Overall Design and Hypotheses
The purpose of the present study is to inform on the feasibility of employing focused modulation (LIFU) of neural circuits to delineate the mechanisms underlying RNT, in a pre-phase 1, double-blind crossover design. The ultimate goal is to inform future trials addressed at 1) establishing a causal relationship between large-scale brain circuits and RNT, and 2) probing such circuits as a novel neuromodulation target with the goal of decreasing RNT in the clinical setting. The study therefore aims to estimate effect sizes for LIFU of brain circuits, as well as to verify the innocuity of this technique in MDD. Specifically, we will compare inter-group differences in (1) resting-state functional connectivity, (2) probabilistic tractography, (3) the relationship between (1) and (2), and (4) overlap of tracts traversing three historical targets of psychosurgery for obsessive-compulsive disorder (OCD) and MDD, in a propensity-matched sample of individuals with depression and high RNT (H-RNT) or low RNT (L-RNT), and healthy controls (HC). These analyses will be used to identify brain regions that will be modulated in a feasibility, proof-of-principle noninvasive LIFU study. The impact of focal LIFU to the right hemisphere will be assessed in terms of 1) resting-state connectivity changes along anatomically plausible white matter tract connections, 2) patterns of change of brain activation during a paradigm of affective self-referential adjective processing, and 3) depressive symptom changes (including RNT and affective status). In addition, we will study pre- vs. post-sonication or sham sonication changes in diffusion-weighted (DWI), FLAIR, and T2. DWI is particularly sensitive to acute neuronal damage, whereas FLAIR and T2 will help document any preexisting parenchymal, glial, or vascular lesion in the brain.
An exploratory hypothesis is that stimulation of anteromedial structures of the telencephalic cortex of the right hemisphere and their connections with subcortical structures will lead to detectable attenuation of RNT in high RNT depressed individuals. We plan to target right-hemisphere structures, but given ellipsoid shape of the low-intensity ultrasound focus, approximately 23 mm in its axial diameter for a 500 kHz and 250 kHz transducer, we cannot rule out some degree of contralateral effects. In addition, interhemispheric connections may result in modulatory changes in left hemisphere functional changes as well. In this study, we will use a double-blind, randomized cross-over design employing either LIFU or sham stimulation. The specific aims for this proposal are:
Aim 1: To establish the efficacy of interference with signaling in right-hemisphere anteromedial structures in decreasing RNT and to estimate the effect size of this intervention (Schoene-Bake et al 2010).
Hypothesis 1a: Relative to sham stimulation, LIFU stimulation will result in a decrease in the intensity of RNT and subjective distress associated to it.
Hypothesis 1b: The effect size of LIFU relative to sham stimulation on RNT greater in H-RNT participants.
Aim 2: To establish the effect of interference with neural activity of right-hemisphere anteromedial structures on changes in the appraisal of adjectives that define the participants’ selves.
Hypothesis 2a: Effect of LIFU relative to sham in the endorsement of self-defining negative adjectives will be associated with the baseline RNT intensity.
Hypothesis 2b: Effect of LIFU relative to sham will result in more frequent endorsement of self-defining positive adjectives in all groups, with an effect size associated to baseline RNT intensity.
Aim 3: To define structural and functional mediators/predictors of RNT and affective response to LIFU of anteromedial structures.
Hypothesis 3a: Anteromedial sonication-mediated decrease in functional/structural connectivity coupling between rostral anterior cingulate cortex/medial orbital frontal cortex (rostralACC/mOFC cortices) and anterior thalamus/ventral striatum will be related to a decrease in RNT intensity.
Hypothesis 3b: Anteromedial sonication-mediated decrease in functional/structural connectivity coupling between rostralACC/mOFC cortices and anterior thalamus/ventral striatum will be related to less negative self-referential processing and more positive self-referential processing in all groups, with maximum effect size in H-RNT patients.
Hypothesis 3c: Increase of high-frequency heart rate variability and decrease of skin conductance level upon sonication will both correlate with decrease in post-treatment RNT and negative self-appraisal adjectives.
The study will consist of two groups/arms subjected consecutively to either LIFU sonication or sham sonication in a crossover fashion. It will be a single-site, pilot study involving 20 participants with depression (10 with low RNT and 10 with high RNT) and 10 healthy controls. There is no planed interim analysis or subgroup analyses. All data analysis will be performed after the recruitment target is reached.
Justification for Dose
The proposed stimulation parameters result in estimated tissue values for spatial-peak pulse-average intensity (ISPPA) = 2.26 W/cm2 and spatial-peak temporal-average intensity (ISPTA) = 0.22 W/cm2 (Zeng et al 2022), below levels reported to be safe in previously published literature on human subjects, including clinical (ISPPA = 11.5-17 W/cm2, Legon et al 2020), and histological (ISPTA ≤ 5.76 W/cm2, Stern et al. 2021) safety examinations. These intensities are estimated derated values in the brain based on simulations. When assuming a linear free field, we calculate a pressure of 259kPa when the ISSPA is set to 2.26 W/cm2. Moreover, this intensity of stimulation is lower than the limits recommended by the FDA for diagnostic ultrasound, namely ISPTA ≤ 0.72 W/cm2 and ISPPA ≤ 190 W/cm2, respectively); http://www.fda.gov/downloads/MedicalDevices/DeviceRegulationandGuidance/GuidanceDocuments/UCM070911.pdf, as cited by Legon et al 2020, and Marketing clearance of diagnostic ultrasound systems and transducers. Guidance for industry and Food and Drug Administration staff. Silver Spring, MD: US Food and Drug Administration, 2019 as cited by Zeng et al 2022; page 14).
We anticipate that administration of LIFU will be through the temporal bone window, given the lesser thickness and increased proximity of this bone to the regions we anticipate we will seek to modulate. A single sonication or a single sham sonication will be employed in each arm of this cross-over study.
As stated, we will employ the parameters used by Zeng K et al. (2022): An 80-second train of 20-millisecond bursts of ultrasound (0.5 MHz), repeated every 200 milliseconds (400 bursts; Figure 2, top panel). Acoustic simulations will be performed with the k-Wave Matlab Toolbox (Treeby and Cox, 2010) to individually confirm the estimated total energy delivered during sonication and verify tissue temperature increases are <1°C. We estimate a 75% tissue attenuation of energy when the ultrasound wave reaches its target, therefore we will set the free-field ISPPA at 9.04 W/cm2 or 518 kPa (to achieve 2.26 W/cm2 derated ISPPA; Figure 2, bottom panel). Since the free-field ISSPA maximum limit of the NeuroFUSTM system is fixed at 30 W/cm2 (free field-PNP at 30 W/cm2 = 944 kPa), i.e. higher than the intended stimulus, in order to minimize the chances of errors during the administration of the stimulus, all subjects will undergo sonication under identical conditions, as shown in Figure 2. These conditions will be fixed with the “Save” option in the operator interphase. In addition, to prevent accidental changes in the parameters of stimulation, the operator and a research assistant will independently verify the parameters as stated herein and shown in Figure 2, prior to each stimulus administration.
Study Population
Inclusion Criteria
To be eligible to participate in this study, an individual must meet all criteria at enrollment:
- Prior participation in the Laureate Institute for Brain Research (LIBR) the Centers of Biomedical Research Excellence (CoBRE) study (WIRB Protocol #20182352) and selected as a component of the propensity-matched sample of patients with major depression and varying intensities of RNT (in the form of Ruminative Response Scale: RRS score; n=20) (Figure 3), or healthy individuals with no psychiatric diagnosis (n=10),
- A Patient Health Questionnaire-9: PHQ-9 score ≥ 10 at enrollment.
- Provision of a signed and dated informed consent form.
- Participant provides verifiable contact information (name, telephone number(s), email and mailing address) for at least 2 persons who agree to be contacted by study personnel as deemed necessary.
- Participant is followed by a licensed physician or a licensed mental health care provider (i.e., psychologist, LCSW) outside of LIBR throughout study participation.
- Stated willingness to comply with all study procedures and availability for the duration of the study.
- Male or female, aged 18 to 60 years.
- In good current general health as evidenced by medical history.
- Ability to comply with the study visit timeline.
- For females of reproductive potential: negative urine pregnancy screening test.
Exclusion Criteria
In the opinion of the Principal Investigator, an individual who meets any of the following criteria at enrollment will be excluded from participation in this study:
- Current use (within the last 30 days) of drugs of abuse or moderate / severe alcohol use disorder.
- Lifetime diagnosis of schizophrenia spectrum disorder, other nonaffective psychotic disorders, or bipolar disorders.
- Presence of cardiac pacemaker or any other MRI contraindication.
- Pregnancy or lactation.
- Febrile illness within the last two weeks.
- Treatment with an investigational drug or participation in any other interventional research protocol in the last 2 weeks.
- The participant is unable to understand the goal of the study, instructions, or the risks associated with the study as judged by a clinically trained assessment team member.
- Clinical and/or imaging evidence of vascular, traumatic, or neurodegenerative disorders of the central nervous system (CNS), or other neurological disorders potentially compromising patient’s participation in the study, or study results. This includes, but it is not limited to, any minor or major neurocognitive disorder including those caused by traumatic brain injury, Parkinson’s disease, significant small-vessel disease, multiple sclerosis, Huntington’s disease, early-onset Alzheimer’s disease, chronic infections of the CNS or the meninges, previous chronic use of alcohol or cerebrovascular accident (CVA) sequelae, or a Montreal Cognitive Assessment (MoCA) score <25 due to any cause if cognitive compromise is clinically suspected. The PI can decide if a potential participant needs to be excluded due to some other cause of structural or functional compromise of the CNS.
- Active suicidal ideation (as measured by Suicide-Risk-Assessment, Columbia Suicide Severity Rating Scale : C-SSRS, “Yes” answers to items 3, 4 or 5 of Suicidal Ideation-Past 1 month section, or any “Yes” answer to any of the items of Suicidal Behavior-Past 3 months section), or any suicide attempt in the last year.
Study Procedures and Assessments
Definition of RNT-relevant structural/functional connectivity coupling:
Whole-brain study: Functional Connectivity (del Cerro et al 2020): All functional connectivity analyses will be performed with the CONN toolbox version 17.f. Significance threshold will be set at p < 0.05, Family-Wise Error (FWE) corrected within our masks of interest. Specifically, statistical significance will be estimated by means of a voxel-wise non-parametric permutation testing with 5000 permutations, using the Threshold-Free Cluster Enhancement (TFCE) method as implemented in the SPM-TFCE toolbox v156 (http://dbm.neuro.uni-jena.de/tfce/).
Voxel-to-voxel analysis: A global degree network measure approach will be used. Network measures of degree summarize the number of connections of any given voxel with the rest of gray matter voxels. Specifically, here we will use the intrinsic connectivity contrast (ICC), which does not require any a priori information. Voxel-to-voxel bivariate correlation coefficients will be computed from the residual blood oxygenation level dependent (BOLD) time series, obtaining a whole-brain gray matter map where voxel intensities accounted for the degree of correlation of each voxel with the rest of voxels. Such maps will be normalized to fit a Gaussian distribution with zero mean and unitary variance. Realignment parameters, scan-to-scan changes in global BOLD signal (Z-scores), framewise displacement time-series (in mm), and outlier scans will be included as nuisance parameters in this first-level analysis. Next, they will be included in a second-level (between-group) analysis, and an analysis of variance (ANOVA) will be performed to derive F-statistic maps comparing global connectivity patterns between HC, L-RNT, and H-RNT. As per our hypotheses, second level analyses will be restricted to voxel masks encompassing precunei, anterior and posterior cingulate cortices, subgenual cingulate, and medial orbitofrontal cortex. Whereas these analyses apply to the whole brain, the stimulus will be delivered to the right hemisphere only.
Whole-Brain Study: Gray- and White-Matter Based Structural Connectivity Analysis (Sanchez et al 2020): Preprocessed diffusion MRI data of every subject will be used to obtain fiber orientations in each voxel. The contribution of different fiber orientations per voxel is represented by a set of spherical harmonics (Tournier et al 2007) obtained through the Constrained Spherical Deconvolution (CSD) method. This model estimates the fiber orientation distribution (FOD). In the next step, each of these local white matter orientations will be connected with the next one thus delineating the fiber pathways, applying an iFOD2 algorithm in a whole-brain tractography approach and generating several million streamlines with seed dynamic option to improve the distribution of reconstructed streamlines density. These streamlines will be filtered using the spherical deconvolution informed filtering of tractograms (SIFT; Smith et al 2013). Finally, implementing a grey matter parcellation, a structural connectome per subject will be obtained. All preprocessing steps will be performed using FSL and MRtrix packages software.
Whole-Brain Study: Fixel-Based (FB), Fractional Anisotropy (FA) and Mean Diffusivity (MD) Analyses (Sanchez et al 2020): A fixel refers to a fiber bundle within a voxel and, each voxel can have more than one fixel. In this way, the FB analysis (FBA) reveals information in a smaller scale than classical DTI models. We will explore three metrics derived from FBA: i) fiber density (FD), the intra-axonal volume of a fiber bundle; ii) fiber-bundle cross-section, number of voxels the fiber bundle occupies by its intra-axonal volume; and iii) fiber density and cross-section that combines both measurements. To achieve this, a FOD template will be designed performed for the HC, L-RNT, and H-RNT. In the next step, all fixels will be extracted from each voxel of the FOD template and the three metrics calculated. Then, we will compare the metrics from the three groups through a General Linear Model and by applying nonparametric permutation testing with 5000 permutations, where family-wise error corrected p-values were assigned to each fixel. Finally, with visualization purposes, we will execute on the FOD template a probabilistic tractography algorithm in order to generate a tractogram template that represents the three groups. We will select right hemisphere streamlines from this tractogram template, that are associated to significant fixels (corrected p < 0.05), and a “significant tract” for the anatomical structures enumerated in Aim 3 (Section 4.1) will be obtained, both at group level, and in individual participants. FA and MD will be also estimated in anatomical tracts defined with FBA, and these metrics will be compared in all three experimental groups.
Structural Correlation Network Analysis: Structural correlation networks (SCNs) have been shown to reflect synchronized maturational changes in connected brain regions and have been considered as putative biomarker for altered connectivity in individuals with neurodevelopmental disorders (Saggar et al 2015). In other words, brain regions that ‘grow together’, i.e., increase or decrease in volume at the same rate over the course of years in the same individual, show structural covariance or anatomical connectivity across individuals and reflect synchronized developmental change in distributed cortical regions (Alexander-Bloch et al 2013).
Individual Functional-Structural Coupling (Honey et al 2009, Jiang et al 2020, Sanchez 2021): We will use the functional and structural connectivity information from Hypothesis 3 to derive a metric permitting to describe their coupling. We will calculate one metric per subject, and it will be compared against coupling metrics from a random sample obtained through a null model. This coupling metric will be defined as the Spearman’s correlation between the two contributions (functional and structural) of the whole brain and, on the other hand only selecting the aforementioned regions of interest. Finally, we will compare all intra-subject outcomes and assess the relevance (or impact) of these regions. We will explore other correlation methods apart from linear correlation (Mohanty et al 2020). Last, we will include MDD duration and response status as covariates in these analyses.
Tractographic-functional coupling analysis of neurosurgical targets for treatment of resistant depression (Schoene-Bake et al 2010, Sanchez 2021): We will refine the findings above by comparing them with the overlap of white matter tracts (and the regional functional connectivity associated to them) traversing the recreated seeds that correspond to the procedures of anterior capsulotomy, anterior cingulotomy, and subcaudate/subgenual tractotomy, as guided by the following Table 1:
Table 1: Location of three simulated historical psychosurgical targets for depression
Target
|
MNI Coordinates Right Hemisphere
|
Anterior capsulotomy
|
x=21 (19–23); y=18 (16–20); z=1 (1–3)
|
Anterior cingulotomy
|
x=8 (6–10); y=30 (28–32); z=16 (14–18)
|
Subcaudate tractotomy
|
x=15 (13–17); y=20 (18–22); z=9 (7–11)
|
We will thus employ a similar method as Schoene-Bake et al (2010), in the propensity-matched sample of patients with depression and L-RNT or H-RNT, in addition to comparable healthy individuals. The site of stimulation will be determined by the most overlap by voxel count of white matter tracts traversing these simulated targets and involving anterior thalamic radiations that reach both orbitofrontal and anterior cingulate cortices.
LIFU experiment (double-blind, sham-controlled design for the study of connectivity changes, clinical changes, and self-evaluation changes):
Study Design and Baseline Neuroimaging Characterization:
A total of 20 MDD participants will be included in this pilot feasibility study. MDD individuals will be extracted from the H-RNT (n=10) and L-RNT (n=10) propensity-matched groups defined above. In addition, we plan to recruit 10 matched healthy comparison individuals. After written informed consent, each participant will be randomized to either a LIFU first or sham first group. All participants will receive both sham and LIFU in a randomized, cross-over, double-blind design (Figure 1). The sham will be implemented by interposing a sound-absorbing material between the active transducer and the scalp. Thus, participant and team members performing efficacy and safety measurements will be blind to the stimulus condition, but the operator delivering the stimulus will not. Prior to randomization, all subjects will undergo the following imaging studies:
• T1 to be employed in the neuronavigation-assisted LIFU transducer positioning (see below).
• Baseline resting-state fMRI, to be compared with post-sonication resting-state fMRI and establish functional connectivity changes mapped onto structural connections of interest.
• Multishell DTI with probabilistic tractography will be used to define streamlines in each subject that conform to targets as defined above. Such streamlines will be mapped onto the T1 image-generated 3D coordinates, to position the LIFU focus by employing neuronavigation software (see below).
Clinical Variables: Core Variables
Brief State Rumination Inventory (BSRI, Marchetti et al 2018): The BSRI is a validated scale to assess the momentary rumination status of participants. In contrast with other scales, which usually emphasize trait-like statements, the BSRI assesses current status of ruminations with 8 questions beginning with “Right now (…).”
Positive and Negative Affect Schedule-State (PANAS-State, Watson and Clark, 1994): PANAS-State has been successfully used to evaluate acute changes in affective status during the development of new neuromodulation techniques and targets (Mayberg et al 2005). Apart from anticipating a decrease in negative affective item scores in case of successful modulation of the proposed targets, we will carefully monitor for increases in positive affect items, given anecdotal reports of hypomania with successful neuromodulation of limbic nodes.
Clinical Variables: Additional Variables
Columbia-Suicide Severity Rating Scale (C-SSRS, Posner et al. 2011): Suicidal ideation (SI) and behavior is an important complication of internalizing disorders, only partially related to general severity. We consider it important to carefully evaluate suicidal ideation and purpose both before and after each sonication session, for clinical and safety purposes. Any interaction that involves mental health assessments or intervention has the possibility to exacerbate mental health symptoms, including suicidal ideation. To prospectively assess the impact of both assessment and intervention we will monitor SI using the C-SSRS at the baseline visit, pre- and post-intervention in Visit 1 and Visit 3, in Visit 2, and at the follow-up, i.e., final visit.
Based on the results from the C-SSRS we will determine the clinical significance of possible SI with a follow-up clinical interview that will be conducted by a board-certified psychiatrist. If emerging SI is deemed clinically significant the psychiatrist will determine whether acute clinical care is indicated. This will be discussed with the study participant, and appropriate clinical arrangements will be made in collaboration with the study participant. All incidents will be documented and will be submitted by the PI to, and reviewed by, the data safety monitoring board (DSMB).
Montgomery-Asberg Depression Rating scale (MADRS, O’Reardon, et at., 2007): The MADRS will be used to quantify general symptoms of depression before the baseline imaging study and after the post-sonication imaging study. It will be obtained by a team member blind to the nature of the sonication experiment (i.e., active or sham). A structured MADRS interview will be employed to improve reliability.
Ruminative Responses Scale (RRS, Nolen-Hoeksema and Morrow, 1991): The RRS will be used as a trait rumination measure that will permit splitting the sample of MDD participants into those with high or low RNT.
fMRI self-evaluation task
Task: Each participant will perform a self-referential processing task (Matthews et al 2010) during two separate fMRI sessions (Figure 1). RNT being a prominent form of self-referential mentation, we expect this task to be sensitive to manipulations of circuits subserving RNT. It is by now well established that so-called cortical midline structures
(Northoff et al 2006) are essential for representation, monitoring, evaluation and integration of self-referential stimuli across the lifespan (Jankowski et al 2014). A recent meta-analysis has identified the medial prefrontal cortex as one of these structures, which serves a variety of social, affective, and cognitive functions. In particular, the anteromedial prefrontal cortex has been linked to self-processes (Lieberman et al 2019). It appears that it is the perigenual anterior cingulate cortex that is specifically involved in self-related processing when compared to familiarity, other, and task/stimulus effects and that this region overlaps with default mode network activity during resting-state. In comparison, other midline regions, i.e., medial prefrontal cortex and posterior cingulate cortex were also recruited during the processing of both self-specific and familiar stimuli (Quin and Northoff 2011). Disturbances of self-related processing is important for every psychiatric disorder. For example, individuals with depression suffer from an increased self-focus, attribution of negative emotions to the self (Wang et al 2022), higher degrees of self-relatedness of negative emotional stimuli (Grimm et al 2009), and increased cognitive processing of the own self (Northoff 2007). Some have reported that depressed individuals show a reduced activation in medial prefrontal cortex during the processing of negative, depression-related words (Renner et al 2015) and judgement of self-relatedness (Grimm et al 2011), but exaggerated striatal processing (Quevedo et al 2022). In some studies, these activation differences correlated with depression severity and hopelessness (Grimm et al 2009). Others have reported increased brain activity in the dorsal medial prefrontal cortex when depressed participants were the receiver of self-related negative events (Wang et al 2022). Interestingly, greater activity in the medial prefrontal cortex during self-referential processing is associated with reduced RNT in depressed individuals (Nejad et al 2019). These accounts have been consolidated by proposing that there is evidence for an elevated tonic ventral medial prefrontal activation, which may relate to automatic aspects of depressive self-focus but also an elevated phasic dorsal medial prefrontal activation, which may be associated with more strategic aspects of depressive self-focus, such as comparing the self with inner standards (Lemogne et al 2012). Taken together, these findings support the notion that assessing the self-relevance of adjectives as a common paradigm to measure self-related processing probes several component processes. First, an assessment of valence, second an assessment of self-relevance, which might comprise interoceptive afferents associated with the word, the perception of the external environment as it relates to the word, and episodic autobiographical memories that support the truth of the self-related proposition, and the decision-making process to select an available option. Moreover, detailed analysis of valence-specific responses and contrasting conditions such as others and valence judgments can aid in delineating these processes. Previous studies clearly show that psychological and behavioral measures of self-related processing correlate with levels of activity or degrees of connectivity of brain structures involved in self-related processing. The task itself will have three different types of trials, based upon the Affective Norms for English Words (ANEW) word list (Bradley and Lang 1999): (1) self-evaluation, (2) other-person evaluation, and (3) word evaluation. During each trial, the subject will be shown initially the type of trial for 1.5 s (i.e., “I am…” for (1), “My best friend is…” for (2), and “This word is positive/negative” for (3)). In the next screen the subject will be shown the adjective drawn from the ANEW word list, along with the Yes and No options on the bottom left and right sides of the screen respectively, with the instruction of pressing the left button for “Yes” and the right button for “No.” This screen will be presented for another 1.5 s, or less if a response is provided sooner.
The adjectives for self, other-person, and word evaluation trials will be matched for arousal, valence, and dominance. Once the subject’s response is recorded (marked by the appearance of an “X” below the chosen response), a blank screen will be presented for the remaining duration of the trial. Since the intertrial interval, showing a fixation cross will be of 3 s, each trial cycle will be 6 s long. We will present 16 6-s trials for each of the 6 possible trial types that result from 2 valence possibilities (i.e., positive or negative) x 3 evaluation conditions (i.e., (1), (2), and (3) above). This yields a total of 96 trials, which will be presented in pseudo-randomized order. The first trial of each run will be preceded by a 0.5 s blank screen, and the last trial will be followed by a 3.5 s blank screen. Therefore, total scan time during the task will be 580 s (i.e., 6 x 96 s + 0.5 s + 3.5 s). Responses for all trials will be recorded for behavioral analyses. In this regard, four paired t tests will be performed to compare responses before and after sonication or sham, related to a. positive self, b. negative self, c. positive other, and d. negative other. We anticipate unambiguous identification of word valences, so this will not be used as an outcome to assess the impact of sonication vs. sham.
fMRI data collection and analysis:
Two resting-state fMRI runs will be collected for each subject using a 3T GE scanner. The parameters will be: T2* weighted echo planar imaging, TR=2000 ms, TE=27 ms, flip angle 78°, resolution: 2.5x2.5x2.9 mm. Each resting-state scan will take 6 minutes and the subjects will ask to rest with eyes open. During the task, fMRI acquisitions will be time-locked to the onset of the task, and the acquisition parameters will be the same as resting-state fMRI runs. During the same experimental session, a high resolution T1-weighted image (TR = 7 ms, TE = 2.9 ms, flip angle =8°, ~1 mm3 voxels) will be obtained for anatomical reference.
All structural and functional image processing will be performed with the AFNI software package (Cox 1996). Preprocessed time series data for each individual will be analyzed using a multiple regression model. Regressors will be constructed for positive self, negative self, positive other, negative other, positive word, and negative word. Additionally, five nuisance regressors will be used to account for residual motion (roll, pitch, and yaw) and to eliminate slow signal drifts (baseline and linear trend). These 11 regressors will be applied to the AFNI program 3dDeconvolve in order to calculate the estimated voxelwise response amplitude. To account for individual variation of anatomical landmarks, a Gaussian filter with 4 mm full width at half maximum will be applied to the voxelwise percent signal change data. Data for each subject will be normalized to Talairach-Tornoux coordinates.
To test a priori hypotheses regarding the effect of anteromedial LIFU on regional activation changes related to self-processing, a two-way ANOVA will be performed with task condition (self-evaluation/word evaluation) and stimulation condition (sham/LIFU) as fixed within-subject factors, subject number as a random factor, sequence (i.e., sham first-LIFU second/LIFU first-sham second), and diagnosis (i.e., H-RNT/L-RNT) as between-subject factors.
For the task effect a whole-brain analysis with p<0.01 will be performed. For the sonication effect, brain activation related to self and word evaluation trials after LIFU relative to brain activation related to self and word evaluation trials after sham exposure. For the LIFU and LIFU x task interaction, a whole-brain analysis will be performed with a p<0.05. Simulations using the AFNI function AlphaSim will be performed to account for multiple comparisons. Clusters of adjacent voxels that maintained the thresholds for the task, sonication, and task x sonication interaction effects will be retained.
In an exploratory analysis we will examine whether brain activation differs between the self- and other-person evaluation conditions by performing a voxel-based two-way ANOVA with task condition (self-evaluation/other-evaluation) and sonication condition (sham/LIFU) as fixed within-subject factors, subject number as a random factor, and sequence (sham first-LIFU second/LIFU first-sham second) and diagnosis (H-RNT/L-RNT) as between-subjects factors.
Positioning and Operation of the LIFU device
We will employ a Sonic Concepts Inc. NeuroFUS™ PRO Version 1.0/transducer power output (TPO)-203 system (Figure 4) to apply low-intensity transcranial ultrasound stimulation to modulate neural activity in the region of interest. Figure 4 shows the pulse generator/interface, a schematic depiction of the transducer, and the technical characteristics of both available transducers (CTX-250 and CTX-500). The choice of the transducer will depend on the size of the target to sonicate; in turn, this will result from the image analyses described above. The transducer will be positioned with the help of a Brainsight® neuronavigation system, which achieves <2 mm precision in the localization of the ultrasound (US) focus relative to the defined target.
For the active sonication, we will use parameters within a range previously described to be appropriate both to effect behavioral and fMRI-BOLD signal changes (Verhagen et al 2019, Folloni et al 2019, Fouragnan et al 2019, Zeng et al 2022) and safe in clinical (Zeng et al 2022), imaging and histological examination (Verhagen et al 2019, Yoon et al 2019) in preclinical studies.
Active sonication will thus involve the following parameters, used by Zeng K et al. (2022): An 80-second train of 20-millisecond bursts of ultrasound (0.5 MHz), repeated every 200 milliseconds (400 bursts; Figure 2, top panel). Acoustic simulations will be performed with the k-Wave Matlab Toolbox (Treeby and Cox, 2010) to individually adjust the estimated total energy delivered during sonication and verify tissue temperature increases are <1°C, modifying Power/Channel, in all subjects.
Sham sonication will use the same parameters, but a high acoustic impedance disk will be placed on the face of the transducer (Legon et al 2018), to account for nonspecific indicators of active sonication, such as device noise or subtle vibration.
BrainSightTM neuronavigation software will be used to 3D coordinate definition of sonication targets. The transducer will be manually positioned by the Principal Investigator over the skull (temporal bone) of the right side. US focus will be steered to the tissue depth defined with the neuronavigation software, and appropriate anatomical positioning will be verified with a T1 image.
A nervous system-relevant review of symptoms, and a sensory and motor neurological physical exam will be performed by an certified physician or nurse before and after each sonication (or sham) session (Legon et al 2020). The pre- and post-sonication MRI sessions will include, in addition, the collection of DWI, FLAIR, and T2-weighted images to confirm the innocuity of this intervention, an approach comparable with previous LIFU studies (Darmani et al 2022). Total sonication exposure will be up to 80 seconds, i.e., well below its duration of use (i.e., <60 minutes per Annex VIII of the MDR).
The proposed pattern of stimulation results in values well below recently published tissue energy in human studies (Figure 4, from Legon et al. 2020; Stern et al 2021).
Moreover, ISPTA and ISPPA resulting from the proposed stimulation pattern (0.22 W/cm2 and 2.26 W/cm2 respectively) is below the recommended limit for diagnostic ultrasound (i.e. 0.72 W/cm2 and 190 W/cm2, respectively); http://www.fda.gov/downloads/MedicalDevices/DeviceRegulationandGuidance/GuidanceDocuments/UCM070911.pdf, as cited by Legon et al 2020, and Marketing clearance of diagnostic ultrasound systems and transducers. Guidance for industry and Food and Drug Administration staff. Silver Spring, MD: US Food and Drug Administration, 2019 as cited by Zeng et al 2022; page 14 of the document).
Autonomic nervous system function variables and correlation with BOLD signal changes (Villarreal et al 2021):
Electrocardiogram (ECG) and/or cardiac cycle assessed through oximeter pulse wave, respiratory rate (RS) and skin conductance level (SCL) will be measured simultaneously and continuously 6 min before, during, and 6 min after each session of low intensity fUS administration, and then during resting state fMRI sessions only, using an MRI compatible device (MP150, BIOPAC Systems, Inc., USA). Continuous noninvasive blood pressure will be recorded outside of the scanner only. ECG will be measured by three disposable pre-gelled EL258RT electrodes located in the left area of the chest (positive pole in V5, negative pole in V2, and the neutral pole in between). Electrodes will be connected to the ECG100C amplifier. RS signal will be registered using a Pressure Respiration Transducer (TSD110) and a self-inflating pressure sensor RX100 connected to tubing with no electrical requirement, all interface to the MP150 via a DA100C amplifier. The RX100 will be attached to the chest of each subject where we register the maximum breadth of inspiration. SCL will be measured using two electrodes that record changes in conductance through the skin. The EL509 Ag-AgCl disposable electrodes with isotonic gel (GEL100 BIOPAC) will be attached to the index and middle fingers of the left hand. Both snap electrodes will be connected to an EDA100C amplifier, which operates by applying a fixed 0.5 Volts DC across the two electrodes and then measuring the current flowing between the two electrodes.
The signals will be recorded at a sample rate of 1 kHz and processed with the BIOPAC software AcqKnowledge and MATLAB (The MathWorks, Inc., Natick, MA).
To assess how functional connectivity changes are related to autonomic activity status, we will use a one-way-ANOVA test implemented in SPM12 between groups for each condition, using as covariates of interest: 1) the respiratory sinus arrhythmia for vagal regulation and 2) area under the curve of the skin conductance level changes for the sympathetic inspection (Figure 5 from Villarreal et al 2021). In all cases we will use the signal after sonication with respect to the signal prior to sonication.
Safety assessments
Baseline clinical assessment will include a neurological physical examination, and first baseline scanning will include scans with sequences FLAIR, DWI, and T2 to establish the presence of recent or old parenchymal, glial, or vascular abnormalities before any study-related intervention. A neurological examination and a repeat FLAIR, DWI, and T2 scans will follow each study intervention (either active sonication or sham sonication), and images will be assessed for any changes compared to baseline, before the next intervention (sham if active sonication was administered initially, or vice versa).
Table 2 summarizes the objectives, both hypothesis-testing and safety testing, and endpoints of the present study. Figure 6 depicts the timeline for the verum and sham stimulus application, along with efficacy and safety assessment associated with it.
Table 2 summarizes the objectives, both hypothesis-testing and safety testing, and endpoints of the present study. Figure 6 depicts the timeline for the verum and sham stimulus application, along with efficacy and safety assessment associated with it.
Table 2. Objectives and Endpoints
OBJECTIVES
|
ENDPOINTS
|
JUSTIFICATION FOR ENDPOINTS
|
Primary
|
|
|
To estimate the effect size of LIFU of anteromedial structures in the right hemisphere on the severity of RNT as measured by BSRI.
|
• Change of BSRI score pre- vs post-sonication or sham intervention.
|
BSRI is a standard state rumination measurement of RNT in patients with major depression, and our center has extensive experience administering and analyzing it.
|
Secondary
|
|
|
To establish the effect of LIFU of anteromedial structures in the right hemisphere on:
• Structural integrity of sonicated areas, including neuronal, glial, and vascular.
• Affective load of self-referential adjectives
• Functional connectivity between anterior thalamus, medial orbitofrontal (mOFC), and subgenual and anterior cingulate cortices (sgCC and aCC).
• General symptoms of depression
• Current affective status
|
• Changes in psychiatric and neurological examination, as well as diffusion-weighted (DWI), FLAIR and T2 MRI sequences pre-vs post-sonication or sham intervention.
• Change of distress associated to repetitive self-referential thoughts.
• Change of the affective valence of self-referential adjectives pre- vs post-sonication or sham intervention.
• Change in the module of functional connectivity between anterior thalamus, mOFC, sgCC and aCC pre- vs post-sonication or sham intervention.
• Change in MADRS pre- vs post-sonication or sham intervention.
• Change in PANAS-State pre- vs post-sonication or sham intervention.
|
Proposed MRI sequences are sensitive and specific determinants of the existence of recent and old damage to neurons, white matter, and brain blood vessels.
Clinical and imaging variables chosen in this project are standard or well-validated methods to assess the symptom and brain circuit structure and function constructs of interest.
|