Transcranial focused ultrasound (tFUS) is a novel approach to brain stimulation that holds immense promise for the treatment of brain-based disorders (Pasquinelli et al., 2019; Stern et al., 2021). At high intensities, tFUS can be used to locally ablate brain tissue and provide irreversible treatment (di Biase et al., 2021; Moosa et al., 2019), and at low intensities, tFUS can be used to transiently modulate the functioning of specific brain regions by non-invasively applying acoustic energy (Tyler et al., 2008; Dell'Italia et al., 2022). Unlike other neuromodulation approaches, low-intensity tFUS can reach deep brain structures with high spatial precision to inhibit or enhance neural activation (Romanella et al., 2020). By manipulating ultrasound parameters, it is also possible to change brain function without causing tissue damage (Spivak et al., 2022). Low-intensity tFUS therefore overcomes many of the limitations of existing neuromodulation techniques and is being increasingly explored as a novel treatment approach for a variety of neurological and psychiatric disorders.

tFUS has been applied to a variety of brain regions, including somatosensory and visual cortices, insula, thalamus, and striatum (Legon, 2021; Monti et al., 2016; Cain et al., 2021). Recent studies have also begun to target the amygdala, a brain region known to mediate threat and emotion processing (Ohman, 2005; Folloni et al., 2019; Kuhn et al., 2023; Chou et al., 2024). Three decades of neuroimaging research indicate that hyperactivity of the amygdala is involved in the pathophysiology of internalizing disorders, including anxiety and post-traumatic stress disorder (Ressler and Nemeroff, 2000; Gerin et al., 2019). Early studies show that tFUS of the amygdala can modulate amygdala perfusion and alleviate symptoms of anxiety (Zielinski et al., 2021; Mahdavi et al., 2023). For example, a case study of an individual with treatment-resistant generalized anxiety disorder (trGAD) demonstrated that tFUS of the amygdala resulted in immediate decreases in anxiety symptoms (Zielinski et al., 2021). In a larger cohort of patients with trGAD, eight weekly tFUS sessions targeting the amygdala similarly resulted in significant decreases in anxiety symptoms, with 64% of patients achieving clinically significant benefit (Mahdavi et al., 2023). Converging research therefore suggests that tFUS has therapeutic potential and the safety, efficacy, and neural mechanisms of tFUS should be further explored.

How tFUS modulates neural processes is still unclear (Dell'Italia et al., 2022). Studies indicate that tFUS increases or decreases blood oxygen level-dependent (BOLD) signal depending on sonication parameters (Kuhn et al., 2023; Chou et al., 2024). Indeed, a study by our group found that tFUS of the amygdala selectively increased perfusion in the right amygdala but not the left amygdala without engaging the auditory cortex (Kuhn et al., 2023). How these focal changes influence the rest of the brain is still an area under active investigation. Using BOLD data collected simultaneously with the tFUS sonication experiment, we found decreases in functional connectivity between the right amygdala and posterior cingulate, anterior cingulate, medial prefrontal, and posterior parietal regions in a sample of healthy older adults (Kuhn et al., 2023). A separate study by Chou et al. (2024) examined changes in resting-state functional connectivity (rsFC) in healthy adults before and after active or sham tFUS of the left amygdala. Chou et al. (2024) reported that active tFUS resulted in decreased amygdala-insula and amygdala-hippocampal rsFC and increased amygdala-ventromedial prefrontal cortex rsFC. Meanwhile, a recent study by Peng et al. (2024) found that tFUS of the left nucleus accumbens was associated with increased rsFC between the nucleus accumbens and the medial prefrontal cortex.

Existing studies provide important initial evidence that tFUS changes FC between the sonication target and other areas of the brain. However, no study to date has directly examined tFUS-related changes to neural networks beyond the target. The brain is organized into multiple distributed (large-scale) systems, including the default mode network (DMN), dorsal attention network (DAN), frontoparietal network (FPN), and salience network (SN) (Yeo et al., 2011). Synchronized activity within each network is observed under resting conditions and posited to underlie specific cognitive-affective functions, including interception and attentional control (Hellyer et al., 2014). Converging evidence from other techniques suggests that neuromodulation can affect brain networks beyond the focal stimulation target (To et al., 2018; Pini et al., 2018; Santarnecchi et al., 2018). Thus, in order to fully elucidate the neuroplastic changes associated with tFUS, it is necessary to investigate “down-steam” changes within and between well-characterized intrinsic neural networks.

In the present study, we collected resting-state BOLD fMRI before and after tFUS to characterize changes in rsFC. We used a sonication paradigm adapted from previous studies to inhibit/disrupt amygdala activity (Spivak et al., 2022; Folloni et al., 2019). We also included a second (within-subjects) target as an active regional comparison—the entorhinal cortex (ErC). The ErC is implicated in memory formation (Montchal et al., 2019) and hypoactive in diseases characterized by memory disturbance, e.g., Alzheimer’s disease and mild cognitive impairment (Igarashi, 2023). In a randomized, double-blind, within-subjects crossover study design, we enrolled participants to complete two tFUS sessions, separated by a 14-day between-session window. The sonication paradigm for left ErC was hypothesized to excite/stimulate and therefore increase ErC activity and connectivity (Spivak et al., 2022; Dell'Italia et al., 2022). In a previously published study of these participants, we demonstrated that the sonication protocols selectively increased perfusion in the targeted region, but not in the contralateral homolog or either of the bilateral control regions (Kuhn et al., 2023). Data were collected and examined blindly to determine the target-specific network changes in rsFC pre-to-post sonication. We broadly expected decreased rsFC connectivity within and between network nodes following the sonication of the amygdala target and increased rsFC connectivity within and between network nodes following the sonication of the ErC target based on the specific tFUS parameters.

The current study used resting-state BOLD fMRI data before and after tFUS to examine changes in functional connectivity between nodes of large-scale rs-networks. The results revealed that tFUS of the amygdala (and ErC) changed connectivity within and between rs-networks. Pre-to-post sonication of the right amygdala was found to modulate connectivity within nodes of the SAN and between nodes of the SAN and the DMN and FPN. A decrease in the SAN to FPN connectivity was specific to the amygdala target. Pre-to-post sonication of the left ErC was found to modulate connectivity between the DAN and FPN and DMN. An increase in the DAN to DMN connectivity was specific to the ErC target. These preliminary findings suggest that tFUS may result in neuroplastic changes outside of the focal neural target. Previous studies have shown that low-intensity tFUS does not work via thermal or tissue-damaging mechanisms (Stern et al., 2021; Spivak et al., 2021). The ultrasound waves have a mechanical effect, which changes the membrane potential of neurons in the target region (Tyler, 2011). More broadly, low-intensity tFUS impacts synaptic transmission and changes synaptic efficiency, which may influence the observed changes in functional connectivity beyond the sonication target (Blackmore et al., 2019).

tFUS of the amygdala was associated with changes in connectivity within the SAN and between the SAN and the DMN and FPN. The SAN is involved in the integration of emotionally salient information and makes inferences regarding interoceptive awareness, threat/reward value, and outcome probabilities to appropriately guide approach and avoidance behavior (Seeley et al., 2007; Vytal and Hamann, 2010). Relatedly, the functional integrity within the SAN underlies the subjective experience of affective states (De Witte and Mueller, 2017; Prillwitz et al., 2018). Theory and research further show that the SAN interacts with other large-scale rs-networks and plays a critical role in the dynamic switching between the DMN and the FPN (Sridharan et al., 2008; Goulden et al., 2014). This switching function allows for efficient engagement and disengagement of goal-directed resources. Upon detection of a salient event, the SAN facilitates access to attention and working memory via the FPN (Cai et al., 2021). During rest, the SAN engages the DMN, which supports basic self-referential functions and internally focused attention (Yoon et al., 2019). Regarding the present findings, acute tFUS-related increases in rsFC within the SAN may reflect increased salience processing. Meanwhile, increased SAN to DMN rsFC and decreased SAN to FPN rsFC may signal changes in the SAN switching functions between networks. The decreased rsFC connectivity between the SAN and FPN was unique to the amygdala target, and thus, amygdala inhibition may uniquely decrease SAN input to the FPN. Numerous studies have demonstrated that anxiety and other stress-related disorders are characterized by aberrant SAN function, including deficient network switching (e.g., Li et al., 2023). Additional studies are therefore critically needed to replicate these preliminary findings and elucidate how tFUS-related SAN changes may impact anxiety symptoms.

tFUS of the ErC was used as an active within-subjects comparison. Sonication of the ErC resulted in pre-to-post changes in the DAN to FPN and DMN rsFC. The DAN is comprised of regions in the frontal and parietal cortices (Osher et al., 2019), which become engaged when attention is voluntarily shifted to salient objects and/or locations (Corbetta and Shulman, 2002) and during visual exploration (Corbetta and Shulman, 1998). The DAN also underlies attentional control via top-down influences on the visual cortex (Osher et al., 2019). Anti-correlation between the DAN and DMN is characteristic of typical brain function (Fox et al., 2009; Kelly et al., 2008), which reflects the distinct attentional processes these two networks serve: the DMN mediates internally directed attention, whereas the DAN mediates externally directed attention (Spreng et al., 2017). The magnitude of anti-correlation between the DAN and DMN has been associated with the performance of attention-based tasks in healthy young adults (Hampson et al., 2010), though this association is attenuated by age and seems to differ in patients (Wang et al., 2019; Spreng et al., 2016; Owens et al., 2020). In at least one previous study in adults with depression, positive connectivity between the DAN and DMN was associated with increased memory accuracy for objects (Satz et al., 2022). Interestingly, in a sample of older adults, we found that sonication of the ErC resulted in increases in DAN to DMN rsFC, and this increase was specific to the ErC target. We also found increases in the DAN to FPN rsFC, which are two networks that functionally interact to support perceptual attention (Dixon et al., 2018). Although the functional significance of these acute changes is unknown, it is noteworthy that the DAN, DMN, and FPN are all involved in the cognitive processes that mediate learning and memory—core functions of the ErC target. Thus, the ErC as a target for tFUS in patient groups with amnestic syndromes appears to warrant further exploration.

The pattern of results indicates that some FC changes were more robust following the sonication of one active target but not the other. Sonication of the amygdala target using an inhibition protocol resulted in decreases in rsFC between the SAN and FPN. Meanwhile, sonication of the ErC target using an excitation protocol resulted in increases in the DAN to DMN rsFC. These findings are broadly consistent with our study hypotheses and demonstrate that tFUS can have “downstream” effects that alter rsFC patterns within and between large-scale networks. Other neuromodulation techniques such as deep brain stimulation and transcranial magnetic stimulation have shown similar extended effects (Haslinger et al., 2003; Coenen et al., 2014; Esser et al., 2005). It is noteworthy that changes involving the SAN were exclusively found with tFUS of the amygdala, given that the SAN is the core rs-network involved in salience and emotion processing and is implicated in the pathophysiology of anxiety disorders (Pannekoek et al., 2013). Investigating if and how these SAN-level changes influence affective states and anxiety symptoms is a critical next step, particularly given existing research highlighting the potential clinical utility of tFUS of the amygdala for anxiety (Chou et al., 2024). Meanwhile, the ErC is a brain region involved in memory (Guzman et al., 2013; Eichenbaum, 2000), and tFUS of the ErC modulated DAN rsFC with other attention networks. Attention and memory disturbance, along with associated hypoactivity of the ErC, underlies several neurological disorders, including Alzheimer’s disease (Donix et al., 2013; Maass et al., 2018). Given the dissociative findings observed here, an investigation into the functional significance of DAN-level changes, as they pertain to cognitive processes like memory, may also be a fruitful next step.

The current study had many strengths, including the double-blind, within-subjects crossover design. The study also had several important limitations. First, the study focused on healthy older adults given that there is a need for novel psychiatric and neurological therapeutic strategies for this developmental period. Previous research shows that there are age-related changes in rsFC between large-scale networks, including those investigated in the current study (Grady et al., 2016; Meunier et al., 2009). It is therefore unclear whether the present findings would generalize to younger adults. Second, the sample size was small, although consistent with other published studies involving tFUS in humans (Kuhn et al., 2023; Pasquinelli et al., 2019). It is possible the study was underpowered to detect certain target-specific network-level changes. Relatedly, a total of 171 ROI-to-ROI connections were initially tested, and the findings would not survive stringent correction for multiple comparisons. The length of each resting-state scan was also short, which studies show may decrease the reliability of within-subjects comparisons (Mueller et al., 2015), though our analyses comparing the two baseline target scans showed good reliability. The current findings are therefore considered preliminary and require replication. It is also important to highlight that we targeted contralateral regions (the left ErC and right amygdala) given the close physical proximity of these two regions in the brain. It is possible that the pattern of results may differ if the hemisphere of each target was switched, given prior research on the lateralization of certain networks (Agcaoglu et al., 2015). Finally, the study focused on immediate, acute changes (pre-to-post sonication) and did not measure real-time functional outcomes. Additional studies are needed to determine the duration and durability of the observed findings and if and how network-level changes relate to cognitive and emotional outcomes.

The present findings add to a growing body of literature on the feasibility of tFUS and its acute neural effects. Converging research, including our own, shows that tFUS can target desired areas of the deep brain without engaging nearby structures (Kuhn et al., 2023). We also demonstrate that tFUS can evoke broader changes beyond the focal sonication target. There are many important next steps in this line of work, including the investigation of the functional significance and therapeutic value of these network changes.