2.1.

Subjects

Seven healthy adults (two men and five women between 28 and $44\text{years}$ old; mean $\text{age}=35$ ) participated in the experiment. Each gave written informed consent after the nature of the experimental procedures was explained before the experiments. Six of the subjects showed right-handedness, and none reported a history of neurological disorder.

2.2.

NIR Topography Measurement

A multichannel NIR topography system (ETG-100: Hitachi Medical Corporation, Japan), which can take measurements at 24 positions, was used. The system delivers light beams with 780- and $830\text{-}\mathrm{nm}$ wavelengths through an optical fiber to the same position simultaneously. The scattered light was detected every $100\mathrm{ms}$ using an avalanche photodiode (APD) $30\mathrm{mm}$ from the incident position through optical fibers. We regarded the midpoint of the source-detector distance as the measurement position because that is where the NIR topography system is most sensitive to changes in chromophore concentration.^{35, 36, 37} Optical fibers carried the light from the sources to the subject’s head and carried the scattered light from the subject’s head to the APD. The average power of each light source was $1.5\mathrm{mW}$ , and each source was modulated at a different frequency to encode irradiated positions and wavelengths. Ten irradiated positions and eight detection positions were arranged to make 24 measurement positions (Fig. 1 ).

Fig. 1

Arrangement of measurement positions in probe patterns over left and right sensorimotor areas centered on locations C3 and C4, respectively.

The measurement positions were determined manually for each subject and were based on the international 10 to 20 system.³⁸ We measured $60\text{-}\times 60\text{-}\mathrm{mm}$ areas in the left and right parietal areas centered on ${\mathrm{C}}_3$ and ${\mathrm{C}}_4$ (Fig. 1). The $60\text{-}\times 60\text{-}\mathrm{mm}$ square was defined as the measurement area for each hemisphere based on the arrangement of optical fibers (irradiation and detection positions). The centers of the bilateral measurement areas were considered to correspond to each primary sensorimotor area based on previous studies examining the relation between the international 10 to 20 locations and the underlying cortical areas.^{39, 40, 41}

2.3.

Task Paradigm

The two experimental sessions for each subject were separated by about $6\text{months}$ — $167\pm 11\text{days}$ $(\text{mean}\pm \mathrm{SD})$ —and two finger-tapping trials were conducted in each session (one with the left hand and one with the right hand) (Fig. 2 ). In each trial, the fingers of one hand were repeatedly placed on the tip of the thumb in the following order: forefinger–second finger–third finger–little finger–third finger–second finger–forefinger. The subjects were asked to repeat the tapping sequence at $3\mathrm{Hz}$ synchronized to the term “Finger tapping” blinking on a CRT monitor. Each task period lasted for $30\mathrm{s}$ and was followed by $30\mathrm{s}$ of rest (rest period). Each trial consisted of six rest periods and five task periods (Fig. 2).

Fig. 2

Schematic diagram of measurement sequence (session and trial): R; rest period; T; task (finger-tapping) period. A finger-tapping task using either the left or right hand was conducted in each trial. The order of trials, right hand or left hand, was counterbalanced among subjects, and the order was fixed in the second session for each subject.

2.4.

Data Analysis

We divided each trial into five $55\text{-}\mathrm{s}$ “blocks.” A block consisted of $5\mathrm{s}$ of the rest period before a task (pretask rest period), the $30\text{-}\mathrm{s}$ task period, and $20\mathrm{s}$ of the rest period after the task (posttask rest period), as illustrated in Fig. 3 . The data collected for each block, i.e., the detected temporal attenuation changes at each wavelength, were baseline-corrected using the data for the pre- and posttask rest periods. The products of the effective optical path length and the concentration changes of the independent hemoglobin ( $\Delta C_{\mathrm{oxy}}^{\prime }$ and $\Delta C_{\mathrm{deoxy}}^{\prime }$ ) were calculated by applying the modified Beer-Lambert law.¹²

Fig. 3

Schematic diagram of analysis parameters for a block. Mean values during $5\mathrm{s}$ of rest period before task (R) and $25\mathrm{s}$ of activation period were used for statistical analysis. Note that activation period can shift from $5\mathrm{s}$ after task onset to $10\mathrm{s}$ after task completion, depending on the maximum absolute value for each case.

The activation signals were statistically assessed. Assuming that the hemodynamic time courses induced by the task varied among the subjects, we defined a $25\text{-}\mathrm{s}$ activation period for each subject (see Fig. 3). We used a within-subject averaged time course over the five blocks to select the $25\text{-}\mathrm{s}$ activation period with the maximum absolute mean change for each hemoglobin signal. The $25\text{-}\mathrm{s}$ activation period was allowed to shift from the initial $5\mathrm{s}$ after task onset to starting $15\mathrm{s}$ after task onset; the earliest period was from $5\mathrm{s}$ after task onset to task completion, and the latest period was from $15\mathrm{s}$ after task onset to $10\mathrm{s}$ after task completion.

Using the mean changes in hemoglobin signals during the pretask period and those during the activation period for each block, we calculated the $t$ value (paired $t$ test) between the mean hemoglobin changes in the pretask rest periods of five blocks and those in the activation periods of the same five blocks.²⁴ We identified measurement positions with significant $t$ values (two-tailed $t$ test, $p<0.1$ ) as activation points by using the independent threshold $(p<0.1)$ for each NIRS signal ( $\Delta C_{\mathrm{oxy}}^{\prime }$ , $\Delta C_{\mathrm{deoxy}}^{\prime }$ , and $\Delta C_{\text{total}}^{\prime }$ ). This statistical analysis was designed to determine the consistency (reproducibility within a session) of changes for the five activation periods. With this analysis, system noise is not detected as an activation because its statistical value does not reach the threshold unless similar changes arise in every (or almost every) activation period. In addition, using activation periods $25\mathrm{s}$ long reduces the possibility of misidentifying an increase or decrease due to spontaneous oscillations⁴² of $0.1\mathrm{Hz}$ . Moreover, a $t$ test using the mean value in the pretask rest period $\left(5\mathrm{s}\right)$ and that in the activation period $\left(25\mathrm{s}\right)$ for each block reduces the effect of high-frequency system noise.

Using the data for the activation points, we examined the reproducibility of spatial information, signal amplitude, and temporal information between sessions for each subject. The spatial information was represented by that for the activation center of the activation area. The coordinates of the activation center $(x_c,y_c)$ were defined as

3.1.

General Aspects of Reproducibility

In the first session, every subject showed a positive $\Delta C_{\mathrm{oxy}}^{\prime }$ , a negative $\Delta C_{\mathrm{deoxy}}^{\prime }$ , and a positive $\Delta C_{\text{total}}^{\prime }$ in the hemisphere contralateral to the tapping hand. Although the same patterns in $\Delta C_{\mathrm{oxy}}^{\prime }$ and $\Delta C_{\mathrm{deoxy}}^{\prime }$ were reproduced in the second session for every subject, for three subjects the $\Delta C_{\text{total}}^{\prime }$ changes in one hemisphere were not significant in the second session. In these three cases, we could not derive the indices for the activation center and so on because no activation point was found there. Reproducibility was therefore further examined using 14 data sets $(\text{two}\text{hemispheres}\times 7\text{subjects})$ for $\Delta C_{\mathrm{oxy}}^{\prime }$ and $\Delta C_{\mathrm{deoxy}}^{\prime }$ , and 11 data sets ( $14\text{minus}$ the three exception cases) for $\Delta C_{\text{total}}^{\prime }$ .

3.2.

Reproducibility of Spatial Information

Figure 4 shows representative activation maps for the left-hand finger tapping for the two sessions of a subject. For both sessions, similar patterns of activation (positive $\Delta C_{\mathrm{oxy}}^{\prime }$ , negative $\Delta C_{\mathrm{deoxy}}^{\prime }$ , and positive $\Delta C_{\text{total}}^{\prime }$ ) were observed in the hemisphere contralateral to the tapping hand. By comparing the activation locations between the two sessions, we found that the activation center was reproduced within 16.0 $\left(\Delta C_{\mathrm{oxy}}^{\prime }\right)$ , 18.6 $\left(\Delta C_{\mathrm{deoxy}}^{\prime }\right)$ , and $16.0\mathrm{mm}$ $\left(\Delta C_{\text{total}}^{\prime }\right)$ on average (Table 1 ). A one-way analysis of variance (ANOVA) for hemoglobin species was conducted using the between-sessions distances of the activation centers. It showed no significant effect of hemoglobin species on the activation center distances between sessions [ $F(2,36)=0.24$ , $p=0.79$ ], which indicates that the hemoglobin species does not affect spatial reproducibility.

Fig. 4

Representative activation maps in two sessions during the left-hand finger-tapping period. Limited positions with significant changes (activation points) are shown in color with linear interpolation, and other measurement positions with no significant change are equally zeroed. Note that the activation centers (“+”) were calculated using the raw coordinates of activation points weighted by the signal amplitude (distance between measurement positions $>21\mathrm{mm}$ ).

Table 1

Mean distance of activation centers between two sessions.

3.3.

Reproducibility of Signal Amplitude

The mean absolute values of $\Delta C_{\mathrm{oxy}}^{\prime }$ , $\Delta C_{\mathrm{deoxy}}^{\prime }$ , and $\Delta C_{\text{total}}^{\prime }$ in the activation periods were used as the signal amplitude for each hemoglobin species. Averaged across subjects, they were $0.095\pm 0.045$ $(\text{mean}\pm \mathrm{SD})\mathrm{mM}\mathrm{mm}$ for $\Delta C_{\mathrm{oxy}}^{\prime }$ , $0.044\pm 0.018\mathrm{mM}\mathrm{mm}$ for $\Delta C_{\mathrm{deoxy}}^{\prime }$ , and $0.067\pm 0.031\mathrm{mM}\mathrm{mm}$ for $\Delta C_{\text{total}}^{\prime }$ . We found a main effect of hemoglobin species on the signal amplitude [ $F(2,78)=17.13$ , $p<0.0001$ ] by one-way ANOVA. A post hoc test [Fisher’s protected least-squares difference (PLSD)] revealed significant differences in every comparison (between $\Delta C_{\mathrm{oxy}}^{\prime }$ and $\Delta C_{\mathrm{deoxy}}^{\prime }$ , $p<0.0001$ ; between $\Delta C_{\mathrm{oxy}}^{\prime }$ and $\Delta C_{\text{total}}^{\prime }$ , $p<0.005$ ; and between $\Delta C_{\mathrm{deoxy}}^{\prime }$ and $\Delta C_{\text{total}}^{\prime }$ , $p<0.05$ ).

The signal amplitudes of each hemoglobin species were compared between sessions for each subject (Fig. 5 ). The signal amplitude was variable within each subject, and no consistent tendency was found across subjects. A one-way ANOVA indicated no significant differences between the two sessions ( $\Delta C_{\mathrm{oxy}}^{\prime }$ : $F(1,26)=1.75$ , $p=0.19$ ; $\Delta C_{\mathrm{deoxy}}^{\prime }$ : $F(1,26)=1.04$ , $p=0.32$ ; $\Delta C_{\text{total}}^{\prime }$ : $F(1,23)=0.15$ , $p=0.90$ ), although the amplitude of the $\Delta C_{\mathrm{oxy}}^{\prime }$ signal tended to be slightly lower in the second session (Fig. 5). In percentage terms, the signal amplitudes varied: on average, $-18\pm 31\%$ $(\text{mean}\pm \mathrm{SD})$ for $\Delta C_{\mathrm{oxy}}^{\prime }$ , $-1\pm 60\%$ for $\Delta C_{\mathrm{deoxy}}^{\prime }$ , and $9\pm 45\%$ for $\Delta C_{\text{total}}^{\prime }$ .

Fig. 5

Signal amplitudes of each hemoglobin signal ( $\Delta C_{\mathrm{oxy}}^{\prime }$ , $\Delta C_{\mathrm{deoxy}}^{\prime }$ , and $\Delta C_{\text{total}}^{\prime }$ ) for each subject in two sessions.

3.4.

Reproducibility of Temporal Information

We examined the reproducibility of temporal information by comparing the time courses for the three activation signals between the two sessions. As shown in Fig. 6 , the time courses for each signal were similar for each subject, with a high correlation coefficient. The mean between-sessions correlation coefficients for the activation point with the maximum signal amplitude are listed in Table 2 .

Fig. 6

Comparison of time courses in two sessions for either hemisphere of each subject. Time courses of $\Delta C_{\mathrm{oxy}}^{\prime }$ , $\Delta C_{\mathrm{deoxy}}^{\prime }$ , and $\Delta C_{\text{total}}^{\prime }$ are shown in left column, middle column, and right column, respectively. Measurement positions were 18 (first session)/18 (second session) (see Fig. 2) for subjects 1, 3, 6, and 7, and $19∕19$ for subject 2, in left-hand finger-tapping session. Measurement positions were $9∕10$ and $6∕9$ for subjects 4 and 5, respectively, in right-hand finger-tapping session. Correlation coefficients (Pearson’s product-moment correlation coefficient) of time courses were as follows: 0.955 $\left(\Delta C_{\mathrm{oxy}}^{\prime }\right)$ , 0.911 $\left(\Delta C_{\mathrm{deoxy}}^{\prime }\right)$ , and 0.945 $\left(\Delta C_{\text{total}}^{\prime }\right)$ for subject 1; 0.975 $\left(\Delta C_{\mathrm{oxy}}^{\prime }\right)$ , 0.935 $\left(\Delta C_{\mathrm{deoxy}}^{\prime }\right)$ , and 0.966 $\left(\Delta C_{\text{total}}^{\prime }\right)$ for subject 2; 0.950 $\left(\Delta C_{\mathrm{oxy}}^{\prime }\right)$ , 0.956 $\left(\Delta C_{\mathrm{deoxy}}^{\prime }\right)$ , and 0.886 $\left(\Delta C_{\text{total}}^{\prime }\right)$ for subject 3; and 0.814 $\left(\Delta C_{\mathrm{oxy}}^{\prime }\right)$ , 0.740 $\left(\Delta C_{\mathrm{deoxy}}^{\prime }\right)$ , and 0.896 $\left(\Delta C_{\text{total}}^{\prime }\right)$ for subject 4; 0.939 $\left(\Delta C_{\mathrm{oxy}}^{\prime }\right)$ , 0.846 $\left(\Delta C_{\mathrm{deoxy}}^{\prime }\right)$ , and 0.880 $\left(\Delta C_{\text{total}}^{\prime }\right)$ for subject 5; 0.951 $\left(\Delta C_{\mathrm{oxy}}^{\prime }\right)$ , 0.785 $\left(\Delta C_{\mathrm{deoxy}}^{\prime }\right)$ , and 0.84 $\left(\Delta C_{\text{total}}^{\prime }\right)$ for subject 6; and 0.835 $\left(\Delta C_{\mathrm{oxy}}^{\prime }\right)$ , 0.812 $\left(\Delta C_{\mathrm{deoxy}}^{\prime }\right)$ , and 0.835 $\left(\Delta C_{\text{total}}^{\prime }\right)$ for subject 7.

Table 2

Mean correlation coefficients (Pearson’s product-moment correlation coefficient) between time courses of two sessions.

4.1.

General Aspects of Reproducibility

The three cases that did not show significant $\Delta C_{\text{total}}^{\prime }$ activation in the second session could have been due to a slight difference in the ratio of $\Delta C_{\mathrm{oxy}}^{\prime }$ to $\Delta C_{\mathrm{deoxy}}^{\prime }$ . The ratio of $\Delta C_{\mathrm{oxy}}^{\prime }$ to $\Delta C_{\mathrm{deoxy}}^{\prime }$ was somewhat unstable between sessions even for the same subject, and this instability sometimes kept the rest-task differences from reaching the threshold of statistical significance. This result is consistent with our previous study²⁴ showing that, for measurements using 31 subjects, the probability of significant activation is less for $\Delta C_{\text{total}}^{\prime }$ than for $\Delta C_{\mathrm{oxy}}^{\prime }$ . These findings suggest that we must analyze the various relationships between $\Delta C_{\mathrm{oxy}}^{\prime }$ and $\Delta C_{\mathrm{deoxy}}^{\prime }$ before using $\Delta C_{\text{total}}^{\prime }$ to evaluate the activation amplitudes.

4.2.

Reproducibility of Spatial Information

Each subject showed similar activation-map spatial patterns between sessions. For each hemoglobin species, the activation center locations determined in the two experimental sessions were reproduced within $20\mathrm{mm}$ (Table 1), which is comparable to the minimum distance between measurement positions in NIR topography $\left(21\mathrm{mm}\right)$ . The high reproducibility of activation centers over time is consistent with the findings of a previous PET study³² that showed consistent anatomical locations of motor activation after $6\text{months}$ , and of an fMRI study³¹ that also showed small distances between the activation centers determined in experimental sessions separated by 30, 49, and $60\text{days}$ . Although these previous studies showed intersession gaps of millimeter order, their results are not comparable with our NIR topography results because of the difference in analytical methods and the spatial resolutions of those methods. For example, the previous fMRI study used $1.17\times 1.17\times 1\mathrm{mm}$ voxels when identifying the coordinates of the maximum peak in the activation area as the activation center,³¹ while the current NIR topography used measurement positions separated by $21\mathrm{mm}$ when identifying the activation center.

In addition, we should keep in mind that the spatial difference between sessions included unavoidable errors caused by the manual method for probe setting. The error in the manual setting was roughly estimated by 10-times repeated setting of the probe holder for one subject and marking of the measurement position each time. We found that the difference among the repeated settings was about $4.8\pm 3.0$ $(\text{mean}\pm \mathrm{SD})\mathrm{mm}$ . This small difference can result in a maximum shift of the activation center of $21\mathrm{mm}$ (the distance between measurement positions) depending on the relationship between the location of the activation area and the probe positions.^{35, 37} Consequently, we can say that the spatial reproducibility of the current NIR topography system is less than $20\mathrm{mm}$ after $6\text{months}$ , including both physiological shifts and technical artifacts. A more accurate method to place the measurement probes, and a probe arrangement with higher density measurement positions^{36, 37} is necessary to distinguish the difference caused by a technical artifact from an actual physiological shift of the activation centers.

4.3.

Reproducibility of Signal Amplitude

Our results showed that the mean $\Delta C_{\mathrm{oxy}}^{\prime }$ amplitude was approximately twice the mean $\Delta C_{\mathrm{deoxy}}^{\prime }$ amplitude, which is consistent with the results of our previous studies.^{12, 20}

The differences in the signal amplitudes between sessions within the same subject showed large variability (SD: 31, 60, and 45% for $\Delta C_{\mathrm{oxy}}^{\prime }$ , $\Delta C_{\mathrm{deoxy}}^{\prime }$ , and $\Delta C_{\text{total}}^{\prime }$ , respectively). In addition, we did not find significant consistency in the difference in signal amplitudes between the two sessions, indicating that habituation has little effect on tasks repeated after $6\text{months}$ . These results suggest variability of the signal amplitude in the NIR topography measurement.

The variability could reflect physiological phenomena and/or technical artifacts. One possible physiological phenomenon is the effect of the baseline differences. The existence of spontaneous oscillation in the hemoglobin oxygenation state has been shown by NIRS studies of adults⁴² and infants,¹⁹ and a difference in the baseline state could affect the activation amplitude, as suggested in previous fMRI and PET studies.^{43, 44}

Technical artifacts may result from small differences in measurement positions. As already described, the measurement position can vary about $4.8\mathrm{mm}$ on average due to manual setting. The difference in measurement positions could affect the signal amplitude, since the current method has uneven spatial sensitivity due to the relationship between the activation area (location and size) and the probe positions.³⁷ Although the tendency of a lower $\Delta C_{\mathrm{oxy}}^{\prime }$ in the second session might indicate a habituation effect similar to the one shown in a previous fMRI study,³¹ where habituation effects for sensorimotor activation were observed between two sessions separated by $5\mathrm{h}$ and by 1 or $2\text{months}$ , it would be difficult to estimate the activation level by using the signal amplitude in current NIRS systems.

4.4.

Reproducibility of Temporal Information

We found high correlation coefficients for the time courses in repeated sessions for the same subject (Fig. 6). Although we know of no other NIRS studies reporting such a similar time course reproduced after more than a few days, our findings could be important with regard to the development of new NIRS applications in the clinical area. The qualitative characteristics of the subject might be more clearly inferred from temporal information, such as the time course of the activation, than from spatial information and amplitude information, which vary with the experimental conditions. A recent study used the characteristics of the NIRS time course $\left(\Delta C_{\mathrm{oxy}}^{\prime }\right)$ to assess patients exhibiting depression and schizophrenia.²² Moreover, examining the shape of the time course in the same subject can be used to evaluate the subject’s state, which would help in longitudinal studies as well as in monitoring rehabilitation effects.