Alterations in prefrontal cortical activity in the course of treatment for late-life depression as assessed on near-infrared spectroscopy


*Yasunori Onishi, MD, Department of Psychiatry, Jichi Medical University, 3311-1 Yakushiji, Shimotsuke, Tochigi 329-0498, Japan. Email:


Aim:  To evaluate the severity of depression by measuring alterations in prefrontal cortical activity associated with mood disorders, as assessed on near-infrared spectroscopy.

Methods:  Ten of 27 subjects hospitalized for late-life depression from May 2006 to June 2007, were examined. In these 10 subjects changes in hemoglobin concentration were measured on near-infrared spectroscopy during two types of the rock, paper, scissors game as the cognitive tasks affecting prefrontal cortical activity on 2 days, including 1 day on which depressive symptoms had slightly improved due to treatment, and then on another day >4 weeks later. Severity of depression and cognitive impairment were also simultaneously assessed.

Results:  The change in oxygenated hemoglobin concentration during a difficult task (intentional loss task) was significantly larger than that during an easy task (try to win task) on the left side (left, P = 0.010; right, P = 0.059). On the left side there was a significant negative correlation between the ratio of oxygenated hemoglobin measurements on the second day to those on the first day, and the severity of depression on the second day (left, P = 0.012; right, P = 0.090). Thus, the more left prefrontal cortical activity tended to increase, the fewer depressive symptoms tended to be present on the second day of testing.

Conclusions:  Measuring of alterations in prefrontal cortical activity associated with mood disorders, as assessed on near-infrared spectroscopy, is feasible in subjects with depression.

LATE-LIFE DEPRESSIVE patients often differ clinically from patients with depression that begins early in life in tending to be more somatic and psychotic, and to exhibit more dementia-like symptoms, but are also more disabled, more prone to relapse, and more chronic.1–3 The more severe syndromes of depression, classified as endogenous depression or melancholia, are more common in older individuals. It is theorized that more severe symptoms in late-life depression tend to be associated with a larger number of more severe changes in brain function.

A number of structural and functional neuroimaging studies have suggested the importance of prefrontal cortical changes in mood disorders. Decrease in prefrontal cortical volumes has been reported in structural neuroimaging studies.4,5 In functional neuroimaging studies, using positron emission tomography (PET), single-photon emission computed tomography (SPECT), and functional magnetic resonance imaging (fMRI), the most widely replicated finding in depression is decreased anterior cerebral blood flow and glucose metabolism. These abnormalities have been observed in both unipolar and bipolar depression and appear to be somewhat state-dependent.6–9

In near-infrared spectroscopy (NIRS) studies, attenuated prefrontal activation may also be evident when depressive patients perform cognitive tasks that engage prefrontal cortex.10–14 But few functional neuroimaging methods are available to evaluate the severity of late-life depression. We therefore examined alterations in prefrontal cortical activity as assessed on NIRS in the course of treatment for late-life depression, to elucidate the alterations associated with severity of depression and cognitive impairment.



We consecutively examined 10 of 27 subjects who were older than 60 years of age and who were hospitalized for major depressive disorders at Jichi Medical University Hospital from May 2006 to June 2007 (Table 1). All subjects met the DSM-IV criteria for major depressive disorders. All were right-handed, as determined using the HN Handedness Inventory.15 In these 10 subjects we measured changes in hemoglobin concentration as assessed on NIRS during two types of the rock, paper, scissors (RPS) game as cognitive tasks affecting prefrontal cortical activity on 2 days, including 1 day when depressive symptoms had been slightly improved by medication (eight patients) or modified electroconvulsive therapy (mECT, two patients), and then on another day more than 4 weeks afterwards in the hospital. Only a few studies have used NIRS for a period of approximately 10 min after mECT,16,17 but no such studies have followed up patients undergoing mECT for any significant interval of time. In the present study, for patients treated with mECT, measurements were performed more than 2 weeks after the last treatment to reduce the impact of it on regional cerebral blood flow (rCBF). Disease onset occurred at a mean age of 57 ± 18 years, and six were late-onset depressive patients, for whom onset occurred after 60 years of age. We assumed that effects of habituation did not need to be considered because the mean duration between first and second evaluations was 42 ± 13 days.

Table 1.  Patient characteristics
CaseAge (years)GenderHAMD 1stHAMD 2ndMMSE 1stMMSE 2ndMedications (mg/day) on second day of testing
  1. HAMD, 17-item Hamilton Depression Rating Scale; MMSE, Mini Mental State Examination; hypnotics with short half-lives of elimination are not listed.

 162M12102425Mianserin 50, sodium valproate 200
 271M1462927Paroxetine 20, lithium 600
 370F9121820Amoxapine 75, olanzapine 15
 463F31122930Milnacipran 125, olanzapine 5
 576F1042627Paroxetine 40, milnacipran 50
 668M12102430Maprotiline 65
 780M20142927Milnacipran 50, olanzapine 5
 877F072028Milnacipran 25
 978M2492727Milnacipran 50, lithium 200
1067F962424Mianserin 20, sodium valproate 200
Mean ± SD71 ± 6 14 ± 99 ± 325 ± 427 ± 3 

Although absolute rCBF or metabolism could be reduced by medications, such as benzodiazepines and antipsychotic drugs,18,19 pharmacotherapy was administered based on the judgment of the attending psychiatrists, who were not authors of the present study. A list of medications being taken by the patients at the second examination is given in Table 1. No significant difference was observed in imipramine equivalents of antidepressant dosages between the first and second days of testing.20

The degree of severity of age-related white matter changes on MRI was rated using the revised version of the scale of Fazekas et al.;21,22 seven of 10 were less than grade 1, and three were grade 2. Although NIRS signals can be affected by brain atrophy, because near-infrared light travels differently inside the skull in the presence of this condition, no significant brain atrophy was found in the 10 patients.

Depression was rated using the 17-item Hamilton Depression Rating Scale (HAMD), and cognitive impairment was rated using the Mini Mental State Examination (MMSE) at the same time. Ratings were recorded by one of the authors. Patients were not selected based on any specific educational or sociological background, and all had more than 9 years of schooling.23

All patients were native Japanese, very familiar with authentic RPS, and had played it for a long time. Before the examinations, we confirmed their competence in RPS, and each gave written informed consent to participate in the study. Seventeen of the possible 27 subjects did not consent to undergo the examination. The study was approved by the Ethics Committee of Jichi Medical University.

Near-infrared spectroscopy

Since Jobsis first reported that near-infrared light could be used to measure the state of oxygenation of hemoglobin in the heart and brain of animals non-invasively in 1977, NIRS has been developed as a perfusion-dependent brain mapping modality, similar to fMRI.24–27 NIRS, which can obtain information only for the brain surface, is non-invasive and is easy to use for repeated evaluation of the same subjects. It can be continuously performed for long periods of time, at various places, such as at bedside, and in various physical postures. New methods of use of NIRS have been developed, especially in situations in which PET or fMRI scans are difficult to use, such as for infants or adults during various tasks accompanied by body movement.

We measured changes in oxygenated hemoglobin (oxy-Hb), deoxygenated hemoglobin (deoxy-Hb), and total-Hb, the summation of these, using a 48-channel NIRS imaging system (ETG-4000, Hitachi Medical, Tokyo, Japan) as described in a previous study.28

This machine produced two wavelengths in the near-infrared light spectrum. It has been demonstrated that the 692-nm wavelength is better than a wavelength around 780 nm for pairing with 830 nm to measure the changes in hemoglobin concentration induced by cortical activation. We used these settings (695 nm and 830 nm), which yielded a higher signal-to-noise ratio than those used in previous NIRS studies (such as 780 nm and 830 nm).29 The time resolution was set at 0.1 s. Changes in oxy-Hb and deoxy-Hb were calculated from the difference in light absorption characteristics of these two hemoglobin species based on the Beer-Lambert law. The unit of change was molar concentration multiplied by length (mmol × mm), because the length of the light path is still not precisely known.

The distance between injection and detection probe pairs was set at 3.0 cm, which allowed for cerebral blood flow measurements 2–3 cm below the scalp, spatially coinciding with the cerebral cortex.30,31

The probes of the NIRS machine were fixed using 9 × 9-cm2 rubber shells on the subject's left and right frontal areas, and lateral positioning of them from the midline of the brain was approximately 9 cm. The shells were covered with a nylon-net to keep them attached to the head. Each shell of 16 probes, consisting of a pair of 4 × 4 arrays with eight injectors and eight detectors, was capable of measuring the relative concentrations of hemoglobin at 24 points (Fig. 1). The lowest probes were positioned along the Fp1–Fp2 line in accordance with the international 10/20 system used for electroencephalography.32

Figure 1.

Cortical projection of near-infrared spectroscopy measurement points. The points were mapped onto an anatomical template brain using MRIcro software (MRIcro: developed by Dr Chris Rorden, available at Numbers denote channel numbers for points of measurement.

Cognitive task procedures

In functional brain mapping, various tasks are designed according to how changes in rCBF corresponding specifically to a target function are to be extracted. Task design may thus affect whether extraction is successful or not. Various tasks have long been utilized in PET or fMRI studies, and are similarly used in NIRS.

In recent functional neuroimaging studies with NIRS, the Word Generation Task (WGT) has commonly been used as a cognitive task. But we used the RPS task as reported previously,33 because adequacy of performance of the WGT might vary according to the severity of depression. Adequacy of performance of the RPS task did not depend on patient condition, but did depend on the computer program used. We wanted to maintain steady performance in the course of treatment in order to avoid affecting prefrontal cortical activity.

RPS is often used as a means of selecting who goes first in an activity, like tossing a coin. The game requires at least two players. Each player shapes one hand into a fist for rock, shows a flat palm for paper, or splays the second and third fingers apart for scissors. The two players simultaneously present one of their hands in one of the shapes described above. The relationship between the hands decides the outcome, which may be win, lose, or draw. Rock beats scissors because a rock beats metal; paper beats rock because paper wraps around a rock; and scissors beats paper because scissors cut paper. If the same hands are presented, the result is a draw and the game is replayed until a player wins.

We used a difficult ‘to lose’ RPS task coupled with an easy ‘to win’ RPS task to compare these tasks, because the former requires inhibition of behavior, one of the most important functions of prefrontal cortex. The subjects had to present one of three RPS hands in response to hands displayed on a computer screen, and were required to show hands that lose to the computer, or that beat the computer.

This method involved a one-condition block design, with a combination of control block and stimulation block. After a preliminary 30-s control block, five pairs of stimulation and control blocks followed (e.g. ABABABABABA, control block = A, stimulation block = B). Each task took 5 min 30 s (30 s × 11 blocks = 330 s). The subjects attempted two tasks: deceitful RPS (dRPS), which was deceitful ‘to win’ RPS, and deceitful reverse RPS (drRPS), which was deceitful ‘to lose’ RPS. In the stimulation block of dRPS they were required to show hands that beat the computer, while in drRPS they were required to show hands that lose to the computer. Because they had been conditioned and motivated to win RPS, attempting to lose at drRPS required much more inhibition of action. In the control block of both tasks, the subjects were instructed to show the same hands as the computer. The order of performance of the two types of tasks was randomized, and the games were performed twice on each day.

The RPS presentation program was compiled using the Hot Soup Processor 3.0 software program, a text-based interpreter system for Windows (ONION software,, Japan). The program was controlled not to produce the same hands twice in succession in order to maintain the interest of the subject, that is, to present RPS hands quasi-randomly. This RPS presentation program is freely downloadable from our website (Software Depot for Optical Topography,, Japan).

Because we needed to maintain the interval between computer hand presentations across different tasks and days, we set the subjects' own interval as the shortest interval in which they could perform the exercise of the drRPS task on the first day, and used it during any other tasks on both days. Four of the 10 subjects' own intervals were 2 s (yielding 165 repetitions of RPS per 5 min 30 s), while for four they were 2.5 s (132 repetitions per 5 min 30 s) and for two they were 3 s (110 repetitions per 5 min 30 s).

Data analysis

We analyzed the hemodynamic waveforms of changes in oxy-Hb, deoxy-Hb, and total-Hb in each of 48 channels with ETG4000 Ver.1.61K software (Hitachi Medical, Tokyo, Japan) and MATLAB 6.5.1 (Mathworks, Natick, MA, USA). The waveforms were calculated by averaging the differences between the waveforms for the stimulation and control blocks with the linear compensation method. The grand average waveforms of the three types of hemoglobin concentration changes were obtained for all the subjects, based on each subject's waveforms in all 48 channels. Thirteen noisy channel recordings out of a total of 960 had to be excluded from further analyses (48 channel recordings × 2 days × 10 subjects).

Changes in deoxy-Hb are associated with a slight increase immediately after the task, commonly followed by a rapid decrease. In contrast, changes in oxy-Hb begin after those in deoxy-Hb, the peak of oxy-Hb is much higher than that of deoxy-Hb, and elevated values continue to be observed until the end of the task.34 In animal studies oxy-Hb is the most sensitive indicator of change in rCBF because the direction of change in deoxy-Hb is determined by the degree of changes in venous blood oxygenation and volume.35 We therefore decided to focus on changes in oxy-Hb alone.

The oxy-Hb measurements were made bilaterally by integrating the mean oxy-Hb waveforms of 24 channels between 0 and 30 s, and those during the dRPS task were compared with those during the drRPS task using the paired t-test. Spearman's correlation coefficients were determined to examine the associations among HAMD scores, MMSE scores, differences between the 2 days, the average of two oxy-Hb measurements during drRPS in the same day, the ratio of measurement on the second day to that on the first day, age, and number of repetitions of RPS, which could influence changes in oxy-Hb. We did not use the difference and instead used the ratio of oxy-Hb measurements because NIRS data were influenced by individual path length and anatomical differences. Because two patients (3 and 8) were treated with mECT, we also performed analysis excluding them. For statistical analysis, SPSS 11.0.1 J for Windows (SPSS Japan, Tokyo) was used, with P < 0.05 considered significant.


The 10 subjects could perform both RPS tasks on 2 days in the hospital. The rates of correct performance of dRPS and drRPS were 97.7 ± 1.5% and 95.1 ± 2.5% on the first day and 97.2 ± 2.1% and 95.7 ± 2.0% on the second day. There were no significant differences between rates for the RPS tasks on the first and second days (dRPS, t = −0.535, P = 0.606; drRPS, t = 0.552, P = 0.595), although a significant difference was observed between those for dRPS and drRPS (t = 3.588, P = 0.002).

We produced grand averaged waveforms of changes in oxy-Hb for all the points of measurement from the first and second days, and examined those during dRPS and drRPS tasks on both sides (Fig. 2). Mean oxy-Hb measurements are shown in Fig. 3. The differences between those for drRPS and dRPS were significant on the left side (drRPS, 6.38 ± 6.14; dRPS, 3.28 ± 3.24; t = 2.862, P = 0.010) and marginally significant on the right side (drRPS, 5.73 ± 6.36; dRPS, 3.43 ± 2.89; t = 2.006, P = 0.059).

Figure 2.

Grand averaged waveforms of changes in oxygenated hemoglobin (oxy-Hb) for the 48 measurement points during deceitful rock, paper, scissors (dRPS) and deceitful reverse rock, paper, scissors (drRPS) tasks on the left (L) and right (R) sides. The black line represents oxy-Hb during drRPS, while the gray line represents oxy-Hb during dRPS.

Figure 3.

Mean oxygenated hemoglobin (oxy-Hb) measurements during deceitful rock, paper, scissors (dRPS) and deceitful reverse rock, paper, scissors (drRPS) tasks on the left and right sides. That for drRPS was significantly larger than that for dRPS on the left side (**P < 0.01; n.s., not significant; paired t-test).

On correlational analysis, the negative correlation between the ratio of oxy-Hb measurements and the HAMD score on the second day was significant on the left side (Spearman's ρ = −0.752, P = 0.012), and marginally significant on the right side (ρ = −0.563, P = 0.090; Table 2). On analysis of the eight subjects who were not treated with mECT, there was a similar, significant correlation on the left side (left, ρ = −0.711, P = 0.048; right, ρ = −0.458, P = 0.254). There were no significant correlations between number of repetitions of RPS and oxy-Hb measurements. No other significant correlations were observed between clinical data (HAMD scores, MMSE scores, differences between them on the 2 days, and age) and NIRS data (oxy-Hb measurements on two days and ratio of oxy-Hb measurements).

Table 2.  Spearman's ρ correlations with ratio of oxy-Hb measurements
  1. HAMD, 17-item Hamilton Depression Rating Scale; MMSE, Mini Mental State Examination; *P < 0.05.

HAMD first−0.2260.531−0.0670.854
HAMD second−0.7520.012*−0.5630.090
Difference in HAMD−0.1950.589−0.1650.649
MMSE first0.0680.8510.1550.668
MMSE second−0.1820.615−0.1320.717
Difference in MMSE−0.2900.416−0.4820.159


We examined alterations in prefrontal cortical activity as assessed on NIRS in the course of treatment of late-life depression. Although functional hypofrontality as assessed on NIRS in late-life depression has already been reported,10,14 no reports have described follow-up of prefrontal cortical activity on NIRS over the course of treatment. We used RPS as a cognitive task to maintain steady task performance and task burden.

First, we found that prefrontal cortical activity during the drRPS task was significantly larger than that during the dRPS task in patients with late-life depression, particularly on the left side. We suspect that the most likely reason for the high degree of prefrontal cortical activity during the drRPS task was the need for inhibition of the habitual or stereotyped behavior of trying to win at RPS. An fMRI study of normal subjects showed that activation of the left inferior frontal gyrus was increased during the drRPS task; the authors of that study interpreted this finding as being associated with inhibition of habitual or stereotyped behavior.36 The inferior frontal gyrus has been described as one of the regions most strongly related to Stroop interference, which requires inhibition of habitual and automatic responses, as with the drRPS task.37 The reason why a more extensive region was activated on NIRS than on fMRI may be related to differences in modality or performance rate and the requirement of more complex processing in the drRPS task.

Second, we examined whether prefrontal cortical activity during the drRPS task varied during the course of treatment of late-life depression. A significant negative correlation was found between the ratio of oxy-Hb measurements and HAMD score on the second day on the left side. It appeared that the fewer the number of depressive symptoms noted on the second day, the more the left prefrontal cortical activity tended to increase. Although activation hypofrontality may not be observed in all patients with depression, the findings of this study suggest the possibility that some patients who initially exhibited activation hypofrontality subsequently exhibited improvement with treatment. But the correlation between the ratio of oxy-Hb measurements and the difference in HAMD was not significant. These findings suggest that oxy-Hb measurements do not increase even if the HAMD score decreases greatly, although it remains above a certain threshold. In other words, activation hypofrontality will not exhibit recovery unless depressive symptoms are fully improved. In depressed patients, most PET studies have indicated resting state hypofrontality, particularly in left dorsal prefrontal cortex, with clinical improvement with pharmacotherapy associated with increases in dorsal cortical metabolism.38,39 It has also been found that increased paralimbic metabolism appears to be reversible with effective pharmacotherapy, although NIRS could not measure limbic or paralimbic activity. It has been proposed that disruption of the reciprocal relationship between the dorsal cortical and ventral limbic regions contributes to depressive symptoms.40

Although no differences in rates of correct performance of the drRPS task were found between the first and second examinations, changes in oxy-Hb were larger in the second examination. The improvement of hypofrontality might have been associated with another index of task performance, such as reaction time.

There are several limitations of the present study to be considered in interpreting our findings. It is a disadvantage of NIRS that absolute concentrations of oxy-Hb and deoxy-Hb cannot be calculated. Next, the sample size was small. In addition, anatomical differences among individuals may have affected our results, because differences in NIRS signals might reflect structural rather than functional differences. But in the case of comparison of findings for the same patient, the impact of these factors may be small. There may be more appropriate scales of cognitive impairment and depression. Although NIRS can be performed at the bedside non-invasively, patients are unable to perform any cognitive tasks appropriate for testing while symptoms of depression are severe.

In conclusion, we were able to evaluate prefrontal cortical activity as assessed on NIRS using the rock, paper, scissors game in late-life depression, and obtained findings showing that it may be possible to examine alterations in this activity over the course of treatment using NIRS. The pathophysiology of late-life depression is complex, and development of methodologies to examine it is urgently needed. NIRS may thus be useful for evaluating late-life depression.