequency band (between 70 and 200 Hz), a range that happens to overlap the fundamental frequency of speech. Here, we analyzed human electrocorticographic (ECoG) and intracortical recordings during speech production and perception as well as rat microelectrocorticographic ({micro}-ECoG) recordings during sound perception. We observed that electrophysiological signals, recorded with different recording setups, often contain spectrotemporal features highly correlated with those of the sound, especially within the high-gamma band. The characteristics of these correlated spectrotemporal features support a contamination of electrophysiological recordings by sound. In a recording showing high contamination, using neural features within the high-gamma frequency band dramatically increased the performance of linear decoding of acoustic speech features, while such improvement was very limited for another recording showing weak contamination. Further analysis and in vitro replication suggest that the contamination is caused by a mechanical action of the sound waves onto the cables and connectors along the recording chain, transforming sound vibrations into an undesired electrical noise that contaminates the biopotential measurements. This study does not question the existence of relevant physiological neural information underlying speech production or sound perception in the high-gamma frequency band, but alerts on the fact that care should be taken to evaluate and eliminate any possible acoustic contamination of neural signals in order to investigate the cortical dynamics of these processes.

There has been no study on the relationship between chronic tinnitus and harmonic templates. Harmonic templates are harmonically structured receptive fields in the auditory system in which all frequency components are integer multiples of a common fundamental frequency (F0). In this study, data from 19 harmonic templates from each of 196 chronic tinnitus patients were analyzed and mathematical modeling was performed to quantify the loudness of chronic tinnitus. High-resolution hearing threshold data were obtained by algorithmic pure tone audiometry (PTA) conducting automated PTA at 134 frequency bands with 1/24 octave resolution from 250 Hz to 12,000 Hz. The result showed that there is an intriguing relationship between the auditory instability of harmonic templates and simplified tinnitus severity score (STSS). This study provides several mathematical models to estimate tinnitus severity and the precise quantification of the loudness of chronic tinnitus. Our computational models and analysis of the behavioral hearing threshold fine structure suggest that the cause of severe chronic tinnitus could be a severe disparity between different temporal capacities of each neural oscillator in a certain harmonic template.

The features of dopaminergic neurons (DAns) of nigrostriatal circuitry are orchestrated by a multitude of yet unknown factors, many of them genetic. Genetic variation between individuals at baseline can lead to differential susceptibility to and severity of diseases. As decline of DAns, a characteristic of Parkinsons disease, heralds a significant decrease in dopamine level, measuring dopamine can reflect the integrity of DAns. To identify novel genetic regulators of the integrity of DAns, we used the Collaborative Cross (CC) mouse strains as model system to search for quantitative trait loci (QTLs) related to dopamine levels in the dorsal striatum. The dopamine levels in dorsal striatum varied greatly in the eight CC founder strains, and the differences were inheritable in 32 derived CC strains. QTL mapping in these CC strains identified a QTL associated with dopamine level on chromosome X containing 393 genes. RNA-seq analysis of the ventral midbrain of two of the founder strains with large striatal dopamine difference (C57BL/6J and A/J) revealed 24 differentially expressed genes within the QTL. The protein-coding gene with the highest expression difference was Col4a6, which exhibited a 9-fold reduction in A/J compared to C57BL/6J, consistent with decreased dopamine levels in A/J. Publicly available single cell RNA-seq data from developing human midbrain suggests that Col4a6 is highly expressed in radial glia-like cells and neuronal progenitors, indicating possible involvement in neurogenesis. Interestingly, the lowered dopamine levels were accompanied by reduced striatal axonal branching of striatal DAns in A/J compared to C57BL/6J. Because Col4a6 is known to control axogenesis in non-mammal model organisms, we hypothesize that different dopamine levels in mouse dorsal striatum are due to differences in axogenesis induced by varying COL4A6 levels during neural development.

Recent studies suggest that 40Hz photoacoustic stimulation is beneficial for improving cognition, and is safe and effective in patients with Alzheimers disease (AD) and mild cognitive impairment (MCI). It has been reported that the healthy population responds more strongly to the 40 Hz intervention compared to MCI and AD patients. Therefore, the aim of this study was to explore the effectiveness of 40 Hz photoacoustics in preventing cognitive impairment and improving sleep quality in the healthy population. Four weeks of 40 Hz intervention significantly improved sleep quality (PSQI) (p=0.028) and enhanced functional connectivity in the hippocampus and default mode network (DMN) (in the cognitively impaired population, functional connectivity showed a decreasing trend). These results were consistent with findings from the FDA clinical trial on MCI/AD patients that showed enhanced connectivity of the core regions PCC and mPFC within the DMN network. No serious adverse effects were reported throughout the intervention. This study is the first to validate the efficacy and safety of a 40 Hz photoacoustic intervention in a healthy population, suggesting its potential in preventing cognitive decline in a healthy population.

The Small Optic Lobe (SOL) family of calpains are intracellular cysteine proteases that are expressed in the nervous system and play an important role in neuronal development in both Drosophila, where loss of this calpain leads to the eponymous small optic lobes, and in mouse and human, where loss of this calpain leads to eye anomalies. Some human individuals with biallelic variants in CAPN15 also have developmental delay and autism. However, neither the specific effect of the loss of the Capn15 protein on brain development nor the brain regions where this calpain is expressed in the adult is known. Here we show using small animal MRI that mice with the complete loss of Capn15 have smaller brains overall with larger decreases in the thalamus and subregions of the hippocampus. These losses are not seen in Capn15 conditional KO mice where Capn15 is knocked out only in excitatory neurons in the adult. Based on {beta}-galactosidase expression in an insert strain where lacZ is expressed under the control of the Capn15 promoter, we show that Capn15 is expressed in adult mice, particularly in neurons involved in plasticity such as the hippocampus, lateral amygdala and Purkinje neurons, and partially in other non-characterized cell types. The regions of the brain in the adult where Capn15 is expressed do not correspond well to the regions of the brain most affected by the complete knockout suggesting distinct roles of Capn15 in brain development and adult brain function.

Problem StatementPain has a crucial function in the human body acting as an early warning signal to protect against tissue damage. However, both assessment of pain experience and its clinical diagnosis rely on highly subjective methods. Objective evaluation of the presence of pain under analgesic drug administrations becomes even more complicated. ObjectivesThe aim of this study was to propose a transfer learning (TL) based deep learning (DL) methodology for accurate detection and objective classification of the neural processing of painful and non-painful stimuli that were presented under different levels of analgesia. MethodA publicly available fNIRS dataset of 14 participants was obtained during an experimental protocol that involved painful and non-painful events. Separate fNIRS scans were taken under the same nociceptive protocol before analgesic drug (Morphine and Placebo) administration and at three different times (30,60 and 90 min) post-administration. By utilizing data from pre-drug fNIRS scans, a DL architecture for classifying painful and non-painful stimuli was constructed as a base model. Knowledge generated in pre-drug base model was transferred to 6 distinct post-drug conditions by adapting a TL approach. The DeepSHAP method was utilized to unveil the contribution weights of nine R OIs for each of the pre-drug and post-drug models. ResultsMean performance of pre-drug base model was above 95% for accuracy, sensitivity, specificity and AUC metrics. Each of the post-drug models had mean accuracy, sensitivity, specificity and AUC performance above 90%. No statistically significant difference across post-drug models were found for classification performance of any of the performance metrics. Post-placebo models had higher decoding accuracy than post-morphine models. ConclusionKnowledge obtained from a pre-drug base model could be successfully utilized to build pain decoding models for six distinct brain states that were altered with either analgesic or placebo intervention. Contribution of different cortical regions to classification performance varied across the post-drug models. ImportanceThe proposed methodology may remove the necessity to build new DL models for data collected at clinical or daily life conditions for which obtaining training data is not practical or building a new decoding model will have a computational cost. Unveiling the explanation power of different cortical regions may aid the design of more computationally efficient fNIRS based BCI system designs that target other application areas.