Methods for diagnosing and treating depression

Abstract

Provided herein are methods of treating depression. The methods include obtaining intracranial neural activity data from a subject, determining an aperiodic slope of power spectral density from the intracranial neural activity data, determining a depression severity score from the aperiodic slope, and treating the subject based upon the depression severity score. Also provided herein are non-invasive methods using proxies of intracranial neural activity data from a subject.

Description

[0001] METHODS FOR DIAGNOSING AND TREATING DEPRESSION

[0002] CROSS-REFERENCE TO RELATED APPLICATIONS

[0003] The present application claims priority under 35 U.S.C. § 119(e) to U.S. Provisional Patent Application No. 63 / 592,453, filed October 23, 2023, which application is incorporated herein by reference in its entirety.

[0004] STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT

[0005] This invention was made with government support under R01MH127006 awarded by the National Institutes of Health. The government has certain rights in the invention.

[0006] BACKGROUND

[0007] Major Depressive Disorder (MDD) is a debilitating disorder with a poor treatment success rate despite its annual global cost approaching $92.7 billion. Around one in three individuals experience no meaningful symptom improvement after multiple therapeutic attempts, and are diagnosed with treatment-resistant depression (TRD). The current reliance on clinician- administered scales for monitoring changes in symptoms has limitations, including impaired expression of emotional experience (alexithymia), environmental influences, and socio-cultural factors.

[0008] The emergence of electrical stimulation-based treatments like transcranial magnetic stimulation (TMS) and deep brain stimulation (DBS) has offered hope to patients with unmet clinical needs and presents an opportunity for direct brain recording, potentially facilitating the identification of a physiological biomarker in neural electrophysiology. Previous efforts to find a biomarker have primarily explored fluctuations in oscillatory power bands within specific affective or reward processing regions during affective tasks to assess the efficacy of MDD and TRD treatment. However, oscillatory markers for DBS treatment tracking or mood diagnostics exhibit substantial inconsistency in frequency range and anatomical location across patients and studies, or may only be detected after weeks to months of treatment.

[0009] Accordingly, there is a need in the art for articles and methods that improve on existing therapeutic processes by providing a reliable biomarker for depression to aid in selection of optimal therapeutic modalities, as well as enhance the tracking of symptom improvement and treatment efficacy. The present invention addresses this need.

[0010] SUMMARY

[0011] In one aspect a method of treating depression includes obtaining intracranial neural activity data from a subject, determining an aperiodic slope of power spectral density from the intracranial neural activity data, determining a depression severity score from the aperiodic slope, and treating the subject based upon the depression severity score. In some embodiments, obtaining the intracranial neural activity data includes recording the intracranial neural activity data from the subject. In some embodiments, the intracranial neural activity data includes electroencephalography (EEG) data. In some embodiments, the EEG data is recorded from target regions of the affective and reward processing networks. In some embodiments, the target regions include ventromedial prefrontal cortex (vmPFC), dorsal prefrontal cortex (dPFC), dorsal anterior cingulate cortex (dACC), amygdala (AMY), mid-temporal regions (MT), or combinations thereof. In some embodiments, the EEG data is recorded from implanted stereoelectroencephalography (sEEG) electrodes.

[0012] In some embodiments, determining the aperiodic slope includes estimating power spectral density, resampling the spectra to spaced frequency bins, performing a fit in log-log over a desired frequency range, and determining the slope of the fit. In some embodiments, the power spectral density is estimated with Welch’s method or Morlet wavelet decomposition. In some embodiments, the spaced frequency bins are linearly or logarithmically spaced frequency bins. In some embodiments, the fit is linear or polynomial. In some embodiments, the aperiodic slope is calculated in a frequency range over alpha frequencies and under gamma frequencies. In some embodiments, frequency range is between 9 and 96 Hz. In some embodiments, the aperiodic slope is calculated in the range of between 20 and 45 Hz. In some embodiments, the aperiodic slope is averaged across electrodes in a region of interest. In some embodiments, the depression severity score is lower when the aperiodic slope is flatter.

[0013] In some embodiments, treating the subject based upon the depression severity score includes administering deep brain stimulation (DBS) when the depression severity score indicates severe depression, treatment resistant depression, or a combination thereof. In some embodiments, the DBS is applied to areas exhibiting abnormalities. In some embodiments, the DBS is applied to the ventral capsule / ventral striatum (VC / VS); subcallosal cingulate (SCC), superolateral medial forebrain bundle (slMFB); lateral habenula (LHb), or a combination thereof.

[0014] In some embodiments, the method further includes determining the depression severity score from the aperiodic slope during treatment of the subject, determining changes in the depression severity score during treatment, and quantifying efficacy of the treatment.

[0015] In some embodiments, the method further includes determining the depression severity score from the aperiodic slope during treatment of the subject, determining changes in the depression severity score during treatment, and modifying parameters of the treatment according to the changes in the depression severity score.

[0016] In another aspect, a method of treating depression includes obtaining proxies of intracranial neural activity data from a subject, determining an aperiodic slope of power spectral density from the proxies of intracranial neural activity data, determining a depression severity score from the aperiodic slope, and treating the subject based upon the depression severity score. In some embodiments, the proxies of intracranial neural activity data include data from non- invasive measures. In some embodiments, the non-invasive measures include pupillometry, fMRI, scalp EEG, heart rate, skin conductance, magnetoencephalography, or a combination thereof. In some embodiments, the non-invasive measure includes pupillometry data.

[0017] In some embodiments, determining the aperiodic slope includes estimating power spectral density of pupil dilation frequency, resampling the spectra to spaced frequency bins, performing a fit in log-log over a desired frequency range, and determining the slope of the fit. In some embodiments, the power spectral density is estimated with Welch’s method or Morlet wavelet decomposition. In some embodiments, the spaced frequency bins are linearly or logarithmically spaced frequency bins. In some embodiments, the fit is linear or polynomial. In some embodiments, the desired frequency range is over alpha frequencies and under gamma frequencies.

[0018] BRIEF DESCRIPTION OF THE DRAWINGS

[0019] For a fuller understanding of the nature and desired objects of the present invention, reference is made to the following detailed description taken in conjunction with the accompanying drawing figures wherein like reference characters denote corresponding parts throughout the several views.

[0020] FIGS 1A-B depict graphs illustrating CAT -DI score and data collection timeline during the 9-day in-patient monitoring period. (A) The CAT-DI assessment times varied per day, dependent upon clinician and patient needs. Clinical mood scales were collected in the early AM and immediately followed by a CAT-DI assessment. Mealtimes occurred at approximately 9 am, 1 pm, and 5 pm, and were also immediately followed with a CAT-DI assessment. In between mealtimes, CAT-DI assessments were collected approximately every 1-2 hours between AM and PM scheduled tasks. Days 1, 2, and 9 consisted primarily of cognitive and affective tasks, collected independently of stimulation paradigms. Days 3 - 8 consisted of DBS stimulation tasks. Finally, CAT-DI was collected at the end of the day before research closeout. (B) Each row shows the temporal progression of CAT-DI scores throughout the stay for each patient. Each bolded line along the horizontal axis represents midnight for days 1-9. Severity scores for each CAT-DI assessment are plotted on the y-axis at the corresponding x-axis time during the stay.

[0021] FIGS. 2A-C depict images and graphs illustrating aperiodic slope as depression mood severity biomarker. (A) The aperiodic slope was computed for each depression severity (CAT- DI) assessment block per electrode and correlated with the depression severity scores. Electrode coloring indicates slope-depression correlation magnitude; magenta outlines designate electrodes belonging to the putative ‘biomarker’ ventromedial PFC region of interest (vmPFC ROI). (B) Power spectra averaged across the ‘biomarker’ vmPFC region plotted for the highest (most depressed) and lowest (most euthymic) CAT-DI scores for each patient, showing that as slope becomes less negative (flatter) depression symptoms seem to alleviate. (C) Depression severity vs slope (averaged over vmPFC) for all scores collected during the 9-day in-patient monitoring period.

[0022] FIG. 3 depicts graphs illustrating depression severity vs slope averages for all scores collected during the 9-day in-patient monitoring period. The first row shows comparisons of depression severity to aperiodic slope averaged over all electrodes. Subsequent rows compare to averages across each ROI subregion.

[0023] FIGS. 4A-D depict images and graphs illustrating biomarker regional and frequency specificity. (A) Electrodes were divided into five ROI groups indicated by fill color. (B) Similar electrode placements allowed inter-subject electrode registration. Electrode fill color shows the average correlational value across subjects (r = -0.5 to 0.06), and electrode size indicates the effect size of the relationship at each contact (Cohen’s d range = -6.5 to 0.3). (C) The distribution of regional depression severity: slope correlations indicates that the vmPFC exhibits the strongest relationship with depression severity score (median = -0.35, IQR = 0.18), and that the temporal contacts exhibit the weakest relationship (median r = -0.06, IQR = 0.24). Each marker represents one electrode from one subject. (D) Severity-slope correlations were calculated for every frequency range from 1 to 100 Hz and averaged in the vmPFC region. The white square designates the location of the a priori range (20-45 Hz).

[0024] FIG. 5 depicts images illustrating similarity of electrode placement across subjects. Marker size indicates the root-mean square dispersion of each electrode across patients after registration across similar trajectories. Note reference / ground electrodes (left dorsal region) were omitted from all analyses leading to focally decreased dispersion statistics for some contacts.

[0025] FIG. 6 depicts desired slope-mood tracking per patient according to an embodiment of the disclosure. Aperiodic slope was calculated for each possible frequency range (1-100 Hz) within the vmPFC for each patient and correlated with depression severity scores. The frequency range for a given correlation is identified by the intersection of the lower (y-axis) and upper (x- axis) bounds.

[0026] FIG. 7 depicts a graph illustrating comparison of aperiodic slope and oscillatory power for tracking depression severity. Violin plots demonstrate the distribution of correlations between CAT-DI scores and spectral parameters (aperiodic slope and beta power) for every vmPFC electrode (grey circles). Box plots indicate median and IQR for each distribution. Correlations were also computed after averaging spectral parameters across vmPFC electrodes within each subject (numbered circles) Aperiodic slope correlations with CAT-DI were generally negative and correlations with beta power were positive. Low and high beta estimates are further divided into total power and oscillation-only components, and correlations were non-significant (p>0.05) after removal of the aperiodic component. ** p < 0.01; * p < 0.05 (not significant after multiple comparisons correction).

[0027] FIG. 8 depicts relationship of spectral slope and depression severity to time of day. Depression severity and spectral slope (averaged over vmPFC region) were sorted by day (dividing days at 0600) and Z-scored within day. Neither measure showed a significant relationship across patients to time of day. FIGS. 9A-C depict connectivity of biomarker region with affective network structures. Whole brain deterministic tractography was initiated from the vmPFC biomarker region to estimate the structural connectivity of this region to other limbic structures. (A) Regions of interest (ROIs) used in the current study observed in one cerebral hemisphere: biomarker ventromedial prefrontal (vmPFC) region (magenta), dorsal prefrontal (dPFC) (orange), anterior cingulate (ACC) (green), amygdala (red), and mid-temporal (blue). (B) ROIs overlaid on vmPFC-affective network connectivity in the left hemisphere. (C) Bilateral affective network connectivity demonstrating innervation patterns from the uncinate fasciculus (blue), forceps minor (purple), and cingulum bundle (green), isolated through tract clustering algorithm in DSI Studio. Recruitment of these three major fiber systems contributes substantially to treatment efficacy in deep brain stimulation, and appear to work by unifying affective function across the distributed network space of the dorsal ACC, dPFC, amygdala, and vmPFC.

[0028] FIGS. 10A-D depict images and graphs illustrating biomarker properties. (A) Spatial distributions of the putative biomarker of mood by individual intracranial electrode implanted in a neurosurgical patient undergoing a clinical trial for treatment-resistant depression. Specifically, the correlations between depression severity (operationalized by the Computerized Adaptive Test, a validated measure of depression) and 1 / f slope (20-45 Hz) are plotted. (B) Summary of the Depression score (CAT) vs power relationship parametric in frequency and electrode, showing strong high-alpha and beta correlates. (C) Correlation between the power template and the 1 / f template. (D) Scatter plots demonstrating the relationship between depression scale scores and our electrophysiological biomarker (slope in the power spectral density plots averaged across all intracranial recording contacts).

[0029] FIGS. 11A-B depict graphs illustrating the relationship between pupil slop and mood. (A) Plot of the approximate PSD curve for each run of resting state per patient, all patients have one run of resting state except for DBSTRDOIO who has 2 runs. X axis = log(frequency), Y axis = log(power). Shaded region shows the range over which the slope was calculated (20-45 Hz), which indicates that much like in the depression / vmPFC neural data, this range shows a nice, relatively flat / linear slope, which will have a more accurate fit using the polyfit function than if we tried to fit the slope over the full range of frequencies. (B) The relationship between the average slope for each resting state run (x axis) compared to depression severity score (y axis). Again, much like the biomarker paper, as pupil slope becomes flatter (less negative), depression severity decreases between patients (and within our single patient who has 2 runs). The overall correlation is approximately -0.61, with a p-value that is nearly significant (p = 0.051).

[0030] FIGS. 12A-F depicts graphs illustrating mood scores by sex and age. (A) CAT -DI. (B) BDI. (C) PANAS Positive. (D) PANAS Negative. (E) PANAS Ratio. (F) Table of scores.

[0031] FIGS. 13A-B depicts graphs illustrating mood score correlations (1 averaged score per sub). (A) Graphs showing mood score correlations for BDI, PosScore, NegScore, and PANAS Ratio. (B) Table of scores.

[0032] FIGS. 14A-B depicts graphs illustrating mood score correlations (1 score per block, pooled across subs). (A) Graphs showing mood score correlations for BDI, PosScore, NegScore, and PANAS Ratio. (B) Table of scores.

[0033] FIGS. 15A-D depicts graphs illustrating average 1 / f vs mood score (1 score per block, pooled across subjects). (A) PANAS Pos. (B) PANAS Neg. (C) PANAS Ratio. (D) CAT-DI.

[0034] DETAILED DESCRIPTION OF THE INVENTION

[0035] Definitions

[0036] The instant invention is most clearly understood with reference to the following definitions:

[0037] As used herein, the singular form “a,” “an,” and “the” include plural references unless the context clearly dictates otherwise.

[0038] Unless specifically stated or obvious from context, as used herein, the term “about” is understood as within a range of normal tolerance in the art, for example within 2 standard deviations of the mean. “About” can be understood as within 10%, 9%, 8%, 7%, 6%, 5%, 4%, 3%, 2%, 1%, 0.5%, 0.1%, 0.05%, or 0.01% of the stated value. Unless otherwise clear from context, all numerical values provided herein are modified by the term about.

[0039] As used in the specification and claims, the terms “comprises,” “comprising,” “containing,” “having,” and the like can have the meaning ascribed to them in U.S. patent law and can mean “includes,” “including,” and the like.

[0040] Unless specifically stated or obvious from context, the term “or,” as used herein, is understood to be inclusive.

[0041] Ranges provided herein are understood to be shorthand for all of the values within the range. For example, a range of 1 to 50 is understood to include any number, combination of numbers, or sub-range from the group consisting 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 1 1, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, or 50 (as well as fractions thereof unless the context clearly dictates otherwise).

[0042] Detailed Description

[0043] Provided herein are articles and methods of determining the severity of and treating depression. In some embodiments, the method includes obtaining intracranial neural activity data from a subject; determining an aperiodic slope of power spectral density (PSD) from the intracranial neural activity data; determining a depression severity score from the aperiodic slope; and treating the subject based upon the depression severity score. Without wishing to be bound by theory, it is believed that the aperiodic slope of the PSD represents a novel mechanism for determining and tracking depression severity, which facilitates new methods of treating individuals.

[0044] In some embodiments, obtaining the intracranial neural activity data includes recording the intracranial neural activity data from the subject. For example, the intracranial neural activity data may include electroencephalography (EEG) data recorded from implanted stereoelectroencephalography (sEEG) electrodes. This data may be recorded from any suitable target region(s) for determining depression severity score therefrom. In some embodiments, the target region(s) includes the affective and / or reward processing networks. In some embodiments, the target region(s) include ventromedial prefrontal cortex (vmPFC), dorsal prefrontal cortex (dPFC), dorsal anterior cingulate cortex (dACC), amygdala (AMY), mid-temporal regions (MT), or combinations thereof.

[0045] After obtaining the intracranial neural activity data, determining the aperiodic slope includes first estimating the PSD. The PSD may be estimated using any suitable method, such as, but not limited to, Welch’s method or Morlet wavelet decomposition. Next, the PSD is resampled to any suitably spaced frequency bins, such as, but not limited to, linearly spaced frequency bins, logarithmically spaced frequency bins, or a combination thereof. Following the resampling, a fit is performed in log-log (log power and log frequency) over a desired frequency range. The fit may be linear, polynomial, or any other suitable fit. In some embodiments, the desired frequency range includes frequencies over alpha frequencies to avoid the knee and under gamma frequencies to avoid line noise. For example, in some embodiments, the desired frequency range is between 9 and 96 Hz, between 15 and 45 Hz, between 20 and 45 Hz, or any other suitable combination, sub-combination, range, or sub-range thereof. In some embodiments, the aperiodic slope is averaged across electrodes in a region of interest. The slope of the resulting fit is then used to determine the aperiodic slope.

[0046] The aperiodic slope provides a consistent, task-independent, and visually discernible measure of depression severity, where a flatter slope indicates less depression (a lower depression severity score). This measure of depression severity is consistent both within and between patients. Additionally, unlike in some embodiments, a subject’s depression can be quantified based upon the aperiodic slope. For example, in some embodiments, depression may be quantified according to the CAT -DI metric, where a depression severity score of less than 50 is normal, 50-65 is mild, 66-75 is moderate, and greater than 75 is severe depression. However, as will be appreciated by those skilled in the art, the disclosure is not limited to the CAT -DI metric and includes quantification according to any other mood inventory using different numbers for the stratification between mild / moderate / severe.

[0047] Unlike the prior focus on periodic, oscillatory data, which necessarily relied upon group averages, advanced statistical modeling, and required weeks-long assessment periods, the aperiodic slope disclosed herein can be calculated simply using short recordings (e.g., seconds to minutes). As such, the methods disclosed herein facilitate real-time assessment of depression during recording and treatment paradigms, enabling improved treatment. In some embodiments, the methods disclosed herein enable improved selection of treatment approaches, such as, but not limited to, deep brain stimulation (DBS), ECT, TMS, medication, or any other modality. For example, the method may include modifying and / or administering deep brain stimulation (DBS) when treatment resistant depression is detected. In some embodiments, modifying and / or administering DBS includes measuring a subject’s slope, determining that the subject is not improving, and modifying the stimulation parameters (e.g., increasing amplitude or frequency of stimulation). In another example, the slope may be measured non-invasively and treatment may be started or modified (e.g., increasing dose of medication, adding medication, or other adjuvant therapy). In some embodiments, the slope measurement may be used during or after each augmentation of therapy (e.g., combinatorial therapy starts with SSRI only, then increases to SSRI + SNRI / tricyclic, then increases to SSRI+SNRI +tricyclic + antipsychotic) to see whether further additional therapy should be added or if the patient should be maintained at the current therapeutic setting for a trial period. In some embodiments, the methods disclosed herein enable improved selection of target regions to stimulate (e.g., applying electrical stimulation such as DBS, transcranial magnetic stimulation (TMS), temporal interference stimulation (TINS), focused ultrasound (FU), etc to areas / regions exhibiting abnormalities), such as, but not limited to, ventral capsule / ventral striatum (VC / VS); subcallosal cingulate (SCC), superolateral medial forebrain bundle (slMFB); lateral habenula (LHb), or a combination thereof.

[0048] Additionally or alternatively, the methods disclosed herein may include real-time modifications to treatment. For example, in some embodiments, after administering treatment, the method includes monitoring the depression severity score from the aperiodic slope during treatment of the subject; determining changes in the depression severity score during treatment; and quantifying efficacy of the treatment and / or modifying parameters of the treatment according to the changes in the depression severity score. In some embodiments, the method includes closed-loop neural stimulation therapeutics for depression, such as, but not limited to, treatmentresistant depression (TRD).

[0049] Also provided herein, in some embodiments, are methods of determining the severity of and treating depression using proxies of intracranial neural activity data. In some embodiments, the method includes obtaining proxies of intracranial neural activity data from a subject; determining an aperiodic slope of power spectral density from the proxies of intracranial neural activity data; determining a depression severity score from the aperiodic slope; and treating the subject based upon the depression severity score. In some embodiments, the proxies of intracranial neural activity data include data from non-invasive measures. Suitable non-invasive measures include, but are not limited to, pupillometry, fMRI, scalp EEG, heart rate, EKG, skin conductance, magnetoencephalography, or combinations thereof.

[0050] As will be appreciated by those skilled in the art, other than using proxies of the intracranial neural activity data, these methods mirror the methods using direct intracranial neural activity data discussed above. In some such embodiments, the only difference is in the determining of the aperiodic slope. For example, in some embodiments, where the non-invasive measure includes pupillometry data, determining the aperiodic slope includes estimating power spectral density of pupil dilation frequency; resampling the spectra to spaced frequency bins; performing a fit in log-log over a desired frequency range; and determining the slope of the fit. Unless explicitly stated otherwise, all other features / embodiments disclosed above for direct intracranial neural activity apply to the methods using proxies as well.

[0051] The methods and embodiments disclosed herein describe the use of an electrophysiological signal (aperiodic slope) to robustly predict human depression severity and / or psychiatric disorders. As compared to oscillatory features of electrophysiology, the aperiodic slope reflects more of overall excitatory / inhibitory balance or arousal functions of the brain. The aperiodic slope is also easily quantifiable through basic computational transformations, thereby eliminating the need for complex machine learning techniques. The use of aperiodic slope to determine mood has broad applicability to ongoing clinical trials and efforts to treat non-responders to existing therapies for major depressive disorder (MDD). Additionally, the aperiodic slope may by combined with oscillatory components to further understand / distinguish depression and its subtypes.

[0052] The following examples further illustrate aspects of the present invention. However, they are in no way a limitation of the teachings or disclosure of the present invention as set forth herein.

[0053] EXAMPLES

[0054] EXAMPLE 1 - Aperiodic neural activity is a biomarker for depression severity

[0055] ABSTRACT

[0056] BACKGROUND: A reliable physiological biomarker for Major Depressive Disorder is essential for developing and optimizing neuromodulatory treatment paradigms. This study investigates a passive electrophysiologic biomarker that tracks changes in mood state on the order of minutes to hours.

[0057] METHODS: We analyze brief recordings from intracranial electrodes implanted deep in the brain during a clinical trial of deep brain stimulation for treatment-resistant depression in 5 human subjects (nfemaie= 3, nmaie = 2). This surgical setting allows for precise temporal and spatial sensitivity in the ventromedial prefrontal cortex, a challenging area to measure. We focused on the aperiodic slope of the power spectral density, a metric reflecting the balance of activity across all frequency bands and serving as a proxy for excitatory / inhibitory balance in the brain.

[0058] RESULTS: Our findings demonstrate that shifts in aperiodic slope correlate with depression severity, with flatter (less negative) slopes indicating reduced depression severity. This significant correlation was observed in all N=5 subjects, particularly in the ventromedial prefrontal cortex.

[0059] CONCLUSIONS: This biomarker offers a new way to track patient responses to Major Depressive Disorder treatment, paving the way for individualized therapies in both intracranial and non-invasive monitoring contexts.

[0060] INTRODUCTION

[0061] Major Depressive Disorder (MDD) is a debilitating disease with a poor treatment success rate despite its annual global cost approaching $92.7 billion. Even with a range of cognitive and pharmacological treatments, about one in three individuals experiences no meaningful symptom improvement after multiple attempts, and are diagnosed with treatment -resistant depression (TRD). Electrical stimulation-based treatments like transcranial magnetic stimulation (TMS) and deep brain stimulation (DBS) offer hope and present opportunities for direct brain recording. This can facilitate the identification of a physiological biomarker in neural electrophysiology, potentially bypassing limitations of current mood- and treatment-tracking assessments.

[0062] A reliable biomarker would allow quicker therapeutic titration or determination of nonefficacy, enabling patients to rapidly progress through treatment options, thereby moving toward individualized medicine and improving outcomes. Efforts to find a biomarker have explored oscillatory power fluctuations in canonical frequency bands measured from specific affective or reward processing regions during affective tasks. However, an explicit biomarker robustly tracking moment-to-moment mood changes is still lacking. Oscillatory markers for treatment monitoring show inconsistency in frequency range and anatomical location across patients and studies, limiting generalizability.

[0063] In contrast, the aperiodic component (exponent) of power spectral density, known as the 1 / f slope, is a marker of neural activity computed across reliable frequency ranges and robust to narrow-band 60-Hz line noise artifact. It is thought to quantify the excitatory to inhibitory (E / I) balance of neuronal networks engaged in coordinated activity. Variations in the aperiodic slope are linked to age, disease states, and inter-subject variability of depression severity. Notably, a flattening of the aperiodic slope is observed in patients receiving pharmacological treatment for MDD, though this study did not distinguish between responders and non-responders. The aperiodic slope differentiates TRD responders from non-responders when measured from chronically implanted DBS electrodes, though shifts in the slope were only assessed after weeks to months of stimulation. While prior studies demonstrate its capacity to capture long-term mood changes, its potential for tracking depressive symptom severity with high temporal resolution remains unexplored. The aperiodic slope can track large changes in arousal between sleep and wake states, and shifts rapidly in response to visual and auditory stimuli, indicating its potential for rapid mood tracking.

[0064] In this study, our objective is to examine the aperiodic slope as a putative physiological biomarker of mood state in patients with MDD. We investigate whether the aperiodic slope correlates with depressive mood severity, as measured by a rapid adaptive depression inventory (Gibbons RD, Weiss DJ, Pilkonis PA, Frank E, Moore T, Kim JB, et al. (2012): Development of a computerized adaptive test for depression. Arch. Gen. Psychiatry. 69(11): 1104-12.), in patients undergoing intracranial electrophysiological investigation as part of a clinical trial for TRD

[0065] METHODS

[0066] Study Design

[0067] Patient enrollment and depression assessment was equivalent to procedures described in Xiao et al., 2023. Briefly, five TRD patients (age: 32-56 years; biological sex: n = 3 female; race and ethnicity: n=l Hispanic, n=4 non-Hispanic, white) were enrolled in a clinical trial (Study enrollment: August 20, 2019 - August 19, 2025) of individualized deep brain stimulation approved by the Baylor College of Medicine Institutional Review Board. Each subject provided written informed consent to participate in this FDA-approved (IDE number G180300) trial (ClinicalTrials.gov number NCT03437928). The participants had histories of severe treatmentresistant Major Depressive Disorder, meeting inclusion criteria of severity, chronicity, and treatment-refractoriness to be included in the trial. Patients were implanted with permanent deep brain stimulation leads and temporary stereoelectroencephalography (sEEG) recording electrodes in depression-relevant prefrontal regions. Following surgical implant, subjects underwent intensive inpatient monitoring for 9 days, during which frequent measures of depression severity were obtained using the Computerized Adaptive Test-Depression Inventory (CAT-DI), which was previously shown to successfully track mood changes in TRD. CAT-DI delivers a series of individual questions indexing the patient’s current mood state, adaptively sampling from a large question bank based on prior responses and computes a summary estimate of disease severity (0- 100) with scores categorized as normal (<50), mild (50-65), moderate (66-75), or severe (>75). CAT-DI has high test-retest reliability (r = .92), strong correlation (r = 0.75 vs. HAM-D, r = 0.81 vs. PHQ-9) to established clinical scales, and is typically completed within 1-3 minutes. CAT -DI typically asks patient mood over a time frame of 2 weeks. For the current study, the time frame was customized to reflect a 1-hour period. CAT-DI inventories were collected throughout the inpatient stay at the convenience of the patient and the clinical staff. An approximate collection schedule for CAT-DI assessments each day is detailed in FIGS. 1A-B. CAT-DI inventory collection would begin after clinical mood assessments in the morning, occur after each meal or break time and within 1-2 hours in between these periods and finish before task closeout in the evening. Over the 9-day stay, days 1, 2, and 9 were primarily cognitive or affective tasks independent of stimulation paradigms, and days 3-8 were focused on DBS paradigms. FIGS. 1A-B also presents the scores over all 9 days of the inpatient stay to demonstrate the trend in mood scores for each patient.

[0068] Intracranial recordings

[0069] Subsequent to each depression severity (CAT-DI) measurement, signals were recorded from sEEG electrodes while the patient fixated on a crosshair for approximately 1 to 3 minutes. Signals were recorded with a Cerebus data acquisition system (BlackRock Microsystems) at 2 kHz. Some recordings in patient TRD010 were recorded at 30kHz for reasons pertaining to other studies (in which case signals were resampled by decimation to 2 kHz after application of a lowpass Chebyshev Type I infinite impulse response (HR) filter of order 8). All analyses were performed in MATLAB (Mathworks, inc.). Signals were inspected for artifact in the time and frequency domains. Temporal periods with excessive noise across channels were excluded. Individual electrodes with excessive line noise were removed from analyses. Two channels from intracranial leads were used as reference and ground and were thus not included in analysis. Remaining sEEG channels were re-referenced by subtracting the common median across channels at each time point. No stimulation was delivered during any of the recordings. Power spectral density estimation was performed with Welch’s method using one second windows with 50% overlap. Spectra were resampled to logarithmically spaced frequency bins and a linear fit was performed in log-log (log(power) and log(frequency)) using the polyfit function with a degree of one and constrained between 20-45 Hz. The slope of this fit was used to compute correlations between spectral slope and depression severity (FIGS. 2A-3). The same fitting procedure was performed, parametric in bin edge frequencies, for retrospective biomarker optimization (FIGS. 4A-5) and for examining diurnal fluctuations (FIG. 6). We then reanalyzed our dataset using the “fitting-oscillations-and-one-over-f ’ (FOOOF) method (Donoghue T, Haller M, Peterson EJ, Varma P, Sebastian P, Gao R, et al. (2020): Parameterizing neural power spectra into periodic and aperiodic components. Nat. Neurosci. 23(12): 1655-65.) as implemented in FieldTrip (Oostenveld R, Fries P, Maris E, Schoffelen J-M (2011): FieldTrip: open source software for advanced analysis of meg, eeg, and invasive electrophysiological data. Comput. Intel!. Neurosci. 2011 : 156869.). To explicitly decompose beta frequencies (FIG. 7), the lower limit of our FOOOF frequency bin was reduced (15-45 Hz) to cover the edge of the beta oscillatory range.

[0070] Electrode localization and ROI Grouping

[0071] Electrode coordinates were obtained as previously described (Metzger BA, Kalva P, Mocchi MM, Cui B, Adkinson JA, Wang Z, et al. (2023): Intracranial stimulation and eeg feature analysis reveal affective salience network specialization. Brain. 146(10): 4366-77.). After transformation to MNI space, electrodes were visualized on the FSaverage surface. Group level results (FIGS. 4A-D) were obtained on an electrode-wise basis (using patient as repeated measure in FIG. 4A) given the near identical electrode configuration across patients. While electrodes were targeted to the same anatomic structures, there were differences in final affine- registered electrode positions due to residual anatomical variability and the need to adjust trajectories to avoid vasculature. The RMS variability in electrode coordinate locations is shown in supplemental FIG. 8 regions were manually parceled and verified by two experts (RM, BS or RM, GB). Regions of interest (ROIs) in the current study include 1) the ventromedial prefrontal cortex (vmPFC) across Brodmann areas 10, 24, 25, and part of 32, which contains the most anterior and ventral aspect of the anterior cingulate cortex (ACC); 2) the dorsal ACC (dACC) across area 24 and part of area 32; 3) the dorsal PFC (dPFC) across areas 8, 9, and 46; 4) the amygdala (AMY); and 5) the mid-temporal regions (MT, Temporal). Deterministic fiber tracking initiated from patient-defined bilateral vmPFC contacts highlights the engagement of three major affective network fiber systems, including the cingulum bundle (green), forceps minor (purple), and uncinate fasciculus (blue), in facilitating structural connectivity across ROIs (FIGS. 9A-C). The vmPFC region was transformed into ICBM152 space, and tractography was conducted in a population-averaged diffusion template available in DSI Studio (https: / / dsi-studio.labsolver.org). Individual fiber systems were isolated and identified through a clustering method in DSI Studio software, then visually-inspected.

[0072] Statistical Analyses

[0073] To test for correlations between depression severity scores and spectral slope estimates, slope estimates for individual electrodes averaged across the ventral PFC region. Correlation between the slope and the depression severity were estimated as the Pearson product-moment correlation coefficient. To assess significance of findings on an individual subject bases, permutation testing was performed for each electrode per patient. A distribution of 5000 pseudocorrelation values using the actual, ordered vector of depression severity scores and a randomly shuffled vector of the aperiodic slope values. A p-value was determined by identifying the percentile of the actual Pearson correlation for each electrode within the corresponding pseudodistribution, and the resulting total number of electrodes which showed a significant slopeseverity relationship after permutation testing are listed in Table 1. A binomial test was performed within each patient to identify if the number of electrodes showing a slope-severity relationship after permutation testing was statistically significant, given the total number of channels. Regional analyses were performed to determine aggregate statistics across ventral PFC, dorsal PFC, amygdala, ACC, and Temporal regions. Visualization was performed with Violinplot (https: / / github.com / bastibe / Violinplot-Matlab). Regional differences in aggregate correlations were assessed using a one-way ANOVA using region as response variable. Violin plots were also created to compare the distribution of depression severity correlations for slope in the ventral PFC to those of beta power (with and without detrending with FOOOF). To track individual subjects across distributions, the correlation of subject average spectral parameters to depression severity are also plotted (large markers in FIG. 7 were computed similarly to those in FIG. 2C). Significance testing for the population was computed based on the subject-wise results with n=5, (given non-independence of multiple electrodes within a patient) using one- sample Student’ s t-test against a null hypothesis of zero mean and a Bonferroni factor of 5 for multiple comparisons. Table 1. Permutation testing for slope-severity correlation per channel.

[0074] Subject 1 Subject 2 Subject s Subject 4 Subject s

[0075] Total # Channels 179 176 184 180 172

[0076] Original # Significant ?0 68 29 62

[0077] Channels

[0078] Permutation Test

[0079] ^PosWv coiSadon1 / 887(1 / 86) 69(3 / 66) 67(7 / 60) 25(0 / 25) 63(24 / 39)

[0080] Negative Correlation)

[0081] Patient Significance

[0082] (Binomial Result of p- <0.001 <0.001 <0.001 <0.001 <0.001 value)

[0083] Original statistical results were calculated by correlating aperiodic slope values for each electrode with depression severity score per patient. Permutation testing was then performed for each electrode per patient by creating a distribution of 5000 pseudo-correlation values using the actual, ordered, vector of depression severity scores and a randomly shuffled vector for the aperiodic slope values. There are a similar number of channels with a significant slopeseverity relationship before and after permutation testing, and each patient showed a significant number of electrodes which detected the slope-severity relationship.

[0084] RESULTS

[0085] Aperiodic slope as an index for depression severity

[0086] We found that depression severity scores strongly correlated with the aperiodic slope, especially in the ventromedial prefrontal cortex, a prominent node in the Affective Salience Network (ASN). By focusing on short-term mood changes, we identify a marker for moment-to- moment changes in ASN cortical state that reliably tracks short-term mood fluctuations and is generalizable across patients. Notably, this biomarker is detected from brief recordings (1-5 minutes long).

[0087] This investigation included five patients undergoing simultaneous implantation of stereo¬

[0088] EEG and DBS electrodes for treatment of TRD as a function of an experimental clinical trial (NCT03437928). Stereo-EEG electrodes placed in target regions of the affective and reward processing networks (vmPFC, dPFC, dACC, AMY, MT; see methods for regional definitions) were used to record local field potentials (LFPs) during each completion of the CAT-DI across a 9-day in-patient hospital visit (Table 2). Studies examining the aperiodic component typically constrain slope calculations to under ~50 Hz to avoid the impact of line noise on the slope calculation. We further imposed a limitation on the lower end of the range to help mitigate the effect of large alpha and low-beta features that give rise to a ‘knee frequency’. Examination of the slope metrics within and across our patients showed a consistent fit within the 20-45 Hz subrange for each patient. CAT-DI depression severity scores were then correlated with slope values at each electrode contact within patient (FIG. 2A), showing the strongest average correlations within the vmPFC. Each patient showed a flattening (reduction of magnitude of negative value) of the aperiodic slope in the ‘biomarker’ region (vmPFC) between the highest [70-96] and lowest [27-68] depression severity scores from the entire in-patient stay (FIG. 2B). Aperiodic slope was averaged across all electrodes within the vmPFC region and related to depression severity scores in each patient in FIG. 2C, yielding r=[-0.58 to -0.34] (p=[0.0004 to 0.017]), indicating significantly flatter slope during more euthymic (less depressed) mood states in every subject.

[0089] Table 2. Frequency and average length of CAT-DI assessments per patient.

[0090] Subject 1 Subject 2 Subject s Subject 4 Subject s

[0091] Assessment3 ? 30 48 50 65

[0092] Count

[0093] Assessment

[0094] Duration 201.80(84.38) 75.77(32.38) 120.30(80.80) 101.96(31.96) 114.27(48.75) ( sec j

[0095] Mean(SD)

[0096] We also examined the relationship of aperiodic slope to depression severity when averaging across all electrodes and for individual ROIs (see ROI definition in FIG. 4A and supplemental results in FIG. 3). Control analysis using permutation testing confirm these findings, and a binomial cumulative distribution function shows a significant slope-severity relationship within each patient (p=0.00 to 7.5 le-11; Table 1).

[0097] Affective-regional specificity of the aperiodic biomarker across patients.

[0098] Typically, locations for sEEG electrode placement are idiosyncratic to clinical hypotheses for individual patients undergoing invasive monitoring for epilepsy, resulting in a high degree of spatial variability in sampling. Here, sEEG target regions were determined a priori and trajectories are only altered slightly based on anatomical variation and vasculature, thus allowing for consistent sampling across regions. This offers a unique opportunity for a group-level electrode co-regi strati on, such that each electrode site can be averaged in space across patients (FIG. 4A). Despite some variability in location after co-regi strati on (FIG. 8), combining correlations across electrodes again indicates that the largest effect sizes for the slopedepression correlations were consistently located in the vmPFC (FIG. 4B). To further characterize any regional specificity, electrodes were grouped by subregion (vmPFC, dPFC, AMY, dACC, and MT). A violin plot showing point aggregation represents the spread of correlations for all electrodes across all subjects grouped by ROI, and indicates the aperiodic biomarker for mood is strongest in the vmPFC (slope vs. depression severity correlation median=-0.35, IQR=0.18) and weakest in lateral temporal contacts (median=-0.06, IQR=0.24) (FIG. 4C). A one-way ANOVA revealed a significant effect of group on the response variable, F(4,540)=16.03, p<3e-12. Notably, while the overall effect in the amygdala showed a moderate relationship between aperiodic slope and depression severity, group-wise visual inspection revealed a consistent lateralization in the amygdala; the left amygdala showed a strong, negative relationship between aperiodic slope and depression severity over subjects (r=-0.35±0.19, p=4e- 6), whereas the right amygdala showed little to no effect (r=-0.17±0.29, p=0.02), as demonstrated in FIG. 4B. To confirm that our a priori frequency range was appropriate for slope calculations, we next correlated depression severity scores with vmPFC aperiodic slope across a broad frequency range (fiow=l to 99 Hz, fhigh= flow to 100 Hz). The a priori range spanned the beta and low gamma ranges (white square, 20-45 Hz; FIG. 4D) was well within the broad range of frequencies showing a significant relationship between slope and depression severity. The boundary designated by the darkest blue region in FIG. 4D indicates there is a set of frequency ranges that may be more optimal than (yet still similar to) the a priori range for detecting the slope-severity relationship.

[0099] Patient-specific biomarker selection

[0100] We next sought to examine the strength with which a selected biomarker would track each individual’s mood over the in-patient stay. We examined the relationship between each vmPFC spectral slope and depression severity for each combination of lower and upper frequency bound (similar to FIG. 4C) for each patient (FIG. 5).

[0101] Biomarker free of diurnal fluctuations

[0102] Given previous findings suggesting that aperiodic slope may be a proxy for arousal state, we aimed to identify if the fluctuations seen in the slope biomarker were due to circadian changes in arousal throughout the day. We binned severity and slope by time of day to inspect the relationship between depression severity and spectral slope (both normalized by Z-scoring within day, FIG. 6) and did not find any significant diurnal relationship for either variable.

[0103] DISCUSSION

[0104] The poor treatment success rate in MDD highlights the necessity for individualized therapeutic approaches. A task-independent physiological biomarker allows an unprecedented opportunity to identify and intervene with at-risk patients. It helps mitigate biases in symptom quantification from social, physical, and emotional factors and enables faster identification of treatment-resultant mood changes compared to the typical 2-week assessment period, facilitating more efficient and data-driven adjustments of treatment parameters across individuals. In this study, we demonstrate that the aperiodic component of the intracranial electrophysiological power spectral density function effectively tracks fluctuations in conscious, patient-reported accounts of mood on the order of minutes to hours, displaying a relatively global effect within and across patients when fitting the slope over one pre-specified range (20-45 Hz). We further show that the aperiodic slope fit can be performed with minutes-long recordings, its relationship to depressive mood severity is evident in each patient, and the pre-specified range can be further refined to increase the SNR within each patient.

[0105] Periodic and Aperiodic Biomarkers

[0106] Previous studies searching for physiological biomarkers have primarily focused on periodic, oscillatory data, identifying potential biomarkers for treatment response and mood in various frequency bands (theta, alpha, beta, high and low gamma) across affective regions. Previous studies have necessarily relied on group averages (i.e., depressed vs. non-depressed) and often require advanced computational modeling for detection and prediction. Variability in anatomical location and frequency range of oscillatory biomarkers limits clinical feasibility. In contrast, the relationship between the aperiodic slope and depression severity is broadly evident across vmPFC electrodes and should be easily reproducible with limited stereotactic accuracy. The optimal frequency range for slope estimation varies per individual at the group level, but a broad range of frequency pairs (darkest blue area in FIG. 4D) produces a reliable feature for detecting the slope-severity relationship using only 2-4 minute-long recordings. This allows for real-time assessment during recording and stimulation paradigms.

[0107] Our previous study using a canonical frequency band-based approach revealed a consistent ‘high-high low-low’ pattern in prefrontal regions across an initial cohort of 3 participants (Subjects 1 - 3 in the current study), indicating increased high-frequency power coupled with reduced low-frequency power in the ACC correlated with decreased depression severity. However, this approach did not remove the aperiodic slope, potentially reflecting broad-band shifts rather than the oscillatory components themselves. In the current study, we are able to examine the extent to which previous findings suggesting beta power as a potential biomarker for mood state may indeed demonstrate a shift in aperiodic slope, rather than in the oscillation itself. While total low or high beta power do show a similar magnitude of correlation with CAT -DI scores the aperiodic slope estimate (FIG. 7), removal of the aperiodic slope from these estimates shows that beta oscillations alone are not sufficient to explain this relationship. Thus, we provide further evidence indicating that former oscillatory band biomarkers can be better described by changes in aperiodic activity rather than in terms of power of oscillations in canonical bands.

[0108] Other studies, such as that by Veerakumar et al. (2019), show a shift in the aperiodic component after weeks to months of treatment in responders, but in the opposite direction of the current study (e.g. steeper slope in response, flatter slope in non-response). The difference in direction between this early report and the group’s more recent findings is a matter of substantial scientific interest. The group’s most recent study suggested that the directionality of neural activity responses to DBS may change between acute stages (as observed in the present study) and chronic stages (Veerakumar et al. (2019)). Alagapan et al. (2023) further found a trend in spectral features which generally reflected an increase in low-frequency activity and a decrease in high-frequency activity between the ‘sick’ (first 4 weeks of the study), and ‘stable’ (last 4 weeks of the study) states. One interpretation is perhaps changes in the acute stage of illness correspond to increased phasic activity in the affective circuit due to the immediate implementation of stimulation, which is followed by a compensatory reorganization of this network due to remyelination of targeted tracts or engagement of homeostatic plasticity mechanisms to produce long-term changes. Future studies should aim to examine the longitudinal changes in aperiodic activity with chronic stimulation to identify if and how this biomarker tracks mood severity between early and late treatment response stages.

[0109] Regional and Frequency Specificity of the Aperiodic Biomarker

[0110] Prior studies focused on non-invasive methods and recordings from sub-optimal anatomical sites using complex artifact rejection techniques. Our clinical trial captured recordings from grey matter through sEEG electrodes placed in tandem with DBS sensing and stimulating electrodes, enabling enhanced sensitivity to local field potential (LFP) activity. Grouping sEEG electrodes into five ROIs, we observed a consistent phenomenon across a broad frequency range and most anatomical regions that robustly tracked mood (FIGS. 2A-3). While the optimal frequency range for calculating slope (FIGS. 2C and 3) varied slightly across patients and regions, the a priori (20-45 Hz) frequency range was robust across all patients and regions. This global characteristic of the aperiodic slope allows tracking of depression severity by averaging aperiodic slope over all electrodes within this frequency range.

[0111] The strongest relationship between depression severity and aperiodic slope was consistently observed in the vmPFC, aligning with its known role in reward, social, and affective processing, and its position in the default mode network. We visualized the recruitment of affective network-related white matter tracts through whole-brain deterministic tractography initiated from the vmPFC region, highlighting the dense innervation of the cingulum bundle, forceps minor, and uncinate fasciculus. These fiber systems represent three of the four white matter bundles targeted in neurosurgical planning for subcallosal cingulate deep brain stimulation for TRD. Further, the cingulum bundle and forceps minor have been identified as the likely stimulation pathways producing stable, efficacious responses in SCC DBS. Together, the cingulum bundle, forceps minor, and uncinate fasciculus facilitate structural connectivity from the vmPFC to ACC, DPFC, and amygdala regions. Thus, the “biomarker region” in the current study may represent a hub region of integration across multiple streams of input and output of the affective networks (FIGS. 9A-C).

[0112] The left amygdala showed a strong relationship between aperiodic slope and depression severity, consistent with amygdala asymmetry reported in MDD. Prior studies demonstrate emotional processing responsivity in the left amygdala in MDD which normalized after effective treatment. Further, left amygdala activity during positive memory recall has tracked with depression severity. The lateral temporal lobe showed the weakest relationship with depression severity, serving as an effective negative control.

[0113] The aperiodic slope overcomes limitations of current diagnostic and treatment-tracking paradigms by being passively collected during patient monitoring, robust across frequency and anatomical regions, and highly correlated with short-term changes in mood severity. CONCLUSION

[0114] This study presents compelling evidence for the aperiodic slope as a reliable and temporally sensitive physiological biomarker for depressive mood severity in patients suffering from TRD. Our results expand on prior findings of aperiodic slope responsiveness to long-term mood changes following MDD treatment and demonstrate its efficacy in tracking mood changes at high temporal resolution. Aperiodic slope provides a consistent, task-independent measure of depression severity across patients, eliminating the need for complex machine learning techniques. This biomarker has the potential to revolutionize MDD treatment by facilitating efficient screening of therapeutic approaches, individualizing treatment strategies, and enabling active control of the dysfunctional affective system in TRD.

[0115] EXAMPLE 2

[0116] This Example describes the discovery of an electrophysiological signal that robustly predicts human depression severity. Referring to FIGS. 10A-D, this signal is easily quantifiable through basic computational transformations, but had not previously been identified as predictive of mood. Prior efforts have looked at oscillatory features of electrophysiology, but here the focus is instead on the aperiodic element of the power spectral density curve, which is believed to reflect more of overall excitatory / inhibitory balance or arousal functions of the brain. This biomarker of mood has broad applicability to ongoing clinical trials and efforts to treat 230 million people globally who are non-responders to existing therapies for major depressive disorder. Ongoing clinical trials and device companies have been actively pursuing these biomarkers over the past 8 years, and it is interpreted as widely implementable and commercializeable.

[0117] Having identified this biomarker and developed the software architecture to rapidly monitor it opens the door to closed-loop neural stimulation therapeutics for treatment-resistant depression, a goal of the field over the past 4-6 years. These data have never before been available for characterization, as they were collected in the context of a first-in-human clinical trial of intracranial electrophysiological monitoring in patients with treatment-resistant depression. As shown in FIG. 10B, a summary of the depression score (CAT) vs. power relationship parametric in frequency and electrode show strong high-alpha and beta correlates justifies the selected frequency bin. Turning to FIG. 10C, the correlation between the power template and the 1 / f template is only -0.36, which seem to give somewhat different answers in terms of spatial templates. FIG. 10D shows that the relationship between depression scale scores and the electrophysiological biomarker holds in all five patients in whom intracranial neural data and densely-sampled measures of depression are available.

[0118] EXAMPLE 3

[0119] This Example describes tracking of the slope of pupil dilation with mood ratings across patients. Referring to FIGS. 11A-B, individual patients were plotted with different colors indicated in the legend (strings starting with Y indicate epilepsy patients and those starting with TRD indicate those with treatment resistant depression). More specifically, FIG. 11A shows the approximate power spectral density (PSD) curve for each run of resting state per patient, with all patients having one run of resting state except for DBSTRD010 who has 2 runs. FIG. 11B shows the relationship between the average slope for each resting state run (x axis) compared to depression severity score (y axis). The shaded region in FIG. 11A shows the range over which the slope (FIG. 11B) was calculated (20-45 Hz). Much like in the depression / vmPFC neural data, this range shows a nice, relatively flat / linear slope, which provides a more accurate fit using the polyfit function (as compared to the slope over the full range of frequencies). As pupil slope becomes flatter (less negative), depression severity decreases between patients (and within our single patient who has 2 runs). The overall correlation is approximately -0.61, with a p-value that is nearly significant (p = 0.051).

[0120] These data suggest that the pupil can serve as a proxy of intracranial neural activity, and as such, the discovery of intracranial slope tracking with mood can be extended to encompass peripheral (non-invasive) measures as well. Without wishing to be bound by theory, it is believed that the biomarker disclosed herein can be used non-invasively to track mood, quantify treatment efficacy, and / or identify which patients are most likely to be responders to different therapies.

[0121] EXAMPLE 4

[0122] This Example describes various properties of a non-treatment-resistant-depression (non- TRD) slope biomarker. FIGS. 12A-14B show graphs illustrating demographics and behavioral distribution. As shown in FIGS. 12A-F, age and sex had no effect on mood scores for any inventory. FIGS. 13A-B shows that PANAS Negative scores and the Ratio between Positive and Negative PANAS scores are significantly correlated with CAT -DI scores, while FIGS. 14A-B shows that PANAS scores and the Ratio between Positive and Negative PANAS scores are significantly correlated with CAT-DI scores. FIGS. 15A-D show graphs illustrating slope vs mood scores per region. The regions include average, vmPFC (not including anterior cingulate contacts), ACC, dPFC, amygdala, and anterior insula.

[0123] EQUIVALENTS

[0124] Although preferred embodiments of the invention have been described using specific terms, such description is for illustrative purposes only, and it is to be understood that changes and variations may be made without departing from the spirit or scope of the following claims.

[0125] INCORPORATION BY REFERENCE

[0126] The entire contents of all patents, published patent applications, and other references cited herein are hereby expressly incorporated herein in their entireties by reference.

Claims

CLAIMS1. A method of treating depression, the method comprising: obtaining intracranial neural activity data from a subject; determining an aperiodic slope of power spectral density from the intracranial neural activity data; determining a depression severity score from the aperiodic slope; and treating the subject based upon the depression severity score.

2. The method of claim 1, wherein obtaining the intracranial neural activity data includes recording the intracranial neural activity data from the subject.

3. The method of claim 1 or 2, wherein the intracranial neural activity data includes electroencephalography (EEG) data.

4. The method of claim 3, wherein the EEG data is recorded from target regions of the affective and reward processing networks.

5. The method of claim 4, wherein the target regions include ventromedial prefrontal cortex (vmPFC), dorsal prefrontal cortex (dPFC), dorsal anterior cingulate cortex (dACC), amygdala (AMY), mid-temporal regions (MT), or combinations thereof.

6. The method of claim 4 or 5, wherein the EEG data is recorded from implanted stereoelectroencephalography (sEEG) electrodes.

7. The method of any one of claims 1-6, wherein determining the aperiodic slope includes: estimating power spectral density; resampling the spectra to spaced frequency bins; performing a fit in log-log over a desired frequency range; and determining the slope of the fit.

8. The method of claim 7, wherein the power spectral density is estimated with Welch’s method or Morlet wavelet decomposition.

9. The method of claim 7 or 8, wherein the spaced frequency bins are linearly or logarithmically spaced frequency bins.

10. The method of any one of claims 7-9, wherein the fit is linear or polynomial.

11. The method of any one of claims 1-10, wherein the aperiodic slope is calculated in a frequency range over alpha frequencies and under gamma frequencies.

12. The method of claim 11, wherein frequency range is between 9 and 96 Hz.

13. The method of claim 12, wherein the aperiodic slope is calculated in the range of between20 and 45 Hz.

14. The method of any one of claims 1-13, wherein the aperiodic slope is averaged across electrodes in a region of interest.

15. The method of any one of claims 1-14, wherein the depression severity score is lower when the aperiodic slope is flatter.

16. The method of any one of claim 1-15, wherein treating the subject based upon the depression severity score includes administering deep brain stimulation (DBS) when the depression severity score indicates severe depression, treatment resistant depression, or a combination thereof.

17. The method of claim 16, wherein the DBS is applied to areas exhibiting abnormalities.

18. The method of claim 16 or 17, wherein the DBS is applied to the ventral capsule / ventral striatum (VC / VS); subcallosal cingulate (SCC), superolateral medial forebrain bundle (slMFB); lateral habenula (LHb), or a combination thereof.

19. The method of any one of claims 1-18, further comprising: determining the depression severity score from the aperiodic slope during treatment of the subject; determining changes in the depression severity score during treatment; and quantifying efficacy of the treatment.

20. The method of any one of claims 1-19, further comprising: determining the depression severity score from the aperiodic slope during treatment of the subject; determining changes in the depression severity score during treatment; and modifying parameters of the treatment according to the changes in the depression severity score.

21. A method of treating depression, the method comprising: obtaining proxies of intracranial neural activity data from a subject; determining an aperiodic slope of power spectral density from the proxies of intracranial neural activity data; determining a depression severity score from the aperiodic slope; and treating the subject based upon the depression severity score.

22. The method of claim 21, wherein the proxies of intracranial neural activity data include data from non-invasive measures.

23. The method of claim 22, wherein the non-invasive measures include pupillometry, fMRI, scalp EEG, heart rate, skin conductance, magnetoencephalography, or a combination thereof.

24. The method of claim 23, wherein the non-invasive measure includes pupillometry data.

25. The method of claim 24, wherein determining the aperiodic slope includes: estimating power spectral density of pupil dilation frequency; resampling the spectra to spaced frequency bins; performing a fit in log-log over a desired frequency range; anddetermining the slope of the fit.

26. The method of claim 25, wherein the power spectral density is estimated with Welch’s method or Morlet wavelet decomposition.

27. The method of claim 25 or 26, wherein the spaced frequency bins are linearly or logarithmically spaced frequency bins.

28. The method of any one of claims 25-27, wherein the fit is linear or polynomial.

29. The method of any one of claims 25-28, wherein the desired frequency range is over alpha frequencies and under gamma frequencies.

Images