Subjects were patients with depression who fulfilled the diagnostic criteria for a major depressive episode in the Diagnostic and Statistical Manual of Mental Disorders, 4^th edition [8] and whose symptoms were evaluated using the Hamilton Rating Scale of Depression (HRSD, 24-item) [9] . Thirty-two subjects participated in this study. They received pharmaceutical therapy with psychotropic agents, but the types of drugs and their doses were not controlled. Patients suspected of having a serious physical disorder or who had clear abnormalities on magnetic resonance imaging (MRI), X-ray computed tomography (X-CT), electroencephalography (EEG), or an intelligence test (WAIS-R) were excluded. Patients whose HRSD scores did not drop by 30% or more 3 months after study initiation were also excluded. Thirteen of the 32 (40.6%) subjects had a positive response and participated in the remainder of the study. After 6 months, their HRSD score was less than 10 points. Thirteen patients who met the above criteria were enrolled (9 female, 4 male) to complete the study. We also invited healthy control subjects to the present study who had no physical or mental disorders or history of treatment for mental disorders; individuals in this group were not taking any medications. They had no abnormalities on MRI, X-CT, EEG, or intelligence tests. The control group consisted of 14 subjects (7 female, 7 male). This study was conducted with the consent of the Fukushima Medical University Ethics Committee (No. 133). Written consent for participation in the study was obtained from all subjects.

SPECT was used in rCBF distribution measurements, with ^99mTc-HMPAO (hexamethylpropyleneamineoxime) 1110 MBq as a radioactive agent given in an intravenous bolus injection via a line secured in the median antebrachial vein. Before and after the injection, the subjects rested in a quiet room with the lights dimmed while wearing an eye mask and ear plugs. Imaging was performed 40 min after injection. Because depressive symptoms such as psychomotor retardation in major depression patients show considerable diurnal variation, ^99mTc- HMPAO injection was given between 10 and 11 a.m. SPECT imaging was performed three times, during the depressive phase before treatment (baseline) and 3 and 6 months later. Depressive symptoms were evaluated with the 24-item HRSD within 4 days of each SPECT study.

A Shimadzu Headtome 030 (ring type, HR collimator) was used in SPECT imaging, and image slices were parallel to the orbitomeatal base line (OM line). The obtained images were 64 × 64 pixels, and slice thickness was approximately 13 mm.

¹²³I and ^99mTc are the most commonly used isotopes for imaging cerebral blood flow with SPECT. Absolute cerebral blood flow can be calculated by collecting arterial blood when using ¹²³I; however, this is not possible when using ^99mTc. We calculated rCBF as a relative value based on the cerebellum as the reference region because cerebellar blood flow is not considered to be affected by factors such as age [10] - [15] . Using the whole brain as the reference region produces inherent contradiction because the target regions are within the reference region.

As a reference site for each imaging, the mean count for the entire cerebellum including the vermis on slices that contained the lower-central pons was calculated, and the value obtained through each pixel count divided by this value was defined as the ratio to the cerebellum used in the analysis.

Nine slices were imaged, but the data were rearranged to form 20 slices with resampling (3-dimensional linear method) and expanded (normalized) to a standard brain. They were then compared voxel by voxel using SPM-99 (Statistical Parametric Mapping 99). The significance levels for the intra- and intergroup comparisons were set as p < 0.005 and p < 0.001, respectively.

Measurements at baseline, 3 months, and 6 months were compared in the subject group using paired t-tests (p < 0.005), and between-group comparisons with the healthy control group were performed with two-sample t-tests (p < 0.001). The cluster size threshold was set to greater than 40 voxels for the intra- and intergroup comparisons. Image processing was performed with Medex 3.3 (Sensor Systems, Inc.). From this analysis with SPM-99, the anatomical positions of each cluster and Brodmann area were identified using the Talairach atlas [16] for locations in which differences or changes were seen. As an indicator of differences between groups on SPM, the Z score ([mean of control group-mean of subject group]/standard deviation of control group) was calculated.

The analyses of other variables such as age, education, and HRSD were conducted using t-tests.

A total of 13 patients (9 female, 4 male) who met the criteria were enrolled. They had a mean age of 46 ± 13 (22 - 62) years and 11.5 ± 1.7 years of education. The control group consisted of 14 subjects (7 female, 7 male) with a mean age of 31.5 ± 12.8 years and had 15.4 ± 3.0 years of education. The control group was thus significantly younger (t (24.31) = 2.952, p = 0.007) and had significantly more years of education (t (20.48) = 4.099, p < 0.01). Depressive symptoms in terms of HRSD scores were 31.5 ± 12.0 at baseline, 11.0 ± 4.3 at 3 months, and 6.2 ± 2.7 at 6 months. The mean score had decreased by 65% or more at second and third measurements, and these differences were significant (baseline vs. 3 months, t (12) = 6.187, p < 0.01; baseline vs. 6 months, t (12) = 8.415, p < 0.01). Table 1

a. The doses (mg/day) of antidepressants, antipsychotics, and antianxiety medications are expressed in imipramine, chlorpromazine, and diazepam equivalents, respectively. Equivalents were calculated based on the conversion table formulated by the clinical psycho-pharmacology study group of Keio University [17] . Avg, average; Std, standard deviation.

shows the antidepressant, antipsychotic, and antianxiety drug amounts in equivalent dosages of imipramine, chlorpromazine, and diazepam, respectively, that were taken by patients at each time point. The conversion in equivalent dosages of imipramine, chlorpromazine, and diazepam was calculated based upon the conversion table formulated by clinical psycho-pharmacology study group of Keio University [17] . The mean medication doses per day, in imipramine, chlorpromazine, and diazepam equivalents, were not significantly different between the three time points.

We found that rCBF values in the superior temporal gyrus (temporal pole), cingulate (frontal portion), and inferior frontal gyrus were consistently lower in patients with major depression compared to controls throughout the 6-month treatment period (Table 2). In terms of symptoms, at least 6 months of treatment is considered necessary for full recovery from depression [7] . During the 6-month period, rCBF of the patients consistently increased in the superior and middle frontal gyri (Table 3). At 6 months, patients showed higher rCBF ratios than the controls in the middle temporal gyrus, superior temporal gyrus, fusiform gyrus, middle occipital gyrus, lingual gyrus, and cuneus (Table 2).

Baseline rCBF in our patients was lower than that in controls in a wide range of the frontal regions. While depressive symptoms were significantly improved at 3 months, rCBF ratios in the superior and middle frontal gyri increased. However, rCBF ratios in the middle and inferior frontal regions remained low. At 6 months, depressive symptoms continued to improve, and rCBF ratios increased in the superior frontal gyrus, middle frontal gyrus, cingulate (frontal portion), precuneus, fusiform gyrus, lingual gyrus, and cuneus. Conversely, rCBF ratios decreased in the inferior frontal gyrus, superior temporal gyrus (temporal pole), middle temporal gyrus, and superior temporal gyrus.

In the present study, rCBF increased in the left middle and superior frontal gyri after 3 months, and those on the right side increased more after 6 months. rCBF decreased in the superior temporal gyrus (temporal pole) including the amygdala after 6 months. It remained unchanged in the cingulate gyrus after 3 months but increased after 6 months. Thus, there were temporal differences in brain activity changes represented by rCBF over the course of treatment.

A period of at least 6 months is thought to be required to recover from a major depressive episode [7] . Previous studies on rCBF in depression compared patients with healthy controls at a single time point or assessed rCBF changes before and after treatment. There have been some observations performed at three or more time points in the course of treatment, but all of these were case reports. The present study investigated changes in symptoms and rCBF at baseline and 3 and 6 months after treatment, which allows us to provide more accurate information regarding rCBF changes following treatment.

There have many previous studies on rCBF using SPECT in patients with major depression [1] - [6] . To our knowledge, three have compared patients with healthy controls [1] [3] [18] . Oda et al. used ^99mTc-ECD and reported lower rCBF in the frontal and temporal lobes and anterior cingulate gyrus in patients [3] . In line with this, Bonne and colleagues used ^99mTc-HMPAO and reported that rCBF was lower in the right parietal, occipital, and left temporal lobes in patients compared with controls [1] . Interestingly, Pagani et al. also used ^99mTc- HMPAO and found relatively higher rCBF in the right temporo-parietal cortex, bilateral prefrontal and frontal cortices, and anterior temporal cortex [18] .

Both Arthur et al. and Mayberg et al. conducted comparative studies with healthy controls and patients with major depression using positron emission tomography (PET) [10] [19] . Mayberg et al. reported higher CBF in the subgenual cingulate and lower CBF in the prefrontal, premotor, dorsal anterior cingulate, and anterior insular regions in patients with major depression compared with controls using ¹⁵O as a tracer [19] . Arthur and colleagues used fludeoxyglucose (FDG) and found higher glucose metabolism in the prefrontal cortex, caudate, and thalamus and lower glucose metabolism in the temporal lobe of patients compared to controls [4] . Drevets observed high blood flow in the amygdala before treatment [5] .

In summary, the previous reports are inconsistent regarding some areas including the prefrontal, frontal, temporal, and parietal cortices. Among them, the temporal cortex is consistently reported to show reduced rCBF in depression, with three studies reporting lower rCBF in patients [1] [3] [4] , but discrepant results were reported by Pagani et al. In contrast, the anterior cingulate is consistently reported to have lower rCBF in patients than in controls [3] [19] . The premotor, anterior insula, and occipital cortex are reported to show lower rCBF, while the subgenual cingulate, caudate, thalamus, and amygdala display higher rCBF in depression than controls [4] [5] [19] .

Davies at al. conducted a SPECT study with ^99mTc-HMPAO that assessed the effect of treatment on rCBF in patients with major depression [2] . They reported increased rCBF in the bilateral thalamus and decreased rCBF in the right posterior temporal, left occipital lobe, and cerebellum following venlafaxine treatment. Three groups show changes in PET tracer levels following treatment [4] [19] [20] . Arthur et al. reported normalization of FDG in the prefrontal cortex, caudate, and thalamus to levels comparable with healthy control group following various treatments [4] , and Mayberg et al [19] . reported normalization of rCBF to levels comparable with controls in the subgenual cingulate and prefrontal cortex of treated patients. Mayberg [20] also reported that treatment elevated blood flow in the anterior cingulate gyrus. Kennedy et al. investigated brain metabolism using PET before and after 16-week treatment for major depression [21] . They found reduced brain metabolism in the ventrolateral prefrontal cortex.

At the start of treatment, the left and right superior temporal gyrus (temporal pole) had low rCBF in the patient group (Table 2). The amygdala is thought to be included in the temporal pole. A further decrease in rCBF of the left superior temporal gyrus (temporal pole) was seen after treatment (Table 3). Our observation of treatment-related decrease in rCBF in the left temporal pole including the amygdala has not been reported in previous SPECT studies of patients with depression. However, fMRI [22] and PET [20] [23] studies have described reduced activity in the left amygdala of patients with depression after treatment. This suggests a relationship between strong negative feelings and high amygdalar activity. Consequently, the treatment-related decrease in rCBF of the left amygdala is an intriguing result and may hint at the amygdala’s pathophysiologic role in depression.

In the present study, the different brain areas of depressed patients demonstrated two different patterns of blood flow change during recovery from depression. In the superior and middle frontal gyri and cingulate, blood flow increased, while it decreased in the superior temporal gyrus and temporal pole including the amygdala (Table 3). Interestingly, these areas are all associated with emotion. This suggests that different brain areas related to emotions would are differentially involved in the pathogenesis of depression.

Previous brain imaging studies on depression can be summarized in that there were two common findings after successful treatment: 1) increased metabolism in the DLPFC and 2) decreased metabolism in the amygdala [5] [19] [21] . However, a major problem with these previous studies is that none found both increased and decreased blood flow changes in the same subject group. The present study is unique in that we found opposite changes in blood flow in a single group of patients during the course of treatment. To our knowledge, the findings of 1) a relatively early increase in blood flow in the superior and middle frontal gyri corresponding to the DLPFC and 2) a relatively late decrease in the amygdala have not been previously reported.

In the present study, areas in which higher rCBF was observed in patients compared to controls at 3 months included the superior temporal gyrus, inferior temporal gyrus, middle temporal gyrus, middle occipital gyrus, and fusiform gyrus. At 6 months, rCBF increased more in these areas (Figure 2). The fusiform and superior temporal and middle temporal gyri are part of the social cognitive neural network [24] [25] . Therefore, it is speculated that multiple brain regions related with social cognition are differentially activated during recovery from a depressive state. Social cognition is thought to be involved in deficits associated with depression, which are expected to recover with clinical improvement. The recovery process from disordered social cognition is speculated to be gradual because the recovery processes of the neural substrates of social cognition are gradual, as shown in the above-cited results.

The mean ages of the control and patient groups were 31.5 ± 12.8 and 46.3 ± 13.2 years, respectively, which were significantly different. Therefore, the effect of age is considered a potential problem. The literature contains several reports on the influence of age on CBF flow in healthy adults. Ances et al. [26] reported that the mean elevation rate of CBF and cerebral metabolic rate of O₂ in response to visual stimuli was the same in 10 subjects with a mean age of 28 years and 10 with a mean age of 53 years; thus, age did not seem to affect the response to stimuli [26] . Giovacchini et al. reported no differences in whole gray-matter CBF among eight people with a mean age of 27 years and seven subjects with a mean age of 65 years after partial volume correction [27] . Similarly, Meltzer et al. [28] reported no difference in cortical CBF between groups aged 19 - 46 and 60 - 76 years after partial volume correction. These reports suggest that there is no significant age-dependent difference in rCBF in healthy adults.

Nagafusa et al. assessed the effect of age on rCBF in patients with depression using SPECT imaging [29] . They did not find an effect of age on rCBF decreases in the depression group; rather, the reductions were common to all ages, specifically in subjects from 30 to 79 years old. Based on the above-cited studies, we tentatively conclude that the age difference between the two groups in the present study did not critically distort a real difference between depressive patients and healthy control subjects.

Actually, Terada et al [30] compared rCBF in old patients with depression, very old patients with depression, and healthy controls. They found differences in the frontal lobe, thalamus, and entire cerebrum, but the differences in the cerebellum were not statistically significant. If an age difference of rCBF exists at all, its effect should be cancelled, because rCBF is expressed was the standardized ratio of cerebrum/cerebellum in the present study.

The following points should be considered when interpreting the present results. The number of patients was limited, and the control group was significantly more educated than the patient group. Frontal lobe function is reported to have a general correlation with years of education [31] , but there are also reports of negative correlation between frontal lobe activity and years of education in some young people and a positive correlation in elderly subjects [32] . Thus, we cannot rule out an effect of education on frontal lobe function between the patient and control groups.