(1) Patients diagnosed with first-ever supratentorial stroke according to established stroke guidelines, with imaging confirming supratentorial lesion location. (2) Time since stroke onset ≤6 months. (3) Right-handedness confirmed by the Edinburgh Handedness Inventory (EHI). (4) Stable vital signs and a Mini-Mental State Examination (MMSE) score ≥24, indicating patients could understand instructions and cooperate with assessments. (5) No history of dysphagia caused by other neurological diseases, head/neck structural abnormalities, or injuries. (6) No history of neuropsychiatric or psychiatric disorders. (7) Absence of tracheostomy.

Flexible endoscopic evaluation of swallowing (FEES) following a volume-viscosity swallow test (V-VST) protocol (20) confirmed impaired swallowing efficacy and/or safety. Patients were diagnosed with oropharyngeal dysphagia, with a Penetration–Aspiration Scale (PAS) score ≤3 when swallowing 5 mL of moderately thick liquid.

FEES following a V-VST protocol confirmed no impairment in swallowing efficacy or safety.

(1) History of previous stroke. (2) Non-stroke-related dysphagia (e.g., dysphagia caused by other neurological or systemic conditions). (3) Auditory or visual impairments. (4) Implanted cardiac pacemakers, intracranial metal implants, cochlear implants, or metallic dentures. (5) Skin lesions on the head or neck, or allergies to electrode pads. (6) Skull defects. (7) History of antidepressant, sedative, or other medications potentially affecting cortical excitability.

(1) Development of new infarction or hemorrhagic lesions. (2) Inability to tolerate examinations. (3) Withdrawal from the study for other reasons.

All participants provided informed consent and underwent screening according to predefined inclusion and exclusion criteria. The study protocol was approved by the Ethics Committee of Binzhou Medical University Hospital (Approval No. 2024KYLL-092) and registered with the Chinese Clinical Trial Registry (Registration No. ChiCTR2500098164). The registration was completed after participant enrollment but before data analysis.

Initially, raw fNIRS intensity data were converted to optical density values. Data points exhibiting signal fluctuations greater than six standard deviations were identified as motion artifacts and corrected using spline interpolation. Subsequently, a bandpass filter (0.01–0.2 Hz) was applied to remove high-frequency physiological noise (e.g., respiration and cardiac activity) and low-frequency baseline drift. Optical density data were then converted into oxyhemoglobin and deoxyhemoglobin concentration changes using the modified Beer–Lambert law. To standardize and ensure comparability across all participants, including those with left- or right-hemisphere lesions, fNIRS functional imaging data from patients with right-hemisphere lesions were horizontally flipped, placing the affected hemisphere uniformly on the left side. Compared to deoxyhemoglobin, oxyhemoglobin concentration changes more sensitively and precisely reflect neuron-induced hemodynamic responses and have been widely employed as critical indicators of local cerebral hemodynamics. Therefore, oxyhemoglobin concentration changes were utilized in this study as the primary measure of cerebral hemodynamic responses.

A general linear model (GLM) was used to fit and analyze hemodynamic response functions, estimating β-values to quantify neural activity levels and patterns across brain regions during specific swallowing tasks. Data from five repeated task blocks under each swallowing condition were analyzed using a block-average method. First, the period from 0–3 s was selected as the baseline, and all task blocks were averaged accordingly. Hemodynamic indices, including central value, mean value, differential value, slope, and integral, were calculated for each channel within task blocks, enabling comprehensive evaluation of cerebral hemodynamic responses during different swallowing tasks.

During the SST, the non-dysphagic stroke control group primarily activated the R-S1, bilateral FPA, and R-PSMC. This activation pattern is broadly consistent with that reported in healthy individuals during water swallowing tasks (16), suggesting relatively preserved sensorimotor integration in non-dysphagic stroke patients.

In contrast, the dysphagia group exhibited notably different activation patterns during the SST, primarily involving bilateral PSMC (right: CH15, CH18; left: CH29, CH32), the R-FPA (CH5), and the L-DLPFC (CH27). The extensive activation of bilateral PSMC suggests that PSD patients require increased recruitment of motor cortical resources to execute even basic swallowing movements (22). The involvement of L-DLPFC warrants particular attention, as this region is not typically activated during simple swallowing tasks in healthy individuals. The DLPFC plays a central role in executive control, working memory, and attentional regulation (23). Its activation during the SST in PSD patients suggests that these patients require greater cognitive control and attentional resources even when performing simple swallowing actions. In healthy individuals, swallowing is often described as a semi-automatic process in which, once initiated, the motor sequence proceeds with minimal conscious oversight (24). The significantly greater L-DLPFC activation in PSD patients compared to controls performing the identical task indicates that PSD patients cannot execute swallowing with the same degree of automaticity, reflecting the initiation of compensatory neural mechanisms (25).

It is important to note that auditory cues were incorporated into the experimental paradigm to ensure temporal synchronization between swallowing tasks and fNIRS data acquisition, consistent with established protocols (17). However, in this study, the auditory cue was only used to signal the onset of each task block rather than to control individual swallowing actions. In the SST, participants initiated swallowing at their own discretion within the task period; in the CST and DTT, participants heard the cue only once at task onset and thereafter swallowed at their own natural pace. This differs from externally paced paradigms that require a cue for each individual swallow, thereby reducing associated cognitive demands. According to Ambrocio et al. (26), self-paced continuous swallowing may represent an intermediate state between fully cued discrete swallowing and spontaneous swallowing, more closely resembling natural drinking behavior. Importantly, since both groups performed identical tasks under the same experimental conditions, the between-group differences in cortical activation cannot be attributed solely to cueing effects, supporting our interpretation that the observed patterns reflect dysphagia-specific neural mechanisms.

Based on the neuroimaging literature, complex dual-task conditions involving swallowing are expected to activate widespread cortical regions, including bilateral DLPFC for executive control, M1 for motor execution, S1 for sensory feedback, and PSMC for motor preparation (15, 27). During the DTT, the non-dysphagic stroke control group demonstrated this expected widespread activation pattern, involving bilateral DLPFC (CH11, CH12, CH24, CH26, CH27), sensorimotor cortices (R-IFG: CH3; L-S1: CH14; R-M1: CH30), and multiple prefrontal and motor preparatory regions (L-FPA: CH25; L-PSMC: CH34). This widespread activation is consistent with previous findings in healthy individuals performing complex swallowing tasks (28). The DLPFC plays a pivotal role as a central hub for executive functions, including task switching, attentional resource allocation, and working memory maintenance (29). The effective L-DLPFC activation observed in our control participants indicates preserved ability to flexibly allocate cognitive resources for concurrent swallowing movements and visual processing during multitasking. Furthermore, M1 activation facilitates precise motor execution (30), while S1 activation provides essential sensory feedback (31).

In contrast, during the DTT, the dysphagia group exhibited activation restricted only to the L-S1 region (CH14). The lack of activation in M1 and PSMC indicates significantly impaired motor initiation and execution capabilities (32). This restricted activation pattern highlights the loss of effective compensation in cognitive-motor dual tasks among PSD patients. The isolated activation of sensory cortical areas suggests that PSD patients concentrate limited neural resources primarily on sensory monitoring due to impaired motor coordination capabilities (33). Thus, PSD patients display constrained neural resource allocation, hindering their ability to simultaneously process multiple tasks as efficiently as controls. The observed shift from hyperactivation during simple tasks to markedly diminished activation during complex tasks demonstrates clear limitations in the compensatory neural mechanisms of PSD patients (11).

Compensatory neural mechanisms are critical phenomena observed when comparing PSD patients with non-dysphagic stroke controls during swallowing tasks. While both groups experienced stroke, the between-group differences in cortical activation patterns suggest additional compensatory demands in PSD patients beyond any stroke-related neural reorganization present in the control group. Block-average feature analysis provided additional quantitative evidence regarding compensatory neural load in the dysphagia group by examining activation intensity. During the SST, multiple activation indices, including central, mean, differential, slope, and integral values, of several key channels [e.g., R-FPA (CH5), R-PSMC (CH17, CH18), L-PSMC (CH29)] were significantly higher in the dysphagia group than in the control group (22). These findings indicate that PSD patients require substantially greater neural effort to perform basic swallowing tasks compared to non-dysphagic stroke controls (34). Such hyperactivation reflects compensatory responses to functional deficits, predominantly within the R-FPA and PSMC regions; however, this strategy might represent inefficient compensation, consuming greater neural resources without corresponding improvements in swallowing function (35). This increased cortical activity may originate from cortical dysfunction, disrupted excitation-inhibition balance, and altered synaptic plasticity following stroke (36).

However, a marked shift was observed under more demanding tasks, such as the CST and DTT. It should be noted that all tasks employed the same cueing approach, with auditory cues signaling only task onset and participants swallowing at their own natural pace thereafter. Additionally, 5-min rest periods were provided between tasks to minimize fatigue effects. During the CST, mean and integral activation values for R-FPA (CH6) and L-FPA (CH25) were significantly lower in the dysphagia group compared to controls. This transition from extensive overactivation during simple tasks to selective underactivation during complex tasks suggests a saturation effect in compensatory mechanisms (37), although this interpretation should be considered in the context of the fixed task order (SST → CST → DTT). Despite the rest periods provided, we cannot completely exclude the possibility that accumulated cognitive load from preceding tasks may have contributed to the observed activation patterns. Post-stroke cortical networks possess limited compensatory potential, indicating difficulty in further increasing cortical activation once a ceiling is reached (38). Consequently, the capacity to maintain task performance under increasing complexity is limited. During the DTT, only R-S1 (CH16) showed greater activation in the dysphagia group relative to controls. This selective activation likely reflects a compensatory strategy prioritizing core functions such as sensory monitoring to prevent critical complications (e.g., aspiration), while sacrificing secondary functions, including prefrontal cognitive control, due to impaired motor coordination capacity (34). Nonetheless, this strategy may induce excessive anxiety or fear of swallowing, rather than genuinely enhancing functional recovery (39).

The control group exhibited highly adaptive patterns of cortical neuromodulation across swallowing tasks of varying complexity, with significant changes in activation patterns as task difficulty increased. Relative to the SST, brain regions with significantly elevated activation during the CST included bilateral FPA (right: CH7, CH22; left: CH8, CH9), R-DLPFC (CH23), and L-PSMC (CH34). These regions are closely involved in perceptual processing, motor control, and cognitive regulation, reflecting increased demands for motor coordination and rhythmic control during continuous swallowing (40). Activation expanded further during the video-continuous swallowing dual-task, involving additional areas such as sensory cortices (R-S1: CH2; L-S1: CH14), executive function networks (L-DLPFC: CH11, CH26), motor planning and execution regions (R-PSMC: CH15; L-PSMC: CH28, CH33, CH34), and R-M1. These regions collectively underpin sensorimotor processing, motor planning, and execution. Thus, the brain mobilizes extensive neural resources to simultaneously integrate and coordinate swallowing and visual tasks under dual-task conditions.

The hierarchical expansion of activation aligns with the task-demand hypothesis in cognitive neuroscience, which posits that the brain automatically recruits higher-order and more extensive cortical regions as task complexity increases (41, 42). It is particularly noteworthy that activation of the DLPFC not only contributes to executive control but also coordinates complex motor sequences through functional connectivity with the motor cortex (43). In dual-task conditions, the DLPFC coordinates the parallel processing of visual information and swallowing motor execution through its role in the central-executive network (44).

Block-average feature analysis further revealed the fine-tuning of activation intensity within the control group. Compared with the SST, feature values in multiple FPA and PSMC channels significantly increased during continuous swallowing. This enhanced activation expanded further during the video-continuous swallowing dual-task, presenting a coordinated enhancement across the prefrontal-motor cortical network, where differential values of FPA, DLPFC, and PSMC increased synchronously, suggesting the formation of a functional cooperative network (45). Enhanced sensorimotor integration reflects greater demands for precise sensory feedback in dual-task scenarios (46), while specific involvement of M1 highlights its critical role in dual-task execution (15).

The control group demonstrated efficient coordination and execution of dual tasks through widespread cortical activation patterns, particularly within the DLPFC and M1 (25, 47), reflecting flexible allocation of cognitive resources in multitasking contexts. Collectively, these findings depict a dynamic, adaptive pattern in healthy brain networks that flexibly mobilize neural resources in response to task demands to optimize energy efficiency while maintaining functional performance (48).

In stark contrast to the control group’s hierarchical activation changes, the dysphagia group exhibited markedly limited intragroup activation modulation. No channels showed significantly increased activation during the CST compared with the SST. Likewise, block-average feature comparisons across tasks revealed minimal activation changes, highlighting the restricted compensatory capacity of the dysphagic brain when facing increased task demands.

This severely constrained activation pattern underscores a fundamental deficit in neuromodulation among PSD patients, wherein the brain fails to dynamically adjust activation patterns in response to greater task difficulty due to reduced task adaptability (49), consistent with previous findings demonstrating impaired cortical modulation in stroke patients. The absence of further upregulation under already elevated baseline activation levels likely reflects depleted compensatory reserves (50). Furthermore, impaired neural network plasticity following stroke leads to diminished flexibility and reorganization potential within cortical circuits (25).

In our study, the isolated enhancement observed in R-FPA (CH5) during the dual-task likely represents limited compensatory recruitment, as the right prefrontal cortex attempts to sustain functional output when other cortical regions fail to respond effectively (51). However, such region-specific compensatory activation is insufficient to meet the complex cognitive and motor coordination demands of dual-task conditions, resulting in significant performance deficits in PSD patients (28).

In summary, our findings suggest that PSD patients exhibit impaired task-adaptive neuromodulation, losing the capacity for dynamic cortical regulation in response to changing task demands (52).