Innov Clin Neurosci. 2025;22(4–6):11–13.

Dear Editor:

Approximately 15 years have passed since functional near-infrared spectroscopy (fNIRS) was approved for use as an auxiliary diagnostic tool for the differential diagnosis of depressive state in psychiatry by the Ministry of Health, Labor and Welfare of Japan.¹ Distinctive waveform patterns appearing in the prefrontal cortex during a verbal fluency task (VFT) have been used to differentially diagnose major depressive disorder (MDD), bipolar disorder (BD), and schizophrenia.² However, current fNIRS-based psychiatric diagnostic aids for hospitalized patients with depression might not be able to effectively distinguish between MDD and BD. For instance, our recent report investigated the agreement between fNIRS diagnoses and final psychiatric diagnoses one year after fNIRS in 241 individuals hospitalized for depression. The findings demonstrated that current fNIRS methods do not necessarily function as diagnostic adjuncts. The concordance rate between fNIRS diagnosis and psychiatric diagnosis was 38.2 percent for individuals with MDD and 44.0 percent for individuals with BD, indicating that more than half of individuals with depression were initially classified as having a different disorder by fNIRS.³ Multivariate logistic regression analysis showed that higher serum sodium concentrations in individuals with MDD and lower serum sodium concentrations and higher antidepressant doses in individuals with BD contributed to the diagnostic discrepancy.³ These results suggest that serum electrolytes and antidepressant doses might affect fNIRS waveforms. 

The fNIRS evaluates brain activity based on neurovascular coupling, in which changes in neural activity and blood flow are closely related. Neurovascular coupling is maintained by numerous mediators, including ion channels, nitric oxide, and astrocytes. Electrolytes have been demonstrated to influence vascular smooth muscle contraction and the function of neurovascular coupling through the sodium-potassium-ATPase pump, exhibiting differences between sexes. In female subjects, serum sodium concentration showed a significant positive correlation with the temporal integral value, while in male subjects, serum potassium concentration showed a significant negative correlation with the frontal initial value.⁴ Antidepressants affect neurovascular coupling through the regulation of astrocytic calcium by the neurotransmitters, including serotonin and noradrenaline.⁵ Electrolytes, antidepressants, and sex might affect neurovascular coupling and alter fNIRS waveforms. 

Recent studies using functional MRI (fMRI) have demonstrated that large-scale brain networks of MDD and BD show distinct patterns.⁶ Individuals with MDD are characterized by hyperactivation of the default mode network (DMN), which is associated with self-referential thinking and internal thought processes, and hypoactivation of the anterior cingulate cortex (ACC) and central executive network (CEN), which are involved in emotion processing and cognitive functions.⁷ Individuals with BD are characterized by hypoactivity of the DMN, including the medial prefrontal cortex (mPFC), which is associated with difficulties in focusing attention and exponential thinking, and hyperactivity of the ACC, which might be related to mood fluctuations and impulsivity.⁸ Therefore, measuring the DMN is effective for distinguishing between MDD and BD. However, it is difficult to measure the DMN using fNIRS with verbal fluency task (VFT). The mPFC, a pivotal component of the DMN, offers a promising avenue for distinguishing between MDD and BD through fNIRS-based measurement. Individuals with BD and healthy subjects could be differentiated by converting frontal lobe activity obtained by fNIRS with VFT into spectrograms with convolutional neural networks (CNN), with an area under the receiver operating characteristic curve of 0.94, and areas including the mPFC were important for the differentiation.⁹

Alternatively, given the persistent presence of the DMN, even in the absence of task-related activity, it might be essential to explore the potential of resting-state fNIRS as a means to differentiate between MDD and BD, as opposed to relying on undertask fNIRS methods. Diffuse optical topography analysis of fNIRS data revealed the DMN in spontaneous deoxygenated and oxygenated hemoglobin changes that are strikingly similar to the fMRI networks obtained from the same subjects,¹⁰ and thus it might be possible to infer the DMN with fNIRS. 

The utilization of fNIRS in the differential diagnosis of psychiatric disorders must progress toward more advanced methods to enhance diagnostic precision. Although fNIRS has the disadvantage of low spatial resolution, it has the advantage of being simpler and easier to apply clinically than fMRI. The relationship between fNIRS waveforms and diagnosis is pattern recognition, and it is possible to improve diagnostic accuracy by using machine learning. Resting-state fNIRS is more effective at detecting the DMN than task-based fNIRS and could be useful as a disease-specific trait marker for the differentiation of psychiatric disorders. The development of an algorithm that can differentiate between MDD and BD is predicated on the accumulation of datasets comprising psychiatric diagnoses derived from the Structured Clinical Interview for the Diagnostic and Statistical Manual of Mental Disorders, fifth edition, research version (SCID-5-RV) and resting-state fNIRS waveforms. This algorithm is to be created through the utilization of supervised machine learning methodologies. The diagnostic accuracy of fNIRS should be improved by developing new analysis methods, rather than using the standards that have been used for the past 15 years.

With regards,

Takahiko Nagamine, MD, PhD 

Dr. Nagamine is with the Department of Psychiatric Internal Medicine, Sunlight Brain Research Center in Hofu, Japan.

Funding/financial disclosures. The author has no conflicts of interest relevant to the content of this letter. No funding was received for the preparation of this letter.

Ethical statement. The study was approved by the Ethics Committee of the Sunlight Brain Research Center under approval number SBRC-505. The present study did not utilize novel datasets; rather, it was based on data from extant papers. Consequently, patient informed consent for each dataset was performed in the original paper.

Correspondence. Takahiko Nagamine, MD, PhD; Email: tnagamine@outlook.com

References

1. Takizawa R, Fukuda M, Kawasaki S, et al. Neuroimaging-aided differential diagnosis of the depressive state. Neuroimage. 2014;85 Pt 1:498–507. 
2. Koike S, Nishimura Y, Takizawa R, et al. Near-infrared spectroscopy in schizophrenia: a possible biomarker for predicting clinical outcome and treatment response. Front Psychiatry. 2013;4:145. 
3. Nakamura M, Nagamine T. Current status and problems of neuroimaging-guided diagnosis using near-infrared spectroscopy in Japan. Int J Neuropsychopharmacol. 2025;28(Suppl 1): i182–i183
4. Nakamura M, Nagamine T. Serum electrolyte levels may be associated with prefrontal hemodynamic responses in near infrared spectroscopy. J Near Infrared Spectrosc. 2018;26(4):229–234. 
5. Renden RB, Institoris A, Sharma K, et al. Modulatory effects of noradrenergic and serotonergic signaling pathway on neurovascular coupling. Commun Biol. 2024;7(1):287. 
6. Zovetti N, Rossetti MG, Perlini C, et al. Default mode network activity in bipolar disorder. Epidemiol Psychiatr Sci. 2020;29:e166. 
7. Siegel-Ramsay JE, Bertocci MA, Wu B, et al. Distinguishing between depression in bipolar disorder and unipolar depression using magnetic resonance imaging: a systematic review. Bipolar Disord. 2022;24(5):474–498. 
8. Yang Y, Cui Q, Lu F, et al. Default mode network subsystem alterations in bipolar disorder during major depressive episode. J Affect Disord. 2021;281:856–864. 
9. Alıcı YH, Öztoprak H, Rızaner N, et al. Deep neural network to differentiate brain activity between patients with euthymic bipolar disorders and healthy controls during verbal fluency performance: a multichannel near-infrared spectroscopy study. Psychiatry Res Neuroimaging. 2022;326:111537. 
10. Zhang F, Khan AF, Ding L, et al. Network organization of resting-state cerebral hemodynamics and their aliasing contributions measured by functional near-infrared spectroscopy. J Neural Eng. 2023;20(1):016012.