Changes in resting-state functional connectivity between large-scale brain networks after an intensive arts intervention for adolescents with depression

Abstract

Emerging data support the benefits of creative arts interventions for addressing depression in adolescents, but the neural mechanisms are poorly understood. This study examined changes in connectivity among large-scale brain networks implicated in both depression and creativity, before and after an arts intervention for adolescents with depression. Thirty-nine adolescents with depression completed resting-state functional magnetic resonance imaging scans before and after completing a two-week arts intervention called Creativity Camp. To approximate a “waitlist” control, half the sample also completed scanning 2 weeks before camp. Using a graph theory approach, we calculated the strength of connectivity between four networks at the “nexus” of mood and creativity: Default, Control, Salience, and Limbic. For these between-network connection strength indices, we (1) measured pre- to post-camp changes across the whole sample; (2) compared changes during waitlist versus during camp for the rest of the sample; and (3) explored correlations between changes in connectivity and changes in depression. In the whole sample, the strength of connectivity between three network pairs increased significantly from pre- to post-camp: Default–Control, Limbic–Control, and Salience–Default (all p < 0.05, corrected). The waitlist control comparison revealed that pre-to-post camp changes were greater than those observed during the waitlist for two of these network pairs (uncorrected). Connectivity changes did not significantly correlate with changes in depressive symptoms. Intensive engagement in creative arts may contribute to increases in connections between large-scale neural networks that are implicated in both mood and creative thinking. Future research is needed with larger samples using randomized control trials to further understand how creative arts engagement confers its mental health benefits for adolescents.

Introduction

Adolescence is a time period notable for considerable brain development, including the refinement of neural connections and organization of brain networks [1, 2]. It also represents a risk period for the onset of mental illness including depression [3] and suicide [4], a leading cause of death in youth ages 10–24 [5]. Once depression manifests, successful treatment of depression early in life is our best hope for preventing adverse outcomes across the lifespan. Despite the known efficacy of standard treatments such as antidepressant medication and cognitive behavioral therapy, many still do not respond [6, 7], and relapse is frequent [8, 9]. Furthermore, many teens are resistant to engaging in standard treatments for numerous reasons [10]. Because of these concerns, novel interventions for adolescent depression are needed. In tandem with developing a broader array of effective and accessible treatment options, neurobiological research is needed to facilitate treatment optimization (based on an improved understanding of how treatments work).

Emerging data support the idea that engaging in the arts can benefit mental health [11, 12]. Prior work has suggested that infusing arts activities into standard mental health care [13, 14] and education settings [15] can heal through providing an opportunity for youth to see new aspects of themselves, build a sense of interpersonal connection, promote personal growth and build resilience. Similarly, in adults, arts participation positively impacts mental health, with benefits potentially mediated by improved social connectedness, emotion regulation, meaning-making and re-defining identity, personal growth and empowerment [16]. In young adults, greater frequency of participating in performing arts is associated with greater degrees of positive mental health [17]. Importantly, adolescence is a formative stage for creativity development. Abstract reasoning develops during this time period [18], as does divergent thinking and insight [19, 20]. These creative thinking processes interact with identity development [21], underscoring the importance of understanding and tapping into these ideational processes in adolescents with depression, who frequently struggle with negative self-views. Furthermore, the field has increasingly recognized the importance of environmental and experiential enrichment opportunities for adolescents as a protective strategy to prevent various forms of psychopathology. Such enrichment during this critical neurodevelopmental period can uniquely harness the enhanced plasticity within the higher-order association cortices that support complex thought and prevent the longer-term embedding of psychopathology-related sequelae within cortical organization [22]. Critically, youth have advocated for more arts-based approaches in mental health treatment, to promote healing and growth [23]. A recent review highlighted the potential value of arts-based interventions to address the global youth mental health crisis: “Worldwide implementation of arts- and culture-based strategies to address youth mental health will provide critical resources to support the health, wellbeing and flourishing of countless youth across the globe.” [12] At the same time, there has been a call for more rigorous research to better understand the mechanisms underlying these benefits [11, 24].

Since arts interventions invoke creative imagination, examining the neural circuitry that is central to both creative thinking and depression could provide new insights about the mechanisms of the beneficial effects of arts interventions on depression. Depression has a complex neurobiology, with involvement of large-scale brain networks including the executive control, default mode, and salience and limbic networks [25–30]. Emerging neuroscience research has also implicated these same networks, and the connections between them, in creative thinking [31–39]. Specifically, higher strength of the connection between these networks is associated with stronger creativity performance [31], and the connections between these networks increases during scientific creative thinking as opposed to a control condition [40]. Current theory in creativity research suggests that dynamic interaction between these large-scale networks is of central importance during creative cognition and artistic performance [38, 41]. These interactions are thought to support complex cognitive processes including the generation, evaluation, and implementation of original and authentic ideas [37, 42]. Together, this suggests that the connections between these four large-scale networks may represent a key “nexus” that is critical to both mood and creativity. Potentially, using creative activities to strengthen this nexus could have promise for addressing depression. Specifically, since (1) these networks are implicated in both depression and creativity; (2) creative arts interventions can be helpful for depression; and (3) creativity training can lead to increases in functional connectivity between these networks [43]; we posited that such neural shifts might be a part of the mechanism underlying the beneficial effects of creative arts interventions. However, while creativity training can lead to key brain changes in these networks [43–45], despite the field’s recognition of the importance of understanding the mechanisms underlying art interventions’ impact on emotional systems [46], the relevance of such changes to improvements in mental health and well-being has not yet been tested. The converging factors specific to adolescence, including ongoing development of the executive control, default mode, and salience and limbic networks [47–51], increased depression prevalence [3], and creativity development during adolescence [19, 52] underscore the critical importance of examining these mechanisms during this pivotal developmental time period.

We recently developed and implemented a 2-week creative arts intervention called Creativity Camp. The current paper is the third one to be published from this project. The initial paper (Cullen et al., 2024) reported on the quantitative clinical results, showing a reduction of parent- and child-reported depression symptoms and improvement of child-reported well-being after the 2-week camp, and sustained effects noted for parent-reported depression symptoms 6-months later [53]. The second paper (Taniguchi et al., 2025) reported the results from the qualitative analyses of data collected using interviews and observations during the sessions, highlighting the adolescents’ experiences of personal growth during and after the intervention, and their parents’ report of their own surprise at their children’s level of excitement, engagement and positive behavior changes after the camp [54]. In the current work, we examined the resting-state functional magnetic resonance imaging (fMRI) data collected in this study to explore potential brain mechanisms of this intervention. In particular, we focused on the connections between four networks at the nexus of mood and creativity: Default, Salience, Control, and Limbic. We predicted that the connection strength between these networks would increase after camp, and that these increases would be greater than the changes seen in a similar time frame in which no intervention was delivered. We further predicted that these increases would correlate with clinical improvement.

Method

Ethics approval and consent to participate

The study was approved by the Institutional Review Board at the University of Minnesota (STUDY00014280). All methods were performed in accordance with International Council for Harmonisation Good Clinical Practice guideline. The ClinicalTrials.gov registration number is NCT07605975.1 Parents or guardians of the adolescent participants provided written informed consent; adolescents provided written assent.

Overview of the Creativity Camp study

Over two summers, 6 cohorts (total N = 69) of adolescents (ages 12–17 years) with a history of depression symptoms completed the intensive 2-week (8 days, 4 h each day) “Creativity Camp” intervention encompassing multiple creative activities [53]. A subset (3 cohorts, N = 39) of the participants in the first summer underwent brain imaging at three time points. We implemented a delayed-onset clinical trial design: all adolescents participated in the camp, and all underwent clinical and cognitive assessments pre- and post-camp, allowing for whole-group pre-post camp comparisons of the measures. In addition, half the sample (the “Waitlist” group) were also assessed two weeks before camp, allowing us to conduct between-group comparisons of changes that occurred during the waiting period for the “Waitlist” group versus the changes that occurred during camp in the “Active” group (the other half of the sample). While initial group assignment was randomized, if a participant was assigned to a group that did not fit with their schedule (e.g., family vacations or other obligations interfered with completing the assigned assessments for that group), to maximize enrollment and use a strategy to collect “all possible data”, we allowed group switching after randomization and prior to the start of neuroimaging data collection. Since we could not adhere to randomization in all cases, the study design can most aptly be described as quasi-experimental.

Participants

We recruited adolescents with current depression symptoms by connecting with community schools and community centers, posting flyers, distributing recruitment postcards to clinicians, and social media postings. Inclusion criteria were: age 12–17 years and depression symptom severity as measured by the Parent-report Child Depression Inventory 2 (CDI 2) score ≥ 12 at the time of screening. Given the known high levels of co-occurrence with other psychiatric conditions in adolescent depression [55–57], we intentionally kept our inclusion criteria broad, focusing on the presence of depression symptoms as an entry requirement, and did not exclude participants based on co-occurring psychiatric diagnosis. Exclusion criteria were: history of a neurodevelopmental disorder such as intellectual disability or moderate/severe autism, as indicated by parent report; significant medical conditions that would interfere with participating in study activities; current active suicidal ideation and inability to agree to an appropriate safety plan; and MRI contraindications (e.g., claustrophobia, braces, pregnancy) and/or a history of significant neurological conditions that might confound MRI analyses. To avoid ethical concerns, participants who were receiving other treatments for depression (e.g., antidepressant medication, psychotherapy) were allowed to continue those treatments.

Creativity Camp intervention

As described previously [53, 54], professional artists led Creativity Camp activities consisting of varying mediums of art and music, aiming to help adolescents become aware of their own creativity through reflection and exploration. Adolescents were invited to embark on a journey to explore their unique traits, interests, and imaginations. While each session included guidance and introduced specific artistic mediums, participants were also given space to fully explore and lead their own creative processes by centering how they personally connected to the topic. The curriculum positioned participants as artists who create in ways that are personally meaningful and authentic. Adolescents also participated in sharing their visions and stories through public exhibitions of their creative work.

Clinical assessments

At the initial visit, after completing the informed consent process, adolescents underwent a diagnostic assessment using the Mini International Neuropsychiatric Interview (MINI-KID) [58]. Parents and adolescents provided demographic information. Subsequently, clinical data were collected at repeated visits that took place two weeks before camp (Waitlist group only), pre-camp (both groups), post-camp (both groups), two weeks after camp (Active group only), and 6 months after camp. (The current analyses focused only on the data from the first three of these visits.) At these repeated visits, depression symptoms were assessed using the parent and child self-report versions of the CDI 2 [59]. Using both parent and youth perspectives can provide meaningful and complementary information about the child [60, 61]. While youth can report most directly on their own internal experience, parents have the ability to observe their child’s behavior from the outside, and in treatment contexts can sometimes notice changes that the child may not yet appreciate. In the 28-item Child CDI 2, youth (validated for ages 7–17 years) rate their symptoms over the past two weeks on a scale of 0 (no symptoms), 1 (mild or probable symptoms), or 2 (definite symptoms). The Child CDI 2 has high internal reliability (Cronbach’s alpha = 0.91), high short-term test-retest reliability (coefficient = 0.89), and good construct and discriminant validity [59]. In the 17-item Parent CDI 2, parents rate the severity of their child’s symptoms ranging from 0 (not at all) to 3 (much or most of the time). The Parent CDI 2 has good internal consistency and construct validity, with a moderate positive correlation between Child and Parent scales, suggesting evidence of convergent validity [62]. As previously reported [53], additional measures collected during these clinical visits included questionnaires to measure anxiety and well-being. However, the current paper focuses solely on the depression symptoms, which had shown the most robust change in our initial clinical paper [53].

Neuroimaging acquisition

Brain imaging data were collected using a 3 T Siemens Prisma scanner using a standard 32-channel receive-only head coil. The Siemens auto align head sequence was used to prescribe all scans to improve within- and between-session reproducibility. Structural T1/T2 weighted data were acquired following the acquisition parameters of the Adolescent Brain and Cognitive Development (ABCD) study [63] (1 mm isotropic voxels). Functional data were acquired using a multi-echo fMRI sequence [64]. Parameters included TR = 1761 ms, TE = 14.20/38.94/63.66/88.40/113.12 ms, voxel=2 mm isotropic, MB = 6/IPAT = 2. A pair of opposite phase-encode multi-echo spin echo field maps were acquired with voxel parameters matched to the fMRI data for use in distortion correction. One 12-minute resting-state fMRI scan was collected, during which participants were instructed to keep their eyes open while viewing a fixation cross and to “not think about anything in particular.” An fMRI task to assess brain activation during creative imagination was also collected; those results will be reported separately.

Neuroimaging data preprocessing

Dicom data were converted to the Brain Imaging Data Structure (BIDS) standard format [65]. Initial preprocessing of the structural and functional data was conducted using fMRIPrep [66]. Structural data preprocessing included brain extraction, tissue segmentation, parcellation, and spatial normalization, and projection into CIFTI space. Functional data preprocessing included reference image estimation, field unwarping, head motion estimation, co-registration of BOLD data to T1 and standard space, creation of CIFTI-2 file for HCP Grayordinates, confounds estimation, and optimized multi-echo T2* echo combination using tedana [67]. We then applied the eXtensible Connectivity Pipeline- DCAN (XCP-D) [68] for further processing, including selected steps: surface normalization, identifying outlier volumes using Framewise Displacement, applying a respiratory filter to motion parameters (e.g., specify a band-stop or low-pass filter to remove respiration-related signals from the motion parameters, prior to calculating framewise displacement [FD]) [69], FD calculation [70] and thresholding, confound regressor selection (e.g., 36 P includes 6 motion estimates, white matter, cerebrospinal fluid, global signal, the squares of these terms and the first derivatives of those terms) and removal of the first 4 volumes. Denoising steps included confound regression, bandpass filter (0.01 - 0.25 Hz) and despiking. Preprocessed fMRI images were parcellated using the 400-parcel version of the Schaefer atlas [71], where each cortical parcel is assigned to one of the 7 canonical Yeo networks [72]. To further address confounds due to head motion, in addition to the steps described above, following our established lab protocols, we excluded any scans from analysis for which greater than 30% of the volumes had an FD > 0.2 mm. (We also considered more strict thresholds that led to excluding more data; the results for these samples are reported in Supplementary Materials A.)

Network definition

For this study we focused on four networks of interest, as defined by Yeo’s canonical networks [72] made up of parcels from the Schaefer atlas [71]. This atlas was selected because it has sufficient granularity (400 parcels) to allow consideration of a large number of possible parcel-parcel links between networks, and because the parcels are roughly similar to each other in size, avoiding size-related biases in the connectivity analysis. Specifically, we focused on the Control network (which contains 50 Schaefer parcels), Default network (90 parcels), Limbic network (28 parcels), and Salience network (labeled as “Salience/Ventral Attention” in Yeo / Schaefer, which contains 40 parcels) (see Fig. 1).

Fig. 1.

Defining the “nexus” networks implicated in both creativity and depression using the Yeo / Schaeffer parcellation.

Connectivity analysis using graph theory

The interaction between complex brain networks can be characterized using graph theory approaches, which quantify topologies of mathematical networks defined by a collection of nodes (here, brain regions represented by parcels) and edges (connectivity between nodes) [73]. To characterize the strength of connection between the control, default, limbic and salience networks (as defined above), we first constructed a weighted graph, represented by cross-correlations between the average time series from each of the network’s parcels, where each link contains information about connection strength between the nodes. Second, the strength of the connection between a given pair of canonical networks was derived by calculating the sum of all positive edge weights [73] (Fisher Z - transformed correlations representing connectivity between the parcels of the network) between each pair of networks and then dividing by the number of possible edges (nodal connections) between those two networks. (Note: While we focused here on connectivity defined by positive edge weights, following expert recommendations and for ease of interpretability [73], we also explored how the intervention would impact connectivity using the negative edge weights; those results are reported in Supplementary Materials D.) The number of possible connections is equal to the product of the number of parcels for each network in the pair, divided by 2. Thus the total possible edges connecting each pair of networks are: Default-Control = 2250; Default-Salience = 1800; Default-Limbic = 1260; Control-Salience = 1000; Control-Limbic = 700; Salience-Limbic = 560.

Statistical analysis

For each of the six between-network connections, we first compared pre-camp versus post-camp connectivity for the whole sample using paired t-tests. Next, we conducted independent samples t-tests to compare the change in connectivity observed in the “Active” group during the camp period (calculated by subtracting their pre-camp connectivity data from their post-camp connectivity data) with the change observed in the “Waitlist” group during their “waiting” period (calculated by subtracting their connectivity data collected two weeks before camp from their pre-camp connectivity data). To characterize the magnitude of these effects independent of sample size, Cohen’s d was calculated for all t-tests. Finally, to evaluate the relationship between neural change and clinical change, we conducted a series of linear regression models to evaluate the relationships between pre-to-post camp change in depressive symptoms (child-reported and parent-reported CDI 2 scores used in separate models), and pre-to-post camp changes in between-network connectivity as the main predictor. Following a conditional change approach [74], each model included the corresponding pre-camp CDI-2 score as a covariate to account for baseline severity and potential regression to the mean. In planning the study, the anticipated sample size was 30 participants per group, due to limited funding and the relatively high cost of MRI scanning. A power analysis calculated the minimum detectable effect size given a fixed sample size of N = 15 (for group-specific analyses) and  N= 30 (for combined analyses). Using a paired t-test to account for the within-subject design, we estimated 80% power to detect a moderate difference of d = 0.53 or larger when comparing within-subject changes pre-intervention to post-intervention using both Active and Waitlist groups. Given the limited power afforded by the small sample in this preliminary exploration of the impact of Creativity Camp on brain connectivity, we report both uncorrected and Benjamini-Hochberg corrected p-values, alongside standardized effect sizes, for all the results.

Results

Participants

Forty-three participants were enrolled in the larger study; of these, 39 participants completed the Creativity Camp intervention, and 39 also completed neuroimaging procedures. Usable resting-state fMRI data were available for 15 participants at the 2-week pre-camp visit (the Waitlist group); for 38 participants at the pre-camp visit (18 Active and 20 Waitlist); and 32 participants at the post-camp visit (17 Active and 15 Waitlist). Demographic and clinical data for the Neuroimaging sample are presented in Table 1.

Table 1.

Demographic and Clinical Characteristics of the Sample.

	Full Sample (N = 39)		Waitlist Group (N = 18)		Active Group (N = 21)
Demographic Characteristics
	M (SD)		M (SD)		M (SD)
Age	15.37 (1.71)		15.30 (1.85)		15.42 (1.62)
	N	%	N	%	N	%
Ethnicity
Hispanic/Latinx	6	15.38%	3	16.67%	3	14.29%
Not Hispanic/Latinx	33	84.62%	15	83.33%	18	85.71%
Race
White	31	79.49%	15	83.33%	16	76.19%
Asian	5	12.82%	2	11.11%	3	14.29%
Black	4	10.26%	2	11.11%	2	9.52%
Additionala	2	5.13%	2	11.11%	0	0.00%
Native American / American Indian	1	2.56%	0	0.00%	1	4.76%
Sex
Female	32	82.05%	16	88.89%	16	76.19%
Male	7	17.95%	2	11.11%	5	23.81%
Gender Identity
Man	8	20.51%	2	11.11%	5	23.81%
Woman	18	46.15%	10	55.56%	8	38.10%
Transgender	4	10.26%	1	5.56%	3	14.29%
Genderfluid/Genderqueer	4	10.26%	1	5.56%	3	14.29%
Other	4	10.26%	3	16.67%	1	4.76%
Prefer not to answer	2	5.13%	1	5.56%	1	4.76%
Parent Education
High School	2	5.13%	1	5.56%	1	4.76%
Some College	5	12.82%	3	16.67%	2	9.52%
Associate Degree	5	12.82%	1	5.56%	4	19.05%
Bachelors Degree	9	23.08%	4	22.22%	5	23.81%
Masters Degree	13	33.33%	8	44.44%	5	23.81%
Doctoral Degree	3	7.69%	1	5.56%	2	9.52%
Professional Degree	2	5.13%	0	0.00%	2	9.52%
Clinical Characteristics
Diagnosis	N	%	N	%	N	%
Major Depressive Disorder	27	69.23%	12	66.67%	15	71.43%
Manic Episode	2	5.13%	1	5.56%	1	4.76%
Bipolar Disorder	6	15.38%	2	11.11%	4	19.05%
Panic Disorder	9	23.08%	2	11.11%	7	33.33%
Anxiety Disordersb	37	94.87%	17	94.44%	20	95.24%
Obsessive Compulsive Disorder	2	5.13%	1	5.56%	1	4.76%
Post-traumatic Stress Disorder	3	7.69%	0	0.00%	3	14.29%
Substance Use Disorder	1	2.56%	1	5.56%	0	0.00%
Tic Disorders	5	12.82%	3	16.67%	2	9.52%
Attention Deficit Hyperactivity Disorder	6	15.38%	2	11.11%	4	19.05%
Conduct Disorder	2	5.13%	1	5.56%	1	4.76%
Eating Disorders	6	15.38%	3	16.67%	3	14.29%
Medication	N	%	N	%	N	%
Taking Medication	28	71.79%	12	66.67%	16	76.19%
Not Taking Medication	11	28.21%	6	33.33%	5	23.81%

^aTwo participants selected “some other race” and provided the following responses, respectively: “ADOS” (possibly referring to the term “American Descendants of Slavery”) and “Filipino, Swedish, Portuguese.”

^bAnxiety Disorders include Separation Anxiety, Social Phobia, Specific Phobia, and Generalized Anxiety Disorder.

Clinical results

Similar to the previously reported results for the entire sample of Creativity Camp participants (N = 69) [53], analyses for the Neuroimaging subsample showed significantly decreased depression after the camp. Specifically, depression symptoms as measured by both the parent and child CDI significantly decreased pre-to-post-camp (CDI parent: t(37) = 4.53, p < 0.001; CDI child: t(38) = 3.33, p = 0.002). Comparison of the “Active” group’s clinical pre-to-post camp changes to the “Waitlist” group’s changes during the two weeks before camp was significant for both parent- and child-reported depression (CDI parent: t(17) = −11.13, p < 0.001; CDI child: t(17) = −13.20, p < 0.001). A table summarizing the means and standard deviations for the CDI scores at each visit is provided in Supplementary Materials C.

Changes in between-network connectivity after Creativity Camp

Pre-post increases in between-network connectivity were notable in three of the six between-network connections. The network pairs that showed significant increases in connectivity from pre- to post-camp were the Default–Control, Limbic–Control, and Salience–Default. Increases in all three of these connections were still significant after correcting for multiple comparisons (see the left side of Table 2).

Table 2.

Camp-related changes in connectivity strength between the “nexus” networks.

Network Pair	Pre-post camp (N = 35) (paired t-test pooled groups)					Active (N = 18) vs Waitlist (N = 15) (two-sample t-test)
	t	p	p-adj	95% CI	Cohen’s d (effect size)	t	p	p-adj	95% CI	Cohen’s d (effect size)
Default-Control	−3.040	0.005**	0.025*	[−0.019, −0.004]	0.514	1.483	0.149	0.296	[−0.004, 0.027]	0.518
Default-Salience	−2.749	0.010*	0.038*	[−0.015,−0.002]	0.465	2.581	0.015*	0.091.	[0.003,0.026]	0.902
Default-Limbic	−1.838	0.075.	0.224	[−0.025,0.001]	0.311	0.956	0.347	0.347	[−0.014,0.038]	0.334
Control-Limbic	−3.009	0.005**	0.025*	[−0.024,−0.005]	0.509	2.310	0.031*	0.093.	[0.002,0.037]	0.807
Control-Salience	−1.421	0.165	0.329	[−0.018,0.003]	0.240	1.319	0.197	0.296	-[0.008,0.037]	0.461
Salience-Limbic	−0.502	0.619	0.619	[−0.011,0.007]	0.085	1.099	0.280	0.337	[−0.007,0.023]	0.384

Adjusted p values (p-adj) reflect Benjamini-Hochberg correction for multiple comparisons. N’s correspond to the number of participants who had usable resting-state fMRI data that could be used in each analysis.

p < 0.05 (*), p < 0.01 (**); trend-level p < 0.1 (·)

Effect sizes (Cohen’s d) are categorized as small (d ≥ 0.2), medium (d ≥ 0.5), and large (d ≥ 0.8)

As shown on the right side of Table 2, the increases after camp in Limbic-Control connections and Salience-Default networks in the “Active” group were greater than the increases observed over the two weeks before camp in the “Waitlist” group. After correction for multiple comparisons, these findings were at a trend level of significance.

Results of the linear regression models testing the relationship between change in between-network connectivity strength with change in depression scores were not significant (see Supplementary Materials B). Furthermore, there were no significant changes observed when examining the negative edge weights; see Supplementary Materials D.

Discussion

Emerging data support the potential mental health benefits of creative arts interventions for young people, but very little is known about the neural mechanisms of these interventions. This study examined brain changes associated with participating in “Creativity Camp”, a creative arts intervention for adolescents, focusing specifically on strength of connections between nodes in large-scale neural networks at the nexus of both creative thinking and mood (Default, Control, Salience and Limbic). Strengths of this study include its prospective longitudinal design with a Creativity Camp as a probe, its inclusion of a waitlist comparison, and its innovative emphasis on network-level mechanisms at the intersection of creativity and depression. Key findings included significant pre-to-post increases in connectivity strength within three of six network pairs studied here, with tentative evidence suggesting that increases in two of the connections were greater than those observed in the Waitlist group. However, at a within-subject level, network pair connectivity changes during camp did not significantly correlate with the observed clinical improvement during the same period.

The current work adds to growing knowledge about how arts interventions impact the “nexus” brain networks described here in ways that may be relevant to their therapeutic effects. First, the increase found here in connectivity between the Control and Limbic networks aligns with a recent systematic review of studies investigating the neural mechanisms of creative activities, which concluded that both active and passive creative arts activities engage fronto-limbic neural circuits, suggesting shared mechanisms between creative expression and emotion processing [75]. Frontolimbic circuits have been consistently implicated in depression, including in adolescents [28, 76–78], and changes in this network have at times been implicated in treatment response [79–81]. Second, our findings showing that intensive arts engagement led to increases in connectivity between the Default and Control networks is aligned with a prior study that compared brain connectivity changes between those who engaged in passive versus active arts experiences, and showed that the active (visual arts production) group demonstrated greater increases in connections between the default mode network and frontal and parietal regions (similar to the findings for the Control network described here) [82].

Potentially, art-based interventions may have some “ingredients” that are similar to other kinds of interventions. For example, increases in connectivity between salience, default and executive control networks have been reported in relation to mindfulness meditation [83, 84]. Indeed, there may be aspects to Creativity Camp that have a “mindfulness” quality. Specifically, the Creativity Camp curriculum provided repeated, scaffolded invitations for reflection and self-relevant expression through the creative process. These structured, present-centered experiences promoted grounded attention, potentially facilitating the increase in coordination observed across the “nexus” brain networks. On the other hand, the Creativity Camp curriculum’s emphasis on self-relevant reflection also distinguishes it from other creativity-focused interventions. For example, a study that applied divergent thinking training using the Alternative Uses Task (AUT) did not observe pre-to-post changes in functional connectivity in adolescents after the training [85].

The increases reported here in connection strength among “nexus” networks after Creativity Camp may be informed by a recent review which identified three domains of therapeutic factors unique to creative arts therapies: (1) embodiment (getting in touch with the body, increasing body awareness, fuller experience of the body); (2) concretization (shaping abstract thoughts and experiences into tangible form); and (3) symbolism and metaphor (allowing for projection of emotion and ideas into the art medium) [86]. Given the salience network’s function of detecting and processing both internal and external stimuli to guide behavior [87, 88], the observed increases between the Salience and Default networks here could be relevant to the “embodiment” therapeutic factor. Further supporting the relevance of embodiment processes in art-based interventions, a recent systematic review of participatory arts programs for youth [89] identified embodied mechanisms, such as sensory engagement, bodily involvement, and expressive movement, as core pathways through which creative arts activities promote mental health. Increases in connection between Default and Control networks could be relevant to the “concretization” domain, by helping adolescents take abstract / ruminative thought content and express them through artistic ideas. The observed post-camp increases in Control-Default and Default-Salience connectivity could potentially relate to the third therapeutic factor domain of capacity for symbolism and metaphor, given prior work showed that novel metaphor production involves connectivity between default, control and salience networks [90].

Critically, these observed changes in brain connectivity in relation to the arts intervention are occurring against the backdrop of important ongoing brain development in adolescence. Key brain changes during the adolescent period include decreases in gray matter volume, increases in white matter volume, and refinement of neural connections [47, 91–93]. These ongoing developmental shifts introduce a time-sensitive window that represents both vulnerability (i.e., the increased risk of mental health disorders to emerge during this time period) and opportunity (i.e., increased malleability confers increased plasticity and amenability to intervention) [4]. Furthermore, adolescence is a critical time for exposure to the arts. Converging lines of evidence point to adolescence as a key window of time for creativity development [52]. The observation reported here of measurable change in brain network connectivity after an intensive 2-week intervention is encouraging, as it suggests that even short-term enrichment experiences like these can leverage brain plasticity in ways that could potentially increase the adolescents’ capacity to navigate their worlds in the recovery process.

Although we observed significant clinical improvements pre-to-post camp, and significant changes in connectivity between several large-scale brain networks pre-to-post camp, these clinical and network changes were not significantly correlated. On one hand, this null result could be due to low power; prior work has demonstrated that reliable brain-behavior associations using methods such as resting-state fMRI require very large samples [94]. On the other hand, it may be that these brain changes underpinned more complex phenomena and/or capacities that did not translate directly to (immediate) symptom improvement. We have previously written about how fostering an artist mindset (i.e., protect curiosity; go to discomfort; be bold; require play and exploration; suspend forming an opinion; share) may be useful for adolescents with depression as they navigate their mental health challenges and their recovery process [95]. Potentially, increases in connections between large-scale brain networks could facilitate some of these mindset changes in ways that are not fully captured by the questionnaires used in our study.

Limitations

Several limitations should be noted when considering these findings. First, the sample is small. The field is increasingly aware of the need for large samples to establish reliable brain-behavior relationships [96]. Second, while we were able to leverage an additional datapoint two weeks before camp for half the sample thereby approximating a “waitlist” control, given adolescent/family scheduling constraints the randomization was not fully implemented, which limits causal inference. Third, the study did not incorporate an active control condition. Randomized designs that incorporate an active control condition will be important for future studies to confirm the putative mechanisms studied here. Notably, designing the most appropriate active control condition is not straightforward and will require dedicated research to develop and validate the optimal approach. Fourth, for ethical reasons, participants were allowed to continue their current treatments, including psychotropic medications. This introduces the possibility that the brain changes reported here could have been influenced by ongoing treatments such as psychotropic medication or psychotherapy for some of the adolescents. A larger sample size will be required to furnish the power to incorporate appropriate modeling of other treatment effects in future studies. Fifth, although we did assess for substance use disorders at the beginning of the study (and none of our participants met criteria for one), we did not monitor for substance use during the study. If any adolescents were using substances during the time frame of their study participation, this could have influenced the results. Given the high prevalence of substance use in adolescents more broadly, future work may benefit from tracking and modeling the impact of substance use on the brain, with large enough sample sizes to afford the power to appropriately tease out these effects. Sixth, while our interpretation of the results speculate on creative processes such as embodiment, concretization and metaphor, these processes were not directly tested in the study. We hope that our results and interpretations will inspire future research to more directly test these creative processes as potential therapeutic mechanisms in experimental therapeutics-style clinical trials. Finally, we focused on four large-scale networks that have been implicated in depression and in creativity. The definition of networks can vary considerably according to different methodological choices [97, 98]. Enhanced clarity on how interventions impact these networks will be achieved as the field moves forward toward greater methodological and substantive consensus on these issues.

Conclusion

Arts interventions offer a promising avenue to address the significant public health threat of depression in adolescents. An advanced understanding of how they work is needed to guide research designed to optimize and tailor these interventions according to individual characteristics. The current findings suggest that after adolescents participated in an intensive intervention that fosters repeated practice of taking on the artist mindset and expressing oneself through art, depression symptoms improved and connectivity increased between key large-scale neural networks. This preliminary result suggests a novel target for creative arts interventions. Future work is needed to confirm this idea.

Author contributions

Dr. Cullen conceived of the study, obtained funding for the study, oversaw the conduct of the study, oversaw the analysis of the data and interpretation of the results, drafted the manuscript, and edited the manuscript. Mr. Roediger carried out the neuroimaging data processing and analysis, drafted sections of the manuscript, and edited the manuscript. Dr. Başgöze conducted the statistical analysis, drafted sections of the manuscript, and edited the manuscript. Ms. Padilla carried out the neuroimaging data acquisition, contributed to the conceptualization of the analysis, and edited the manuscript. Ms. Hayee contributed to data collection, drafted sections of the manuscript, and edited the manuscript. Dr. Klimes-Dougan contributed to the design and conduct of the study and the interpretation of the results, drafted parts of the manuscript, and edited the manuscript. Dr. Fiecas contributed to the design of the study, provided consultation on the statistical analysis, and edited the manuscript. Dr. Koutstaal contributed to the design and conduct of the study, including the theoretical approach to the analysis used here, contributed to the interpretation of the results, and edited the manuscript. Ms. Taniguchi co-designed the study, and contributed to the interpretation of results, and edited the manuscript. Dr. Mueller co-designed the study, oversaw the neuroimaging acquisition, processing and analysis, contributed to the interpretation of results, and edited the manuscript.

Funding

This study was supported by the University of Minnesota’s Minnesota Futures grant, the University of Minnesota Medical School, and the University of Minnesota Foundation.

Data availability

De-identified data from this study is not available in a public repository but may be shared with other researchers at the discretion of the principal investigator.

Code availability

The github link for the script that calculates the between network connectivity metric is as follows: https://github.com/droediger/CreativityCamp_BtwNetAnalysis.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41398-026-04352-4.