Low intensity TUS in animal models has been shown to interact with neural tissue via mechanical effects. The sonication wave either directly changes the permeability of ion channels within neuronal membranes, such as voltage-gated sodium, calcium and potassium channels (e.g., K2P, TRP and Piezo1), or it temporary mechanically alters the cell membrane properties (Liao et al., 2021). Several mechanisms have been proposed, including changes in membrane turgidity and the dynamics of lipid microdomains (Anishkin et al., 2014; Babakhanian et al., 2018; Petersen et al., 2016; Suki et al., 2020; Tyler, 2011). TUS also affects the coupling between neurons and glial cells (Oh et al., 2019). The combination of TUS mechanical effects tends to increase action potentials being elicited by excitatory and inhibitory neurons (Tyler, 2011; Yoo et al., 2022). TUS has been shown to be capable of inducing muscle contraction and limb or tail flicking when rodent motor cortex is stimulated with low to moderate intensities (Kim et al., 2020; Lee et al., 2018; Tufail et al., 2010). Similar motor responses have yet to be observed and reported in human and non-human primates (Darmani et al., 2022).

At low intensities for neuromodulation, TUS can influence neural tissue without causing substantial damage, heating or adverse effects, as reported in human and non-human primates (Gaur et al., 2020; Spivak et al., 2021; Verhagen et al., 2019). However, care should be taken with continuous stimulation protocols where the continuity of stimulation (duty cycle; see Box 1) is high (Roumazeilles et al., 2021; Verhagen et al., 2019). Overall, TUS does not appear to cause significant heating or cavitation of brain tissue when the intensity remains low, based on Mechanical and Thermal Index values and recommendations of use (see Box 1 and 2). Temperature changes for low-intensity TUS are commonly <1°C (Baek et al., 2017; Yoo et al., 2011), and thermal effects could at least minimally alter cell membrane capacitance during ‘online’ TUS. However, thermal effects alone are unlikely to play a considerable role for longer-lasting ‘offline’ TUS effects (Ozenne et al., 2020; Verhagen et al., 2019). The mechanism of action for the longer-lasting offline effects is not yet well understood. Because the offline effects can last tens of minutes, or in some cases hours, after the sonication period (Bault et al., 2023; Pasquinelli et al., 2019), offline effects likely engage neuroplasticity mechanisms, such as modulation of AMPA and NMDA glutamatergic receptors and post-synaptic Ca²⁺ mediated changes to receptor properties. Interestingly, TUS effects on neuronal NMDA receptors appear to be indirect via, for instance, TUS modulation of astrocytes that can influence neuronal plasticity (Blackmore et al., 2023). TUS pulsed at a theta (4-8 Hz) frequency (theta-burst TUS; tb-TUS) is being studied for its capability to induce LTP-like plasticity (Darmani et al., 2025; Oghli et al., 2023; Samuel et al., 2023; Samuel et al., 2022; Zeng et al., 2022). We further consider tb-TUS as part of ‘offline’ stimulation protocols at the end of section (I) below. Of importance for future clinical applications, the repeated use of TUS sessions does not appear to negatively impact on the integrity of brain tissue as assessed by MRI (Munoz et al., 2022).

There is substantial interest in understanding the conditions under which TUS could be used to bias the directionality of neuromodulatory effects on the targeted brain area and its network or on behavior (Blackmore et al., 2023; Mihran et al., 1990; Tsui et al., 2005; Zhang et al., 2023). To describe the directionality of neural effects, we use the terms enhancement or suppression throughout, reserving the terms excitation and inhibition for references to reports where it was possible to directly record from identified excitatory and inhibitory neurons with animal models.

Recordings from identified excitatory and inhibitory neurons during TUS with animal models provide important mechanistic insights, because the neuronal recordings can be combined with causal manipulation, such as blocking specific ion channels. For instance, previous studies with murine models have reported that short Sonication Durations (SD, <1 sec) can lead to net neural excitation (attributed to more action potentials being elicited by TUS in excitatory neurons), whereas longer sonication durations (> 1 sec) can lead to net neural suppression (i.e., more strongly driving inhibitory neurons) (Mihran et al., 1990; Tsui et al., 2005). Other TUS studies have suggested that higher sonication Pulse Repetition Frequencies (PRF >100 Hz) can lead to net excitation (Manuel et al., 2020; Zhang et al., 2023). The caveat is that many nonhuman animal studies are conducted under anesthesia, which can alter the balance of excitatory-inhibitory neuronal activity. By comparison, although human TUS studies are often conducted without anesthesia, access to single units (neurons) is only possible with specialist FDA or ethical board approved electrodes for clinical monitoring in neurosurgery patients, with which it is difficult or not possible to identify the recorded cell types. Moreover, there is a paucity of direct neuronal recording studies in humans during TUS.

Researchers with other non-invasive brain stimulation approaches, including TMS, are working to overcome similar challenges in identifying the directionality of effects on neurons and neuronal networks (Fitzgerald et al., 2006). TMS researchers now regularly apply higher duty cycles to tip the directionality of TMS neuromodulatory effects on cortical areas towards net excitation (i.e., potentiation). By contrast, low-duty cycle TMS pulses are associated with net inhibition (i.e., de-potentiation or suppression) of muscle potentials or motor cortical responses (Solomon et al., 2024). Therefore, although the effects of TMS and TUS on neurons and neural systems differ, a certain level of correspondence between the two techniques may be found, even if it is not expected, in parameters such as duty cycle causing net excitation or suppression of neural network function.

Research into TUS mechanisms and effects is informing computational modeling, which allows the more systematic exploration of TUS stimulation parameters, difficult to achieve with any one empirical study alone. In a computational Neuronal Intramembrane Cavitation Excitation (NICE) model developed to study activation and suppression effects on modeled excitatory and inhibitory neuronal populations, TUS effects were simulated as intramembrane cavitation causing changes in ion channel conductivity (Plaksin et al., 2016). Although it remains questionable whether intramembrane cavitation is a key mechanism for TUS, the NICE model stimulation simulation approach can be viewed as general and it allowed the researchers to simulate a broad set of TUS parameters, including TUS intensity and the continuity of stimulation (duty cycle), on modelled neuronal responses.

Box 1 summarizes the common TUS parameters and their measuring units. Key parameters are the average acoustic intensity (intensity spatial peak pulse average, ISPPA), intensity spatial peak time average (ISPTA), sonication duration, duty cycle (DC), pulse repetition frequency (PRF), thermal index (TI) and mechanical index (MI). Box 2 shows guidelines on the ultrasound parameter limits that human low-intensity TUS studies typically follow.

The NICE model was initially evaluated with a more limited set of the previously available data from human and nonhuman animal studies, and the model at that time showed a high level of predictability, albeit with a very small sampling of studies available for comparison. Increases in intensity (ISPPA, Box 1) and duty cycle (DC, Box 1) were argued to be able to tip the balance from suppression to activation in the modeled populations of excitatory and inhibitory neurons (Plaksin et al., 2016). Several reviews have now conducted case-by-case and ad-hoc comparisons of TUS parameters with the NICE model predictions, with mixed support for the NICE model (Ai et al., 2018; Dell’Italia et al., 2022; Forster et al., 2023b; Zhang et al., 2023). In Box 3, we collate a set of net enhancement versus suppression hypotheses linked to TUS parameters that may be able to bias the directionality of neural effects. These hypotheses are generated by the animal research studies discussed in this section. Therefore, the hypotheses are not solely generated by the NICE model and are further considered in Section I on the meta-analysis and resource provided as part of this report.

The uncertainty about the extent to which TUS can be used to enhance or suppress neurobiological function limits its research potential (Chen et al., 1997; Fitzgerald et al., 2006). Rather than exploring the TUS parameter space, many researchers opt to emulate the TUS parameters of prior studies reporting specific positive findings, limiting the necessary exploration of the entire parameter space necessary for a nascent field and contributing to an empirical and reporting bias. We recognize the complexity of neural circuits and systems and the limitations in aiming to evaluate predictions with a relative paucity of data in humans. However, we also recognize that stepwise progress and evaluation are needed as signposts in this research endeavor, not unique to TUS or other brain perturbation approaches with longer histories of use, e.g., invasive deep-brain electrical stimulation, TMS, tACS, tDCS; see Fig. 1 (Derosiere et al., 2020; Klink et al., 2021). Therefore, since there are now many tens of human TUS studies (Figure 3), to us it seems the time is ripe for further research signposts and a database to be established as an open resource that can help to accommodate growth in the field. There are now a range of reported behavioral and neurobiological outcomes with human TUS, ranging from eliciting somatosensory sensations with TUS applied to the somatosensory cortex, the enhancement or suppression of the threshold for motor-evoked potentials (MEPs) with TUS applied to motor cortex (including in combination with TMS), the perception of visual phosphenes or modulation of visual motion perception from TUS applied to the visual cortex, and mood improvement induced by TUS to the prefrontal cortex (these and others are summarized in Tables 1-3). With the accumulation of these and other human low-intensity TUS data, a more extensive review and meta-analysis than previously possible can now be conducted. This could be a needed step towards the next key evaluation period when samples sizes increase.

Our key objectives with this report are twofold. In Part I, we evaluate the initial set of collated net neural enhancement versus suppression hypotheses (Box 3), and we conduct an initial meta-analysis of the available healthy human participant low-intensity TUS reports. We conclude Part I by establishing an open Iowa-Newcastle (inTUS) resource, including a database that can grow with TUS research community input and an analysis script that can (re)analyze the data. The inTUS resource is offered to encourage TUS researchers to report outcomes on a broader and more systematic exploration of the TUS stimulation parameter space. The reporting approach that we use is as recommended by the International Transcranial Ultrasonic Stimulation Safety and Standards (ITRUSST) consortium, which has proposed standards to enable comparison and reproducibility across studies (Martin et al., 2024), also see https://itrusst.com.

In Part II of this report, we selectively summarize the current state of the literature on human TUS applications for perturbing the brain and developing treatments for neurological and psychiatric disorders. This literature review is not intended to be a comprehensive review of the TUS literature on clinical applications. More comprehensive reviews of this literature can be found elsewhere (Matt et al., 2024; Pellow et al., 2024). Instead, this selective summary of TUS in specific clinical disorders aims to illustrate how hypotheses on the directionality of TUS effects need to be patient and brain area/system specific. Therefore, at the conclusion of each of the Part II sections, we motivate testable hypotheses on how TUS could be employed to enhance or suppress function for that patient-specific application. We encourage the reader to continue to develop hypotheses, not limited to the examples provided here.

The hypotheses summarized in Box 3 are based on TUS parameters that have been highlighted in the TUS literature to be capable of biasing TUS effects on a given brain region towards neural enhancement or suppression. These hypotheses were generated both from the NICE model (Plaksin et al., 2016; Plaksin et al., 2014) and preclinical studies with animal models. The preclinical studies of TUS effects have reported conditions under which TUS could potentially bias the directionality of effects towards greater excitation or inhibition. The NICE model hypothesized that two key parameters associated with net activation or suppression (using the authors’ terminology) are sonication intensity in the target brain area (ISSPA in brain) and the continuity of stimulation (duty cycle, DC). The NICE model predictions are shown in Figure 4 with a light blue line defining the border between neural enhancement (higher DC and intensity) and suppression (lower DC and intensity). Other parameters of interest are ISPTA, which mathematically multiplies ISPPA by the continuity of sonication (DC), and the length of the sonication pulse (Sonication Duration, SD). Shorter SDs (<500 ms) have been reported to elicit more action potentials from excitatory neurons, and longer SDs (>500 ms) tend to elicit suppression via greater excitation of inhibitory neurons (Mihran et al., 1990; Tsui et al., 2005). Other studies have suggested that pulse-repetition frequency (PRF), the frequency with which the ultrasound pulse is turned on/off, can bias towards greater net excitation at higher PRFs (Kim et al., 2023; Yu et al., 2020). These hypotheses are summarized in Box 3.

Studies of online or offline effects tend to use different TUS parameters (compare Tables 1 and 2). Offline effects of TUS stimulation are those induced for longer time periods after the sonication period (seconds or minutes). Because many human studies aim to keep the intensity of offline stimulation within FDA and/or ITRUSST guidelines (Box 2), a different parameter range for online and offline studies may be implemented by TUS researchers. Namely, for offline studies, DC and ISPTA values are often kept low (Figure 4B). For this reason, we summarize the online and offline studies in separate tables (Tables 1-2) and include this distinction as a factor in the database meta-analyses.

The meta-analysis approach follows the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines (Figure 3). We searched PubMed/MEDLINE for studies published through 1^st January 2025 by employing the combination of the following keywords: ‘human’, ‘ultrasound’, ‘focused’, ‘low-intensity’, ‘stimulation’, ‘transcranial’, ‘neuromodulation’, ‘TUS’, ‘FUN’, ‘LIFUS’ ‘clinical’, ‘treatment’ (see Figure 3 for the papers in the initial sample). Four authors (HC, NS, CP and MZ) searched for and curated the included studies to ensure that the survey was as comprehensive as possible. We searched only for articles in English. Study inclusion criteria for the meta-analysis focused on human studies involving low-intensity TUS for brain stimulation or neuromodulation. The included studies encompassed only healthy human participants to avoid including clinical studies of patients with brain pathology at this stage of the database development, although we intend to include patient clinical studies in future versions of the database. We excluded studies with moderate-intensity or high-intensity ultrasound used for, respectively, BBB perturbation or thermal ablation, outside of the scope of the current version of the database. Given the currently small size of the database, we expected the analyses not to be robust to missing data. For this reason, we had to exclude 4 diagnostic ultrasound studies (Gibson et al., 2018; Guerra et al., 2021; Hameroff et al., 2013; Schimek et al., 2020), which did not report the parameters we needed for the meta-analysis. After applying the inclusion and exclusion criteria, 32 studies were included in the first inTUS_DATABASE_6-2024, and 47 studies were included in the next key expansion of the database: inTUS_DATABASE_1-2025. Further key database expansions, analyses and results will be reported in the inTUS resource location noted below. For the meta-analysis, we only included studies that either reported a basic set of TUS stimulation parameters or those sufficient for estimating the required parameters necessary for the meta-analysis.

The reported studies’ TUS parameters and reported effects were used to populate the data tables (Table 1 for online studies; Table 2 for offline studies). Most parameters required for the analysis could be found in the reported studies or calculated from the parameters given. If a given study conducted multiple complete experiments, the sample sizes reflect the overall number of experiments rather than the number of studies/papers, and if separate experiments tested different values of a given parameter with the same result, the experiment eliciting the strongest effect was input into the database for meta-analysis. A number of experiments (n = 14 out of 37 in the inTUS_DATABASE_6-2024, and an additional, n = 13 out of the 15 additional experiments added in the inTUS_DATABASE_1-2025), although reporting ISPPA values in water, did not report ISSPA values in the brain, required for our analyses, which is a recognized problem for this field (Martin et al., 2024). For these studies, we applied an accepted approximation of ISPPA values, because the sonic wave passes through and loses much of its energy at the skull, whereby typically 70-75% of the intensity is lost (Fomenko et al., 2020; Lee et al., 2015; Mueller et al., 2014; Oghli et al., 2023; Xia et al., 2024; Zeng et al., 2024). We compared these approximated values entered into the database to simulations using k-plan software (Jaros et al., 2020; Treeby & Cox, 2010), targeting the same regions as in the reported experiments with the reported ISPPA in water values. Comparing the k-plan simulation values to the estimated values showed a low margin of error of ∼5% between the two sets of values in the comparisons. Therefore, we used the approximated values for studies not reporting ISPPA in the brain.

Probable net neural enhancement versus suppression was characterized as follows. Note that our use of the terms enhancement and suppression refers exclusively to the increase or decrease of neural activity, respectively, as measured by, neurophysiological methods (EEG-ERPs, BOLD fMRI, etc.) and do not imply equivalent changes in behavioral responses. Although many studies have reported behavioral influences, these alone were not sufficient for us to determine the probable directionality of underlying neural TUS effects. Although many TUS studies report using a sham control (e.g., no TUS or an inverted TUS transducer that directs the sonic energy away from the head), it is difficult to rule out other sources for placebo effects in the behavioral reports. We thereby focused on the studies’ reporting of neurobiological effects and characterized these effects as probable net neural enhancement versus neural suppression, using the following approach. For net enhancement, we followed the prior approach established by the TMS field, whereby EEG-evoked responses that are magnified in the target area as a function of TUS application could be characterized as probable enhancement (see Tables 1-2). We included positive fMRI BOLD effects resulting from TUS as a probable enhancement. For suppression, we also followed the prior approach established by the TMS field, whereby EEG evoked responses that were reduced indicate likely suppression of neural activity. Wherever possible, we relied on independently characterized directionality of effects, and we cite the original sources that conducted the characterization in Tables 1 and 2. As an example of a study categorized as ‘neural suppression’ of function, Legon and colleagues (Legon et al., 2014) examined TUS targeting the primary somatosensory cortex (S1) in healthy human subjects with EEG. The authors reported that TUS significantly attenuated somatosensory evoked potentials. The effects were specific to the targeted region, because the changes were abolished when the acoustic beam was focused away from S1. As another example, Liu and colleagues applied TUS to S1 of participants performing a sensory discrimination task, and they reported augmented somatosensory spatiotemporal EEG responses, which are interpreted as increased local excitability or ‘neural enhancement’ by our terminology (Liu et al., 2021).

The resulting database tables were submitted to an exploratory univariate logistic regression model tested with R Studio. The R markdown script used to generate the results from the data tables is shared as part of the inTUS resource below. The sample sizes in the initial inTUS_DATABASE_06-2024 were 37 experimental observations of each of the TUS parameters in the database. The sample sizes in the subsequent inTUS_DATABASE_01-2025 were 52 experimental observations. The first statistical model tested the Enhancement versus Suppression Effect separately for ISPPA, Duty Cycle, PRF and SD from the hypotheses with the Mode of stimulation ‘online’ or ‘offline’ as a covariate (e.g., logit ∼ OfflineOnline + Isppa, etc.). A single model with all factors and all interactions would have been preferred, but with these sample sizes is not sufficiently powered to provide meaningful results. This will be revisited in the future when sample sizes increase through the database expansion via the inTUS resource. For the second key database expansion period (inTUS_DATABASE_01-2025), we also conducted an initial multivariate pattern analysis using a Random Forest Walk analysis implemented in python (Scikit-learn: Machine Learning in Python) (Pedregosa et al., 2011).

The meta-analysis used the data in Tables 1 and 2 from the inTUS_DATABASE_1-2025.csv. The tables summarize the range of TUS parameters of interest for the studies reporting probable enhancement or suppression of TUS effects. The rationale for characterization of the directionality of TUS effects, independently evaluated wherever possible, is summarized in Tables 1-2 in the rightmost column. The database is further separated by studies aiming to elicit online (Table 1) or offline effects (Table 2). Further key evaluation periods of the inTUS database, as it expands, may not be reported with updates to this paper, but will continue to be reported online in the inTUS resource. The statistical models, statistical output variables and resulting figures are available and can be recreated in the inTUS resource for any of the datasets available via the R markdown script.

This section reports the results from the first database analysis (inTUS_DATABASE_6-2024.csv), with a sample size of n = 37 experiments (from 33 studies). The statistical model for DC was significantly associated with the Enhancement versus Suppression, our effect of interest (p = 0.046). Sonication duration (Thurman et al.) was also significantly associated with the effect of interest (SD: p = 0.040). Consistent with the hypotheses in Box 3, higher DC values were significantly associated with Enhancement (n = 21; mean DC during Enhancement = 29.952, standard deviation (STD) = 23.252), lower values with Suppression (n = 16; mean DC during Suppression = 16.719; SD = 17.159). Also consistent with the hypotheses in Box 3, lower SD brain values were significantly associated with Enhancement (n = 21; mean SD = 14.6190, STD = 28.736) and higher values with suppression (n = 16, mean SD = 29.612, STD = 41.369). A statistical trend was present for the Online and Offline covariate factor (p = 0.061), as might be expected given the different parameters that are often used for online and offline studies (Figure 4a-b).

There was no significant effect for ISSPA brain (p = 0.256), and the other parameters of interest were also not significantly associated with our effect of interest in this dataset: pulse repetition frequency (PRF: p = 0.324) or ISPTA (p = 0.787). Therefore, in this initial dataset evaluation, only DC and SD were significantly associated with Enhancement versus Suppression in the hypothesized direction (Box 3). No multivariate pattern analyses were attempted with this dataset.

This section reports the results from the second key database expansion meta-analysis (inTUS_DATABASE_1-2025.csv), with a sample size of n = 52 experiments (from 47 studies) as shown in Tables 1-2. The exploratory univariate statistical model for DC (p = 0.077) and for ISSPA brain (p = 0.097) showed a statistical trend in their association with the Enhancement versus Suppression effect of interest. Consistent with the hypotheses in Box 3, higher DC values trended in association with Enhancement (n = 27; mean DC during Enhancement = 25.333, standard deviation (SD) = 22.231), lower values with Suppression (n = 25; mean DC during Suppression = 17.630; SD = 16.052). Inconsistent with the hypotheses in Box 3, lower ISPPA brain values trended in association with Enhancement (n = 27; mean ISPPA brain during Enhancement = 5.276 W/cm², standard deviation (SD) = 4.696), higher values with Suppression (n = 25; mean ISPPA brain during Suppression = 9.221; SD = 10.718).

The Online and Offline study covariate factor was not significantly different in any of the statistical models (p > 0.1). Rerunning the statistical models separately on the Online and Offline effects did not result in any of these parameters showing a significant association or a statistical trend, consistent with the effects not being statistically distinguishable in the current database (1-2025). This might be either because the sample sizes are not sufficiently powered for statistical effects to be observed or the TUS community might be converging somewhat in the TUS parameters used for online or offline stimulation (Fig. 4). The remaining parameters of interest were not significantly associated with the Enhancement versus Suppression effect of interest: PRF (p = 0.324) and sonication duration (SD: p = 0.326).

For this database, we conducted an initial multivariate pattern analysis using a Random Forest Walk procedure (53 observations, 4 variables: PRF, DC, SD, ISPPA brain, from Tables 1 and 2). The mean accuracy of effect classification (200 times 5-fold cross-validation) with all these variables was 0.55 (SD = 0.14), compared to analysis with a shuffled target variable (TUS effect directionality: Enhancement vs Suppression; mean 0.48 and SD = 0.15), see Figure 6. Although the accuracy of the database predicting the effect of interest is numerically above chance, the SD values overlap the original and shuffled data, and the permutation importance of the parameter values (Figure 6, right panel), are low (between 0.04 and –0.06). These observations suggest that the database sample size is too small to provide a meaningful result with this multivariate analysis at this stage. The classification accuracy was similar when all variables from Tables 1-2 were included in the analysis (8 variables: FF, PRF, DC, SD, PD, ISPPA_water, ISPPA_brain, ISPTA_brain; classification accuracy 0.57, SD = 0.14; shuffled accuracy 0.50, SD = 0.14).

Given the still limited sample size of human TUS studies to date, we interpret these meta-analysis results with caution. A key observation is that the NICE model is not as strongly predictive as initially thought (Dell’Italia et al., 2022; Zhang et al., 2023). Nonetheless, in the exploratory univariate analyses, DC and ISPPA are statistical trends for an association with the categorical Enhancement versus Suppression variable in the database. The DC effect is in the hypothesized direction. However, the ISPPA_brain effect is not (see Box 3; compare Figure 5 for DC and ISSPA_brain). Namely, consistent with the hypotheses in Box 3, higher DC values trended in association with Enhancement and lower values with Suppression. Inconsistent with the hypotheses in Box 3, lower ISPPA brain values trended in association with Enhancement and higher values with Suppression (see Figure 5). The multivariate pattern analyses with the four parameters of interest classified at an accuracy level close to chance and the factor importance values were too small to be meaningfully interpreted (Fig. 6). In relation to the initial dataset results (inTUS_DATABASE_6-2024), the DC factor which was significant in the first is now a trend in the second analysis. The ISPPA_brain factor was not significant previously with the univariate analyses, but is a trend with the second dataset evaluated (inTUS_DATABASE_1-2025). Therefore, the effects are still labile and may change, warranting caution in over-interpretation at this stage.

Theta-burst TUS (tb-TUS) is being studied for its capability to induce cortical LTP-like plasticity for an extended period of time (Darmani et al., 2025; Oghli et al., 2023; Samuel et al., 2023; Samuel et al., 2022; Zeng et al., 2022). Theta-burst TUS protocols are identified in Table 2. We observe that tb-TUS protocols tend to consist of a more continuous stimulation approach than online protocols, such as 80s trains of 20ms sonication pulses spaced over 200ms (i.e., pulsed at a 4-8 Hz theta frequency). However, independent replication of these findings remains limited. For example, Bao et al., found reduced motor cortex excitability – measured as decreased TMS-MEP amplitude in M1 -- that lasted up to 30 minutes post- sonication (Bao et al., 2024). Whereas Fong et al. reported no significant effects between tbTUS and sham conditions in M1 excitability (Fong et al., 2024). These studies, although growing, are still too few to consider separately, and as such are currently included as ‘offline’ studies in the database. Of note, some groups are seeing enhancement and others disruption with their tb-TUS protocols (Table 2). As the number of tb-TUS studies grows in the database, it will be important to evaluate tb-TUS outcomes and compare these to other TUS stimulation protocols.

A key limitation of this meta-analysis is the relatively small sample size. Multi-variate analyses of these data could have more predictability with greater sample sizes to provide deeper insights into the interplay of TUS parameters and effects. Moreover, some effects with certain parameters of interest have statistically changed between the initial (6-2024) and second (1-2025) key dataset evaluation periods reported here. Therefore, as noted above, the effects remain labile but may stabilize with future dataset expansion and evaluation periods. Currently it is unclear which if any TUS parameters or sets of parameters together will be good predictors for controlling the directionality of TUS effects. Another limitation is the inherent selection bias of retrospective studies, whereby researchers may limit their exploration of the TUS parameter space based on studies with positive findings. Researchers may also be limiting their brain targeting options, and there may be a positive finding reporting bias. Moreover, we and others have noted that not all TUS researchers share the full set of key parameters necessary for secondary hypothesis testing, even though the ITRUSST consortium has recommended that TUS studies report a basic set of parameters like those included in the inTUS database (Martin et al., 2024). For instance, we had to simulate the derated ISPTA values in the brain, which although comparable to our k-plan simulation values warrant caution when interpreting results. Therefore, the results from the 6-2024 and 1-2025 databases analyses need to be considered as purely explorative. For this reason, we have not evaluated the predictability of the models, which the statistical tests at this stage suggest is low.

We will continue to periodically report updates to these results in the inTUS resource, and the database and results can be re-analyzed at any time using the R markdown script in the freely available software R Studio. We hope that the inTUS resource will encourage more systematic exploration and reporting of the TUS parameter space, complemented with computational modeling to fill gaps in the empirical research. Other caveats worth mentioning are that the TUS community is relatively small, and many researchers work with several TUS research groups around the world. It is important to be able to identify independent replication studies. In the online version of the table, which can expand as needed, we have included additional columns that could be used for future analyses, including a column to identify independent replication or contradiction studies. Another important facet is that the current database has focused on healthy participants. Patient studies and those in animal models (e.g., rodents and nonhuman primates) would be important to include in future iterations of the database, and we welcome input on growing the database via the inTUS Qualtrics survey. Moreover, in the online data table we have added a column to identify cortical versus subcortical sites that were stimulated. Currently most TUS studies in the database have focused on cortical stimulation. These additional columns could be used for further analysis and hypothesis testing as the database grows.

To help to address the noted limitations to the current database, we establish the inTUS Resource as an open database and a set of analysis tools. The inTUS Resource can be found in the Open Science Framework link here: https://osf.io/arqp8/, in the folder associated with this paper (Caffaratti_et_al_inTUS_Resource folder). This repository contains the data tables at each of the key evaluation periods (currently 6-2024 and 1-2025) and the R markdown script to regenerate the statistical tests, results and figures using the freely available R or R Studio software (Posit). The resource also has a link to a Qualtrics form that researchers can use to submit parameters and outcomes as part of their preprint or published work. A more detailed description of the inTUS resource contents, how they are curated and how to contribute to the resource can be found as a READ_ME file in the resource documents page, a copy of which we have appended at the end of this report. Please cite this paper if you use or rely on the inTUS Resource.

As noted here, not all studies are reporting the range of ITRUSST recommended TUS stimulation parameters. We encourage TUS researchers to contribute more accurate values to the database from their studies and to more systematically report the more complete set of values as recommended by the ITRUSST consortium (Martin et al., 2024). This will allow the data to be mined more completely, which may further support or refute these or other hypotheses. We will retain the historic datasets on the resource pages as a point of reference. We welcome input via the online form on improving the criteria for assessing neurobiological or behavioral effects, which will benefit the entire TUS community and is a key objective of the ITRUSST consortium https://itrusst.com/. The original OSF link noted here will point to the location of the repository even if it moves location.

Given the sample size limitations, these retrospective meta-analysis results are tentative, with the possibility that the results may stabilize, strengthen or change. The combination of the meta-analysis and resource are made openly available as an inTUS Resource to further support and encourage the TUS research community to more systematically report TUS parameters and study outcomes. Furthermore, we encourage the TUS research community to explore the full TUS parameter space whenever safe and possible to conduct in human and nonhuman animals or rely on computational modeling for exploration of the extreme reaches of the TUS parameter space. We fully anticipate that this effort will dovetail with a need for further computational modeling and empirical study in human and nonhuman animals. TUS effects could also in the future benefit from being modeled across the cortical depth, in interaction with other brain areas (Thorpe et al., 2024) or insights on the impact on cell and molecular properties of the neural network.

For moderate intensity TUS application, we do not generate directionality of TUS effects hypotheses. This section is only intended to provide a brief introduction to moderate intensity TUS applications, mainly to distinguish them from low-intensity TUS neuromodulation.

For high intensity TUS application, we do not generate directionality of TUS effects hypotheses. This section is only intended to provide a brief introduction to high intensity TUS applications and to distinguish these from low-intensity TUS.

The inTUS Resource consists of the following resource items, linked to Caffaratti et al. Neuromodulation with Ultrasound: Hypotheses on the Directionality of Effects and a Community Resource. Please cite the latest eLife paper when using and referring to the inTUS resource: https://elifesciences.org/reviewed-preprints/100827.

The inTUS Resource documents can be found on the Laboratory of Comparative Neuropsychology data share on the Open Science Framework: https://osf.io/arqp8/ under the folder Cafferatti_et_al_inTUS_Resource. If the location of the repository changes, a link to its location will be left here.