Cognitive impairments have been widely recognized as core pathological features of schizophrenia (SCZ) that contribute significantly to functional disability and demonstrating limited responsiveness to interventions (1). Neuropsychological investigations have consistently identified a broad range of cognitive deficits in SCZ encompassing impairments in attention, processing speed, working memory and executive functions (2–4). Longitudinal studies have established that cognitive decline precedes the onset of psychotic symptoms by nearly a decade (5, 6). Emerging evidence suggests accelerated cognitive aging in specific domains among individuals with psychotic spectrum disorders compared to the general populations (7, 8). Nevertheless, the underlying neurobiological mechanisms remain poorly characterized.

Functional near-infrared spectroscopy (fNIRS), a non-invasive optical neuroimaging modality, quantifies cortical hemoglobin dynamics through infrared light absorption variations (9). It is particularly valuable for assessment of regional neuronal activity in the prefrontal and superior temporal cortices- brain regions that have been consistently implicated in the neuropathology of SCZ (10, 11). During verbal fluency tasks (VFT), SCZ patients exhibit attenuated hemodynamic responses in frontotemporal regions (channels 10, 11, 17, 19, 21, 23–32, 34, 36, 38–42, and 44–52) compared to controls (12). Similarly, both individuals at clinical high risk for psychosis (CHR) and patients with first-episode SCZ demonstrated significantly reduced neural activation during the verbal fluency task (VFT), particularly in the bilateral inferior prefrontal gyrus, right middle temporal gyrus, and left dorsolateral prefrontal cortex(lDLPFC) (13). These findings suggest that VFT may serve as an indicator of disrupted neural processing underlying the cognitive deficits associated with SCZ.

Recent studies have demonstrated the diagnostic sensitivity of fNIRS’s across neuropsychiatric disorders. A cross-sectional study examining psychosis-spectrum trajectories (healthy controls, clinical high-risk, first-episode psychosis, chronic schizophrenia) revealed distinct spatiotemporal activation patterns across clinical stages (14). In several studies, researchers found two indexes obtained from fNIRS measured VFT activity, the integral value (IV) and centroid value (CV), which can be used for diagnosis of major psychiatric disorders. In a Tawain cohort examining 192 patients with psychotic disorder, researchers found the frontal CV at 54 seconds enabled the accurate distinction between patients with major depressive disorder at a rate of 80.0% (sensitivity:78.9%, specificity: 72.5%) and those with bipolar MDD (15). Furthermore, a combined fNIRS index of IV and CV achieved sensitivity in classifying individuals with SCZ from healthy controls (16).

In the present study, we used 52-channel fNIRS to measure the prefrontal and temporal cortical hemodynamics during VFT performance, and assessed clinical symptoms and MATRICS Consensus Cognitive Battery (MCCB) in stable SCZ patients and matched controls. The main aim of the study was to validate and extend previous reports by testing the relationships between brain activation during the fNIRS-VFT task and cognitive deficits and to evaluate the reliability of fNIRS as a clinical aid for diagnostic tools in stable schizophrenia. This study may advance the clinical application of fNIRS in assisting the diagnosis of SCZ and provide an objective measure to facilitate research on neural mechanisms research and treatment.

For many years, electroencephalography (EEG), fNIRS, and functional magnetic resonance imaging (fMRI) have been employed to investigate the functional activity of the cerebral cortex. EEG and fNIRS offer several advantages over fMRI in the study of dynamic brain activity: they are portable, cost-effective, and characterized by higher temporal resolution compared to fMRI (25). EEG exhibits high temporal resolution, and it is particularly susceptible to motion artifacts. In contrast, fNIRS provides distinct advantages including tolerance to motion artifacts, and user-friendly operational requirements (26). fNIRS, a state-of-the-art non-invasive brain functional imaging technology to detect the cortical oxygenation activity in real-time, is widely used in psychiatric populations (27, 28).

The VFT task, considered the gold-standard cognitive probe for fNIRS studies, effectively engages in the ability to retrieve and produce verbal material from the lexico-semantic memory, and to test executive control ability, such as processing speed, working memory, response inhibition and tasks requiring cognitive flexibility (29). The VFT performance activates and crucially relies on the superior medial frontal cortex, ventrolateral prefrontal cortex, and anterior temporal lobe (30). Poor performance on the VFT has been consistently linked to impairments in verbal and working memory, as well as executive functioning and problem-solving abilities in mental illness (31). Many studies have reported the brain dysfunction in patients with mental illness during fNIRS-VFT task. In patients with SCZ, reduced brain activation in the prefrontal cortex and superior temporal cortex was reported (32). Individuals with clinical high risk for psychosis demonstrate abnormal brain activation of rSTG (16). Crucially, differences in the performance of the fNIRS-VFT has been used to distinguish various psychiatric disorders. The present study investigated brain activation deficits during the Chinese version VFT using a 52-channel fNIRS device in stable SCZ and healthy controls, as well as evaluated the potential of these brain activation deficits as a clinical aid for diagnostic tools in SCZ.

Compared with HC, our study revealed that stable SCZ patients demonstrated poorer performance in MCCB total scores and cognitive subdomains (speed of processing, attention/vigilance, reasoning and problem-solving, and social cognition). These results may reflect impaired brain inefficiency in SCZ. Our findings align with previous studies demonstrating widespread cognitive dysfunction in SCZ patients across MCCB domains (3, 8). In this study, the SCZ group showed reduced activation values primarily in the mPFC and other regions including rVLPFC, rDLPFC, and lDLPFC relative to controls. Moreover, task-related β values were positive correlated with speed of processing. Our results are in line with previous studies that reduced resting cerebral blood flow and regional glucose metabolism in the PFC of SCZ patients (33). Findings from TMS/hd‐EEG measurements also indicate that intrinsic defects in both activity and connectivity of PFC in SCZ, and that these defects are specifically associated with cognitive impairment (34). Moreover, the PFC plays a critical role in regulating information processing speed and working memory (35). Hypofrontality in DLPFC has been linked to cognitive deficits in SCZ, and neuromodulation using rTMS over DLPFC has shown potential for improving cognitive function (36). Similar results were also found in EGG studies. Compared with controls, distinct functional connectivity patterns involving a wide range of brain regions- including the inferior, superior, and middle temporal gyri, the fusiform gyrus, the superior as well as the middle frontal gyri -were observed in EEG study of patients with chronic SCZ (37). In addition, previous studies have demonstrated that patients with schizophrenia exhibit reduced cortical inhibition in the motor cortex and DLPFC, with these deficits frequently correlating with symptom severity (38, 39).

Specially, a pronounced reduction in IVs in the frontal and bilateral temporal lobes observed during the VFT in the SCZ group, while no significant inter-group differences were noticed in CV values in our study. The IV of the bilateral temporal lobes exhibited positive correlations with processing speed, attention/vigilance, social cognition, and MCCB total scores. In a global ENIGMA study, convergent functional and structural epicenters across all stages of SCZ were predominantly localized within transmodal regions, including the parietal, temporal, and frontal lobes (10). Hypo-connectivity of the default-mode network (DMN) may reflect the impaired integration of inner activity, thereby offering a potential mechanistic explanation for the reduced IV in SCZ (40). Consistent with previous studies, the above results indicate that fNIRS may serve as a valuable clinical tool for assessing cognition in schizophrenia. The fNIRS technology has been widely utilized to detect biomarkers and evaluate cognitive impairments in various mental disorders. For instance, in major depressive disorder, both verbal learning and working memory positively correlated with functional connectivity of the bilateral frontotemporal cortex (41). During the Trial Making Test the Mazes Test, the mean changes in relative oxyhemoglobin concentration demonstrated negative correlations with PFC activation in the SCZ patients (31).

The IV measures the total hemodynamic response during the 60-s activation phase, and this biomarker has demonstrated diagnostic utility in mood disorders, particularly for differentiating depression subtypes (22). In a study involving ultra-high risk for psychosis individuals, first-episode psychosis, chronic schizophrenia, and healthy controls, IV was a significant biomarker for differentiating psychosis spectrum in various clinical stages with AUC range from 0.511 to 0.633 (p<0.05) (14). In a Chinese clinical population, temporal lobe IV was used to assist the diagnosis of major psychiatric disorders with an AUC of an 0.275, and the AUC improved to 0.923 with combined IV and CV (20). The former two studies indicated the combined use of IV and CV could improve the diagnostic sensitivity. In our results, IV of the bilateral temporal lobes alone distinguished patients from healthy controls with an AUC of 0.781(sensitivity 0.633, specificity 0.844). No significant reduction in CV was observed in stable SCZ patients in this study, suggesting preserved cortical homeostasis following task completion in our SCZ cohort. Notably, task-related β value of channel 48 is valuable in diagnosing SCZ from HC with an AUC of 0.762(sensitivity 0.867, specificity 0.533). Integration of the bilateral temporal IV and β value of channel 48 provided an AUC of 0.820. This may preliminarily suggest the high accuracy and reliability of using fNIRS as a clinical aid for diagnostic tools in SCZ.

Our findings not only highlight the distinct neurofunctional signatures in frontotemporal circuitry in SCZ but also validate fNIRS as a non-invasive and practical clinical tool for cognitive assessments. Nevertheless, all findings must be interpreted with caution as this study has several limitations. Deep learning techniques have demonstrated substantial value in fNIRS research aimed at diagnosis of SCZ, particularly in studies involving limited sample sizes (42, 43). In the current study, the SCZ group only included 45 individuals and further studies with deep learning techniques are required to confirm our results. Second, we only observed the cross-sectional outcome between the groups; the longitudinal changes in cognitive function and brain functional connectivity of SCZ are still not fully understood. Third, potential pharmacological confounders persist as patients maintained heterogeneous antipsychotic regimens, disease stages, highlighting the need for subgroup analyses.

In conclusion, our findings suggest that fNIRS may serve as a valuable clinical measure of cognition assessment, and IV of bilateral temporal lobes and β value of channel 48 can be used as candidate biomarkers to differentiate individuals with schizophrenia.