Shared and disorder-specific resting-state neural activity characteristics in patients with anorexia nervosa and bulimia nervosa

Abstract

Background

Anorexia nervosa (AN) and bulimia nervosa (BN) are two primary subtypes of eating disorders (ED), often presenting with overlapping clinical features that complicate diagnosis. Despite shared symptoms, the underlying neural mechanisms of two subtypes remain incompletely understood. Delineating both shared and unique neural alterations may support biomarker discovery and inform targeted interventions.

Methods

We recruited 28 patients with AN, 26 with BN, and 31 matched healthy controls (HC), aged from 14 to 40 years old. Resting-state functional magnetic resonance image (Rs-fMRI) data were acquired to investigate alterations in spontaneous brain activity. Four voxel-wise metrics were analyzed: amplitude of low-frequency fluctuation (ALFF), fractional amplitude of low-frequency fluctuation (fALFF), regional homogeneity (ReHo), and degree centrality (DC). Symptom severity was assessed using the Eating Disorder Examination-Questionnaire (EDEQ), which includes four subscales: Eating concern (EDEQ_E), Shape concern (EDEQ_S), Weight concern (EDEQ_W), and Restraint (EDEQ_R). Pearson correlation analysis was used to examine associations between altered imaging metrics and clinical variables.

Results

Both AN and BN exhibited convergent alterations, including reduced activity in the bilateral middle frontal gyrus (MFG), insular cortex (INS), superior temporal gyrus (STG), and left parahippocampal gyrus (PHG), alongside increased activity in the bilateral striatum, middle occipital gyrus (MOG), and cerebellum. Disorder-specific alterations in AN included increased activity in the right striatum and right precuneus, increased DC in the right superior frontal gyrus (SFG), and decreased fALFF and DC in the left calcarine. In contrast, patients with BN exhibited elevated fALFF in the right precentral gyrus (PCG_R) and increased DC in the right calcarine. Correlation analyses revealed negative association between the ReHo value of the MOG_L and EDEQ, and positive associations between the DC value of the PCG_R and EDEQ and EDEQ_E in patients with BN.

Conclusion

Our findings revealed both shared and diagnosis-specific alterations in intrinsic brain activity within the cortico- striatal-limbic circuit, underscoring its role in the pathophysiology of ED.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40337-026-01559-0.

Plain language summary

Anorexia nervosa (AN) and bulimia nervosa (BN) are the two primary subtypes of eating disorders (EDs) and often present with overlapping clinical features that complicate diagnosis. This study investigated the neural mechanisms underlying each subtype to identify both shared and disorder-specific neural alterations that may support biomarker discovery and inform targeted interventions. In this study, we examined brain activity in 28 patients with AN, 26 with BN, and 31 healthy controls. We used resting-state functional MRI to assess alterations in spontaneous brain activity, including neuronal activity, synchronization of local activity, and coordination across the whole brain. Our findings revealed both shared and diagnosis-specific alterations in intrinsic brain regions within cortico-striatal-limbic circuit, a network involved in cognitive processing, interoception, reward processing, and body image perception. We also found that some brain activity changes were related to symptom severity. These results improve our understanding of the biological mechanisms underlying eating disorders and may help guide the development of more personalized treatments in the future.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40337-026-01559-0.

Introduction

Anorexia nervosa (AN) and bulimia nervosa (BN) are two primary subtypes of eating disorders (ED), sharing overlapping symptomatology yet displaying distinct behavioral and cognitive profiles [1]. While AN is characterized by severe restriction of food intake or excessive exercise to achieve low body weight, accompanied by an intense fear of weight gain, and some patients may experience intermittent binge eating followed by compensatory behaviors, BN primarily involves repeated episodes of binge eating followed by purging and overconcern with body weight or shape [2]. ED is especially prevalent among adolescents and young adults, with a significantly higher incidence in females, with lifetime prevalence of 2% to 5% [3]. Notably, a global rise in ED prevalence has been documented following the COVID-19 pandemic, including among Chinese populations [4, 5]. Both disorders are frequently comorbid with psychiatric and somatic comorbidities, contributing to increased mortality, and imposing a substantial burden on patients, families, and healthcare systems [6, 7].

Neuroimaging has emerged as a promising tool for elucidating the neurobiological substrates of ED by investigating structural and functional brain alterations, facilitating the identification of reliable neuroimaging biomarkers for precise diagnosis, and subtype classification which was currently relied on self-reported symptoms [8]. Brain structural studies have revealed abnormal gray matter volume (GMV) or cortical thickness (CT) in frontal-striatum circuit, limbic cortex, parietal, temporal and occipital cortex, and cerebellum (CBL) in ED patients [9–11]. Many task-based neuroimaging studies also confirmed altered activity in these widespread cortical-subcortical regions in different task paradigms including reward, cognitive, emotional, and body-image tasks [12, 13], such as during the processing of food stimuli, AN is associated with hyperactivation in the prefrontal cortex, anterior cingulate cortex (ACC), precuneus (PRU), and CBL, and hypoactivation in the amygdala, striatum (STR), insular (INS), thalamus, parahippocampal gyrus (PHG), hippocampus, and parietal cortex in response to food stimuli, and BN is linked to hyperactivation in the STR, INS, visual cortex, and CBL, and hypoactivation in the prefrontal and ACC, superior temporal gyrus (STG), PRU, and temporal cortex [14]. However, these findings remain heterogeneous and inconclusive. For example, one study examined patients with AN found that the volumes of the ventromedial prefrontal cortex (vmPFC), ACC, and INS were positively correlated with symptom severity [15], whereas another study reported that the CT in the middle frontal gyrus (MFG), superior frontal gyrus (SFG), orbitofrontal cortex (OFC), and temporoparietal regions were negatively correlated with symptom severity [16]. Such discrepancies may stem from variability in imaging paradigms, sample sizes, task designs, and methodological approaches. Despite this inconsistency, these regions are consistently linked to ED pathophysiology. A systematic review concluded that restrained eating has been linked to increased GMV in the PHG, inferior parietal lobule (IPL), PRU, temporal cortex, SFG, and MFG. In contrast, disinhibited eating has been associated with increased GMV in the OFC and decreased GMV in the middle cingulate cortex (MCC). Body image disturbance (BID), another hallmark of ED, has been linked to GMV reductions in the superior temporal gyrus (STG), while compensatory behaviors and preoccupations with body weight and shape have shown associations with the STR and INS, respectively [17].

Resting-state functional magnetic resonance imaging (Rs-fMRI) studies enable the exploration of spontaneous brain activity in natural states without task interference. Several voxel-wise metrics have been developed to characterize intrinsic brain function in Rs-fMRI. The amplitude of low-frequency fluctuations (ALFF) reflects the intensity of spontaneous neural activity within the 0.01–0.08 Hz frequency range [18], while fractional ALFF (fALFF) represents the proportion of low-frequency power relative to the full frequency spectrum, offering improved specificity to neural signals [19]. Regional homogeneity (ReHo) assesses local synchronization of BOLD signals using Kendall’s coefficient of concordance (KCC) [20]. Degree centrality (DC) quantifies the global functional connectivity (FC) of each voxel by evaluating its centrality in a whole-brain network [21]. These measures complement one another by capturing different dimensions of brain function. ALFF and fALFF focus on signal amplitude (frequency domain), whereas ReHo and DC emphasize temporal coherence and network integration (time domain). Their combined use enables a comprehensive assessment of both regional and large-scale brain dysfunction and holds promise for identifying neuroimaging biomarkers across neurological and psychiatric disorders [22, 23].

Despite growing interest in FC in ED, which is often explored using region-of-interest or whole-brain analyses [24–26]. Comparatively, fewer studies have examined local spontaneous activity using Rs-fMRI. Prior research reported increased fALFF in the right hippocampus and left SFG, along with decreased fALFF in the left rectus gyrus, and left middle occipital gyrus (MOG) in AN, with some associations with body mass index (BMI) [27]. Another study reported increased fALFF in the hippocampus, PHG, and bilateral occipital cortex, as well as increased ReHo in the thalamus, caudate, STG, INS, and occipital cortex, and decreased fALFF in the bilateral MFG, decreased ReHo in the bilateral inferior temporal/fusiform gyrus and the CBL, with no significant correlations were found between these parameters and clinical variables of AN [28]. A topological brain network study reported decreased DC in the bilateral inferior frontal gyrus (IFG), along with significant negative correlation between DC in the left IFG and state anxiety scores, and between DC in the right IFG and depression sum scores and state anxiety scores [29]. Although studies on BN are limited, previous study identified reduced fALFF in the left INS and increased fALFF in the IPL, as well as a positive association between internal FC within the left IPL and restrained eating [26]. A graph-theoretical study demonstrated disrupted nodal centrality in BN, involving sensorimotor, visual, subcortical, and limbic regions, including the precentral gyrus (PCG), postcentral gyrus (ProCG), MOG, lingual gyrus, inferior temporal gyrus (ITG), PRU, OFC, PHG, amygdala, thalamus, putamen, and INS [30]. These findings were broadly consistent with results from another study employing diffusion tensor imaging [31]. However, many of these investigations have compared AN or BN with healthy controls (HC) relying on a limited set of functional measures. Few studies have directly compared AN, BN, and HC using a multi-metric approach that captures both local and global alterations in intrinsic brain activity.

Therefore, this present study systematically examined resting-state brain activity in AN, BN, and HC participants using a combined analysis of ALFF, fALFF, ReHo, and DC to collectively assess neuronal activity intensity, regional synchronization, and global connectivity. Elucidating these neural signatures may enhance our understanding of ED pathophysiology and contribute to the development of mechanistically informed interventions in the future. We formulated three exploratory hypotheses based on previous findings: (1) patients with AN and BN would exhibit overlapping altered activity within cortico-striatal region circuit compared to HC; (2) existence of distinct, disorder-specific neural activity characteristics of AN and BN, which maybe mainly located in the cortical regions; and (3) brain activity in the prefrontal cortical areas may be associated with BMI and clinical symptoms of AN and BN.

Methods

Participants

This study was approved by the Research Ethics Committee of Nanjing Brain Hospital, affiliated with Nanjing Medical University, and conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants and, where applicable, from their legal guardians for individuals under 18 years of age.

A total of 54 female patients with ED, including 28 with AN and 26 with BN, were recruited from the inpatient and outpatient units of the Department of Medical Psychology at Nanjing Brain Hospital. Within the AN group, 16 patients met criteria for the binge-eating/purging subtype (AN-BP), and 12 for the restricting subtype (AN-R). Diagnoses were independently confirmed by two board-certified psychiatrists based on the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) [2]. Eligibility criteria for ED participants were as follows: (1) female sex; (2) age 14–40; (3) right handedness; (4) BMI ranging from 12 to 18.5 kg/m² for AN and 18.5–25 kg/m² for BN, as defined by age-specific Chinese BMI cutoffs [32]; (5) no history of psychotic symptoms or comorbid psychiatric disorders; (6) no prior suicidal or aggressive behavior; (7) no current use of antidepressant or antipsychotic medications; (8) no detectable cognitive or communicative impairments.

The HC group comprised 31 age-, sex-, and education-matched individuals recruited via advertisements, community outreach, and hospital staff referrals. All HC participants were female, aged 14–40, right-handed, and of Han Chinese ethnicity. Exclusion criteria for all participants included: (1) major physical or neurological illness; (2) current or past substance or alcohol use disorder; (3) family history of psychiatric disorders in first-degree relatives; (4) inability to undergo MRI scanning or inability to tolerate scanning procedures; and (5) pregnancy or breastfeeding.

Questionnaire measures

The Chinese version of the Eating Disorders Examination-Questionnaire 6.0 (EDEQ, 6.0 version) was used to evaluate core ED psychopathology [33, 34]. This 36-item self-report instrument assesses dysfunctional eating-related attitudes and behaviors across four subscales: Eating Concern (EDEQ_E), Shape Concern (EDEQ_S), Weight Concern (EDEQ_W), and Restraint (EDEQ_R). A global score is calculated as the average of the four subscale scores. Each item is rated on a 7-point scale (0–6), with higher scores indicating greater symptom severity. To assess affective symptoms, all participants completed the Zung Self-Rating Anxiety Scale (SAS) [35] and the Zung Self-Rating Depression Scale (SDS) [36], which measure subjective levels of anxiety and depression, respectively. Additional clinical and demographic data, including age, sex, education level, BMI, illness duration, and other relevant variables, were collected using a structured questionnaire. Illness duration was calculated as the interval from the self-reported onset of initial symptoms to the time of enrollment. Participants completed this form independently, and responses were subsequently verified by a trained psychiatrist.

Imaging acquisition and analysis

All imaging data were acquired using a 3.0 Tesla GE MRI scanner at Nanjing Brain Hospital, affiliated with Nanjing Medical University. During scanning, participants wore earplugs to reduce acoustic noise and used foam pads to minimize head motion. They were instructed to stay awake and relaxed, with their eyes closed, and refrained from focused thinking throughout the resting-state scan. Structural T1-weighted images were obtained using a 3D magnetization-prepared rapid acquisition gradient echo (MPRAGE) sequence with the following parameters: Repetition/Echo Time (TR/TE) = 8.2ms/3.2ms, flip angle = 12°, 172 slices, field of view (FOV) = 256 m x 256 mm; matrix = 256 mm x 256 mm, slice thickness =1 mm. Rs-fMRI data were scanned via a gradient recalled echo-planar imaging (GRE-EPI) sequence, with the settings: TR/TE = 2000ms/30ms, flip angle = 90°, FOV = 220 mm × 220 mm, matrix = 64 × 64, slice thickness/gap = 4.0 mm/4.0 mm, 40 slices, time points = 240.

Preprocessing of Rs-fMRI data was conducted using the Data Processing & Analysis for Brain Imaging (DPABI) toolbox [37]. The following steps were applied: discarding the first 10 time points, slice timing correction, and head motion correction (participants with head translation > 2.5 mm or rotation > 2.5° were excluded). T1-weighted images were co-registered to functional data, segmented using DARTEL, and nuisance covariates (Friston 24 motion parameters, white matter, and cerebrospinal fluid signals) were regressed out. Datasets were normalized to MNI space using DARTEL, detrended, and spatially smoothed. ALFF, fALFF, ReHo, and DC metrics were extracted using the DPABI toolbox. ALFF was calculated as the power within the 0.01–0.08 Hz frequency band and reflects spontaneous regional neural activity, and fALFF was computed as the ratio of power in the 0.01–0.08 Hz band to the total power across the full frequency spectrum, indicating the relative contribution of spontaneous fluctuations. Voxel-wise ALFF and fALFF values were standardized by dividing by their respective global means and further transformed into Z-score maps for statistical comparison [19]. ReHo maps were generated by calculating KCC among the time series of a given voxel and its 26 neighboring voxels. Each participant’s ReHo map was normalized by dividing by the individual mean ReHo within a brain mask and smoothed with a 6-mm full-width at half-maximum (FWHM) Gaussian kernel. DC was computed as the binary degree (correlation threshold r > 0.25) of each voxel within a predefined brain mask, reflecting the total number of significant connections (i.e., node strength). Individual DC maps were transformed into Z-scores and smoothed using a 6-mm Gaussian kernel. Motion parameters (translation and rotation) were compared across groups, and no significant differences were observed between patients and HCs. In all group-level statistical analyses, age, BMI, and total intracranial volume (TIV) were included as covariates, given evidence that brain volume reductions are closely associated with lower BMI and may resolve with weight restoration [38].

Statistical analysis

Normality tests indicated that age, years of education, and SAS scores were normally distributed across all participants. These variables were summarized as means ± standard deviations and analyzed using one-way analysis of variance (ANOVA), followed by post hoc comparisons using the least significant difference (LSD) method. In contrast, BMI, illness duration, SDS scores, and total and subscale scores of the EDEQ exhibited non-normal distributions. These variables were reported as medians with interquartile ranges and analyzed using non-parametric methods. Specifically, differences in illness duration between the AN and BN groups were evaluated using the Mann-Whitney U test. Comparisons among all three groups were performed using the Kruskal-Wallis H test, with post hoc pairwise comparisons corrected using the Bonferroni method. Chi-square analysis was used for categorical data by Fisher’s exact test. All statistical analyses were conducted using SPSS version 21.0. A two-tailed p-value < 0.05 was considered statistically significant. For the analysis of resting-state fMRI metrics (ALFF, fALFF, ReHo, and DC), group comparisons among AN, BN, and HC were performed using ANOVA in the DPABI toolbox, controlling for age, years of education, and GMV as covariates. To correct for multiple comparisons, Gaussian Random Field (GRF) theory was applied, with a voxel-level threshold of p < 0.001 and a cluster-level threshold of p < 0.05. Significant clusters surviving GRF correction were extracted, and post hoc comparisons were conducted in SPSS 21.0 to assess differences between the AN vs. HC, BN vs. HC, and AN vs. BN groups. Bonferroni correction was applied to account for multiple testing.

To investigate the neurobiological relevance of altered intrinsic brain activity, we conducted correlation analysis with clinical variables (BMI, illness duration, SAS, SDS, EDEQ and its subscales) to identify significantly different activity among the patients based on the previous studies. Specifically, Pearson correlation coefficient was calculated between the abnormal brain activity and their scores on the clinical variables using SPSS version 21.0. Statistical significance of the correlation was determined by two-sided p < 0.05 after Bonferroni correction.

Results

Demographic and clinical variables

Table 1 summarizes the demographic and clinical characteristics of all the participants. No statistically significant differences were observed among the AN, BN, and HC in terms of age or years of education, marital status, or residential setting. Similarly, illness duration did not differ significantly between the AN and BN groups. However, significant differences were found for SAS scores, SDS scores, and EDEQ total and subscale scores. Post hoc analyses showed that both the AN and BN groups scored significantly higher than the HC group, while no significant differences emerged between the AN and BN groups. BMI also differed significantly across the three groups.

Table 1.

Description and comparison of demographic and clinically psychological characteristics of the included subjects

Variable	AN	BN	HC	Statistics		Effect size	Post hoc test, p-value
	N = 28	N = 26	N = 31	F/t/H/Z/χ2	p	η2	AN vs. HC	BN vs. HC	AN vs. BN
Age (years)	20.86 ± 5.56	22.69 ± 5.85	22.23 ± 6.18	0.723	0.488a	-	0.375b	0.766b	0.255b
Education (years)	12.64 ± 2.75	13.76 ± 2.62	13.77 ± 2.84	1.582	0.212a	-	0.119b	0.995b	0.137b
Married/Unmarried	4/24	4/22	3/28	0.592	0.781f	-	0.698f	0.691f	1.000f
Non-urban/Urban	3/25	4/22	3/28	0.582	0.838f	-	1.000f	0.691f	0.699f
SAS	52.78 ± 11.18	56.39 ± 10.92	45.89 ± 8.12	8.087	0.001a	0.165	0.01b	< 0.001b	0.191b
SDS	62.50(12.25)	66.25(13.13)	53.75(5.00)	32.05	< 0.001c	0.290	< 0.001d	< 0.001d	1.000d
BMI	15.81(2.25)	19.77(3.88)	17.96(1.49)	48.29	< 0.001c	0.462	< 0.001d	< 0.034d	< 0.001d
Illness duration (years)	2.00(3.00)	2.00(2.00)	-	-0.87	0.384e	-	-	-	-
EDEQ	4.28(1.51)	3.48(1.45)	1.41(0.67)	50.04	< 0.001c	0.601	< 0.001d	< 0.001d	1.000d
EDEQ_R	3.40(1.40)	3.80(2.90)	1.40(1.00)	34.69	< 0.001c	0.366	< 0.001d	< 0.001d	1.000d
EDEQ_E	4.60(2.75)	3.20(2.25)	1.20(0.60)	45.64	< 0.001c	0.496	< 0.001d	< 0.001d	0.834d
EDEQ_S	4.25(1.34)	4.25(1.34)	1.63(0.75)	40.14	< 0.001c	0.535	< 0.001d	< 0.001d	1.000d
EDEQ_W	4.20(2.25)	4.00(1.50)	1.40(0.80)	43.10	< 0.001c	0.494	< 0.001d	< 0.001d	1.000d

^a the p values were obtained by one-way analysis of variance tests for normally distributed data; ^b the p values were obtained for post-hoc comparisons by least significance difference (LSD) for normally distributed data; ^c the p values were obtained by k-samples non-parametric tests (Kruskal-Wallis H test) for non-normally distributed data; ^d the p values were obtained for host-hoc comparisons by Bonferroni corrected) for non-normally distributed data; ^e the p values were obtained by two-sample non-parametric tests (Mann-Whitney U test) for non-normally distributed data; ^f the p values were obtained by chi-square analysis and Fisher’s exact test; AN anorexia nervosa, BN bulimia nervosa, HC healthy controls, SAS Zung Self-Rating Anxiety Scale, SDS Zung Self-Rating Depression Scale, BMI body mass index, EDEQ Eating Disorders Examination Questionnaire, EDEQ_R EDEQ- restraint, EDEQ_E EDEQ-eating concern, EDEQ_S EDEQ- shape concern, EDEQ_W EDEQ- weight concern.

Abnormalities in brain ALFF, fALFF, ReHo, and DC in patients with AN and BN

All the results were showed in Table 2; Fig. 1, and the overall effect sizes of results were large and varied between 0.249 and 0.462 (η2). ANOVA revealed significant group differences in ALFF across eight brain clusters, including the right PHG (PHG_R), left and right MFG (MFG_L, MFG_R), left STR (STR_L; encompassing the caudate, putamen, and pallidum), left and right INS (INS_L, INS_R), right STG (STG_R), and left MOG (MOG_L). Compared to HC, both the AN and BN groups exhibited significantly reduced ALFF values in the PHG_R, bilateral MFG, and bilateral INS, alongside elevated ALFF values in the STR_L and MOG_L. No significant differences in ALFF were observed between the AN and BN groups.

Table 2.

Brain regions with significant differences in the patients with AN and BN across all the metrics compared with HC

Metrics	Brain regions	Cluster sizes	Peak MNI coordinates (x, y, z)	Peak F-value a	Effect size (η2)	Post-hoc comparisons (p-values b)
						AN vs. HC	Cohen’s d	BN vs. HC	Cohen’s d	AN vs. BN	Cohen’s d
ALFF	Right parahippocampal gyrus (PHG_R)	29	36, -12, -27	20.783	0.336	<0.001	0.218	<0.001	0.215	1.000	0.118
	Left middle frontal gyrus (MFG_L)	66	-24, 42, -12	20.194	0.330	<0.001	0.244	<0.001	0.243	1.000	0.202
	Right middle frontal gyrus (MFG_R)	96	24, 26, -12	19.351	0.321	<0.001	0.250	<0.001	0.252	1.000	0.192
	Left striatum (STR_L)	34	-9, 6, 0	16.303	0.285	<0.001	0.334	0.002	0.311	0.172	0.383
	Left insular cortex (INS_L)	41	-51, 6, 0	23.453	0.364	<0.001	0.280	<0.001	0.278	1.000	0.258
	Right insular cortex (INS_R)	68	36, -3, 15	26.026	0.388	<0.001	0.183	<0.001	0.188	1.000	0.189
	Right superior temporal gyrus (STG_R)	18	63, -48, 21	13.588	0.249	0.001	1.044	<0.001	0.978	0.530	0.699
	Left middle occipital gyrus (MOG_L)	37	-27, -60, 27	16.809	0.291	<0.001	0.419	<0.001	0.358	1.000	0.472
fALFF	Right striatum (STR_R)	23	24, 9, -15	22.207	0.351	<0.001	0.328	0.151	0.308	<0.001	0.362
	Left calcarine (CAL_L)	93	-21. -69, 9	23.135	0.361	<0.001	0.429	1.000	0.390	<0.001	0.381
	Right precuneus (PRU_R)	18	15, -42, 39	20.080	0.329	<0.001	0.456	0.123	0.419	<0.001	0.485
	Left middle occipital gyrus (MOG_L)	15	-24, -57, 30	17.153	0.295	<0.001	0.523	<0.001	0.495	1.000	0.557
	Right precentral gyrus (PCG_R)	20	42, -12, 45	16.305	0.285	1.000	0.528	<0.001	0.433	<0.001	0.506
ReHo	Right cerebellum_8 (CBL_R)	58	18, -42, -57	20.211	0.330	<0.001	0.091	<0.001	0.088	1.000	0.102
	Left striatum (STR_L)	397	-21, -6, -30	30.636	0.428	<0.001	0.084	<0.001	0.081	0.055	0.074
	Right striatum (STR_R)	241	21, 12, -15	35.269	0.462	<0.001	0.073	<0.001	0.069	0.001	0.070
	Left cerebellum_6 (CBL_L)	35	-18, -60, -21	15.681	0.277	<0.001	0.109	<0.001	0.124	0.470	0.136
	Left middle occipital gyrus (MOG_L)	112	-36, -75, 3	27.646	0.403	<0.001	0.111	<0.001	0.094	0.098	0.096
	Right middle occipital gyrus (MOG_R)	73	39, -69, 3	28.202	0.408	<0.001	0.128	<0.001	0.126	1.000	0.097
	Left insular cortex (INS_L)	48	-51, -3, 9	27.261	0.399	<0.001	0.073	<0.001	0.077	0.062	0.072
	Right insular cortex (INS_R)	54	51, 0, 9	19.067	0.317	<0.001	(0.103)	<0.001	0.111	1.000	0.098
DC	Left superior temporal gyrus (STG_L)	12	-57, 3, 0	25.989	0.388	<0.001	0.492	<0.001	0.497	0.029	0.394
	Right precentral gyrus (PCG_R)	8	51, 0, 9	20.369	0.332	<0.001	0.597	<0.001	0.513	1.000	0.548
	Right calcarine (CAL_R)	12	12, -75, 12	18.998	0.317	0.005	0.525	0.007	0.551	<0.001	0.513
	Left insular cortex (INS_L)	7	-36, -6, 12	17.296	0.297	<0.001	0.646	<0.001	0.619	1.000	0.457
	Left middle occipital gyrus (MOG_L)	13	-21, -57, 39	22.577	0.355	<0.001	0.441	<0.001	0.498	0.032	0.545
	Right superior frontal gyrus (SFG_R)	7	12, 6, 54	17.989	0.305	<0.001	0.603	0.133	0.446	<0.001	0.584

^a One-way analysis of variance (ANOVA) corrected by Gaussian random field (GRF) (voxel-level threshold of p < 0.001 and a cluster-level threshold of p < 0.05); ^b Bonferroni corrected; ED eating disorder, AN anorexia nervosa, BN bulimia nervosa, HC healthy controls, MNI Montreal Neurological Institute, ALFF amplitude of low-frequency fluctuations, fALFF fractional amplitude of low-frequency fluctuations, ReHo regional homogeneity, DC degree centrality.

Fig. 1.

Brain regions with significant differences among the three groups from ANOVA in ALFF, fALFF, ReHo, and DC. Notes: Clusters are shown after multiple comparison by GRF (voxel-level threshold of p < 0.001 and a cluster-level threshold of p < 0.05). ANOVA analysis of variance, ALFF amplitude of low-frequency fluctuations, fALFF fractional amplitude of low-frequency fluctuations, ReHo regional homogeneity, DC degree centrality, L left, R right

Analysis of fALFF revealed significant group differences across five clusters (Table 2; Fig. 1). These clusters were centered in the right STR (STR_R; encompassing the caudate and putamen), MOG_L, left calcarine cortex (CAL_L; extending into the right calcarine cortex and left precuneus), right PRU (PRU_R), and right PCG (PCG_R). Relative to HC, patients with AN exhibited reduced fALFF in the CAL_L (extending to CAL_R) and increased fALFF in the STR_R, MOG_L, and PRU_R. In contrast, patients with BN showed elevated fALFF in the MOG_L and PCG_R, with no regions showing reduced fALFF compared to HC. When directly comparing AN and BN, the AN group showed significantly lower fALFF in the CAL_L and PCG_R, but higher fALFF in the STR_R and PRU_R. No significant group difference was observed in the MOG_L.

For ReHo, the results revealed significant group differences across eight clusters among the three groups (Table 2; Fig. 1). Peak regions were identified in the right cerebellar lobule VIII (CBL_R) and left cerebellar lobule VI (CBL_L), as well as the STR_L, with the peak voxel located in the PHG_L extending to the hippocampus and amygdala (including caudate and putamen). Similarly, in the STR_R, the peak was located in the right hippocampal gyrus with extension to the PHG, amygdala, caudate, and putamen. Additional clusters were observed in the bilateral MOG and bilateral INS. Compared to HC, individuals with AN demonstrated decreased ReHo in the bilateral MOG and bilateral INS, alongside increased ReHo in the bilateral CBL and STR. The BN group showed a similar pattern of altered ReHo relative to HC. When directly comparing AN and BN, the AN group exhibited significantly higher ReHo values in the STR_R.

DC analysis revealed significant group differences across six clusters among the three groups, as identified by ANOVA (Table 2; Fig. 1). Peak clusters were located in the left STG (STG_L), PCG_R, CAL_R, INS_L, MOG_L, and right SFG (SFG_R). Compared to HC, the AN group exhibited decreased DC in the STG_L, PCG_R, INS_L, and CAL_R, along with increased DC in the MOG_L and SFG_R. The BN group demonstrated decreased DC in the STG_L, PCG_R, and INS_L, but increased DC in the MOG_L and CAL_R. Direct comparisons between the AN and BN groups revealed that patients with AN had significantly higher DC in the STG_L and SFG_R, and significantly lower DC in the CAL_R and MOG_L.

Correlation analysis of clinical assessment and abnormal brain intrinsic activity

After correcting for multiple testing [ALFF: 0.05/8 ROIs*9 clinical variables (BMI, illness duration, SAS, SDS, EDEQ, EDEQ_R, EDEQ_E, EDEQ_S, EDEQ_W); fALFF:0.05/5 ROIs*9 clinical variables; ReHo: 0.05/8 ROIs*9 clinical variables; DC: 0.05/6 ROIs*9 clinical variables], no correlation coefficients between the four metrics and clinical variables were significant in the AN group, and we found three significant correlations between ReHo in the MOG_R and EDEQ score (r = − 0.678, 95% CI [-0.841, -0.394], p = 1.43E-04), between DC values in the PCG_R and EDEQ score (r = 0.673, 95% CI [0.387, 0.841], p = 1.64E-04), EDEQ_E (r = 0.736, 95% CI [0.488, 0.874], p = 1.80E-05) in BN group (Fig. 2). Some nominally significant correlations (uncorrected p < 0.05) were also observed in AN and BN groups (details in the Online Supplementary Material, Tables S1-S3). Because of the relatively small sample size, we further validated these findings. We conducted univariate analyses using General Linear Model (GLM) with data from both AN and BN group for the three significant correlation findings separately, and we added an interaction term in the model to investigate a potential moderation effect by different diagnosis between AN and BN. The outcomes were consistent with these findings (details in the Supplementary Material).

Fig. 2.

Scatter plots depicting the correlations between symptom assessments and intrinsic activity of different region in patients with BN. BN bulimia nervosa, EDEQ Eating Disorders Examination-Questionnaire, EDEQ_E EDEQ-eating concern, ReHo regional homogeneity, DC degree centrality, MOG_R right middle occipital gyrus, PCG_R right precentral gyrus. p < 0.05 after Bonferroni correction

Discussions

In this study, we investigated intrinsic brain activity alterations in patients with AN and BN using multiple metrics: ALFF, fALFF, ReHo, and DC, reflecting different dimensions of brain function including neuronal activity, synchronization of local activity, and coordination across the whole brain. Three primary findings emerged. First, patients with AN and BN exhibited overlapping patterns of altered brain activity across all metrics, suggesting shared neurobiological mechanisms, mainly located in the bilateral DLPFC, bilateral INS, PHG_R, bilateral STG, bilateral MOG, bilateral CBL, and bilateral STR. Among them, the activity of the three regions (STG_L, STR_R, MOG_L) exhibited significant differences between AN and BN groups. Moreover, we found distinct, disorder-specific alterations across different metrics. AN group showed increased fALFF in the STR_R and PRU_R, along with elevated DC in the SFG_R, and decreased fALFF and DC in the bilateral CAL, while BN showed increased fALFF in the PCG_R and elevated DC in the CAL_R. Finally, we identified three correlations between the ReHO of MOG_R and EDEQ, and between the DC of PCG_R and EDEQ, and EDEQ_E in BN, and no correlation in AN.

Shared neural activity characteristics for AN and BN

Our findings revealed shared characteristics of activity within cortico- striatal-limbic circuit in AN and BN (Fig. 3), which are highly consistent with previous studies [11, 26, 28, 30, 31], further confirming its role in the pathological mechanisms of ED [39]. As a component of the DLPFC, the MFG is essential for executive control, inhibitory regulation, and behavioral monitoring [40], and the STR plays a central role in reward processing, decision-making, and emotional expression [41, 42]. Altered fronto-striatal circuitry has been implicated in both cognitive control and reward dysregulation in ED [43, 44]. AN is often characterized by excessive control to suppress reward-driven behavior such as restrained eating, and BN may reflect impaired inhibition and enhanced reward sensitivity such as binge eating [44, 45]. Due to diagnosis crossover, recurrent binge eating also occurs in AN, with symptoms overlapping with those of BN [46]. The INS, a hub for interoceptive integration, monitors internal bodily states and coordinates emotional and cognitive responses [47, 48]. Deficits in interoception are central to ED pathology [49, 50]. INS hypoactivity has been reported in AN across multiple contexts [38, 39], including pre- and post- meal states [51] and visual food stimuli [14]. Similarly, the patients with BN also showed decreased activity in the INS [26, 30], with impaired interoceptive accuracy persisting even in recovery [52]. The STG and PHG as key limbic structures are involved in emotional processing and regulation [53, 54], and the MOG and other occipital regions are involved in visual processing, including the perception of body image [55]. Body image disturbance (BID), another hallmark of ED, is closely related to interoceptive dysfunction and INS abnormalities [56, 57]. Structural and functional changes in the STG, MOG, and parietal cortex have been linked to BID [17, 58, 59], influencing eating behavior in AN and BN [60, 61]. Abnormal activity of CBL has been linked to altered responses to food and emotion-related stimuli [14, 62, 63], regulating the appetite and eating behavior [64]. Collectively, our findings may suggest shared dysfunction across the cortico- striatal-limbic circuit underlying common features of ED.

Fig. 3.

This figure shows the shared and disease-specific regions at different metrics in the patients with AN and BN. The regions in the small ellipse represents disease-specific activity characteristics in AN or BN compared with healthy controls, respectively, and the bule means decreased activity while the red means increased activity. The regions in black in the big ellipse represents different activity characteristics between AN and BN. AN anorexia nervosa, BN bulimia nervosa, ALFF amplitude of low-frequency fluctuations, fALFF fractional amplitude of low-frequency fluctuations, ReHo regional homogeneity, DC degree centrality. PHG_R right parahippocampal gyrus, MFG_L left middle frontal gyrus, MFG_R right middle frontal gyrus, SFG_R right superior frontal gyrus, STR_L left striatum, STR_R right striatum, CAL_L left calcarine, CAL_R right calcarine, INS_L left insular cortex, INS_R right insular cortex, STG_R right superior temporal gyrus, STG_L left superior temporal gyrus, MOG_L left middle occipital gyrus, MOG_R right middle occipital gyrus, PRU_R right precuneus, PCG_R right precentral gyrus, CBL_R right cerebellum_8, CBL_L Left cerebellum_6.

Significant differences in activation in STG_L, MOG_L, and STR_R between AN and BN may reflect disorder-specific heterogeneity in BID and reward processing. We speculate that these differences likely reflect distinct patterns of BID and abnormal sensitivity to reward-related stimuli [65, 66]. For example, BID in AN is typically characterized by an unrealistic and excessive assessment of body weight, which may at times approach a delusional state, whereas BID in BN is more often manifested as an excessive emphasis on body shape and weight [67].

Disorder-specific neural activity characteristics of AN and BN

Beyond the shared alterations observed in AN and BN, our findings also revealed disorder-specific patterns of intrinsic brain activity (Fig. 3). In patients with AN, distinct alterations were identified in the STR_R, SFG_R, bilateral CAL, and PRU_R, relative to HC. The increased activation in STR_R in AN might suggest heightened reward sensitivity in response to maladaptive stimuli such as dietary restriction and weight loss, which may be perceived as reinforcing [43]. Compared with the function of the MFG, SFG is more engaged in cognitive control, and behavioral monitoring [40]. Increased cognitive control over eating behavior as a compensatory mechanism were observed in patients with AN, in contrast to diminished inhibitory control observed in patients with BN [14, 44]. The CAL is involved in the functions of visuospatial processing and attentional control [68]. Meta-analytic evidence has documented hypoactivation of the bilateral CAL during emotion and food-related tasks in AN [13]. Consistent with this, a recent high-density EEG study demonstrated enhanced early visual orienting followed by reduced CAL activation, suggesting impaired attentional flexibility and reduced cognitive adaptability in AN [69]. In contrast, patients with BN exhibited unique alterations in the CAL_R, and thus we speculated that the activity in CAL in AN and BN may be related to the different attentional bias [70]. The PRU, as a core hub of the DMN, is implicated in self-referential processing and introspection [71]. It is also linked to the perceptual dimensions of BID, together with the IPL and occipital cortices [72]. Previous studies have reported reduced activation of the PRU in AN during tasks related to food, cognition, and body image [13], and reduced connectivity involving the PRU, prefrontal cortex, STG, and INS in AN [73]. These findings support the notion that hypoactivity in the PRU may reflect persistent internal attentional bias and disrupted interoception, contributing to the phenomenology of BID in AN, compared with BN, who pay more attention to palatable food [74].

The PCG, part of the sensorimotor network (SMN), is involved in somatosensory integration, motor control, and behavioral inhibition [75]. Previous studies have shown increased activation in the PCG, along with DLPFC and INS in patients with BN during exposure to food stimuli [76]. In our study, increased fALFF and DC in PCG_R may suggest that BN may involve heightened sensorimotor engagement and motor impulsivity, consistent with prior reports linking BN to increased habitual responding and impaired motor inhibition [77]. More importantly, our results also showed positive associations between the DC of PCG_R with EDEQ and EDEQ-E, indicating that the more severe the symptoms, the more obvious the impaired inhibitory control and increased motor impulsivity, which accords with the clinical characteristics of BN. These results were in consistent with the previous study [75], but were different with other studies which reported associations between abnormal brain structure and EDEQ assessment [15, 16]. In addition, we also found a negative correlation between the values of ReHo in the MOG_R and EDEQ, which could be considered as another unique characteristic for BN. As mentioned in the discussion, MOG mainly engaged in the processing of BID, which may suggest that patients with severe symptoms exhibit more frequent binge eating behavior and pay less attention to BID [74]. Nevertheless, our study found no significant correlation between abnormal brain activity and clinical variables in AN, which was also consistent with previous study with large sample in AN [25, 28].

Furthermore, some of our findings were inconsistent with previous research. For example, one study examining set-shifting in AN reported increased activation in both the INS and PRU, with higher activation in the AN-R than in the AN-BP [78]. Another meta-analysis noted increased PRU activity in AN during food image viewing [14]. Of note, our study did not observe significant PRU alterations in BN, despite prior reports of hypoactivation in response to food cues [14]. These discrepancies may stem from methodological variations, including sample characteristics and task paradigms.

Limitations

Several limitations should be considered when interpreting the findings of this study. First, the relatively small sample size may limit the generalizability and statistical power of our results. Second, due to insufficient subgroup sizes, we were unable to compare the AN-R and AN-BP subtypes of AN, which may exhibit distinct neural signatures. Third, despite rigorous inclusion criteria were applied, diagnostic crossover and psychiatric comorbidity are common in clinical practice, which may have confounded the results and reduced the representativeness of the sample. Fourth, the cross-sectional design of the current study precluded the observation of dynamic changes in brain activity, as suggested by recent work indicating the absence of “scar” effects following remission [73, 79]. Future multicenter studies with larger sample sizes and longitudinal designs will be necessary to further validate these findings. More detailed subgroup analyses may help reduce potential confounding factors. Longitudinal investigations are also warranted to determine whether the observed intrinsic activity changes persist over time or normalize with clinical recovery. Additionally, although we conducted primarily exploratory correlation analyses, few results survived rigorous multiple-comparison correction. This finding may reflect the complex relationship between behavioral manifestations and brain function and warrants further in-depth investigation using approaches such as multimodal neuroimaging and network-based analyses.

Conclusions

Our multidimensional resting-state fMRI analysis revealed both shared and disorder-specific alterations in patients with AN and BN. The shared dysfunctions within the cortico-striatal-limbic circuits consolidate its central role in the pathological mechanisms in ED. More importantly, there are disorder-specific patterns of neural activity in patients with AN and BN mainly located in the STR, PRU, PCG, and CAL, offering novel insights into the underlying neurobiological mechanisms, potentially informing the future development of targeted biomarkers for clinical assessment and differential diagnosis.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Changjun Teng : writing original draft, resources, project administration, funding acquisition, formal analysis, data curation, investigation, methodology, conceptualization. Jiajia Zhang : resources, formal analysis, data curation. Zheyu Lin : resources, formal analysis, data curation. Xiaomeng Shi : resources, formal analysis, data curation. Xin Wu : resources, formal analysis, data curation. Wei Zhang : resources, formal analysis, data curation. Huan Zhang : resources, formal analysis, data curation. Ning Zhang : writing–review & editing, conceptualization, supervision, resources, formal analysis, conceptualization. Chengbin Guan : supervision, conceptualization, resources, formal analysis, conceptualization ; Huifen Qiao : writing–review & editing, resources, formal analysis, data curation, supervision, conceptualization.

Funding

This work was supported by Health Science and Technology Development of Nanjing (YKK22134, YKK23145); Young Talents Project of Nanjing Brain Hospital(23‑25‑1R8‑1).

Data availability

All research data and methods used in the current manuscript are freely available from the authors upon reasonable request.

Declarations

Ethics approval and consent to participate

This research was approved by the Medical Research Ethics Committee of The Affiliated Brain Hospital of Nanjing Medical University in accordance with the World Medical Association Declaration of Helsinki (approval number:2022-KY069-01). Each participant provided informed written consent to participate in the study. Written informed consent of the patients was obtained from them and their legally authorized representative, while the control participants provided written, informed consent themself after totally understanding the purpose of our study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.