Data-driven optimization of ultrasound parameters for personalized neuromodulation

Abstract

Devices, systems, software, and methods are provided for treating neurological and neuropsychiatric disorders with ultrasound neuromodulation. In particular, a Bayesian optimization approach is used to tune ultrasound stimulation parameters to optimize the efficacy of treatment of neurological and neuropsychiatric disorders with ultrasound neuromodulation.

Description

DATA-DRIVEN OPTIMIZATION OF ULTRASOUND PARAMETERS FOR PERSONALIZED NEUROMODULATIONCROSS-REFERENCE TO RELATED APPLICATION

[0001] This application claims benefit of provisional application 63 / 697,327, filed September 20, 2024, which application is hereby incorporated by reference in its entirety.BACKGROUND OF THE INVENTION

[0002] Focused ultrasound (FUS) neuromodulation is emerging as a promising technology for targeted and noninvasive neurointervention (Rabut et al. (2020) Neuron 108:93-1 10, Beisteiner et al. (2023) Adv Sci (Weinh) 10(14):e2205634, Kubanek (2018) Neurosurg. Focus. 44(2): E14). The ability of FUS to transiently modulate neural activity in deep brain structures with high spatiotemporal specificity complements current clinical modalities like transcranial magnetic stimulation (TMS) and electrical deep brain stimulation (DBS). Preliminary studies indicate that FUS neuromodulation can be effectively applied to treat many neurological and neuropsychiatric disorders (Fan et al. (2024) Brain Stimul. 17(5):1001 -1004; Riis et al. (2023) J Med Case Rep 2023;17(1 ): 4-9, Reznik et al. (2020) Neurol Psychiatr Brain Res 2020;37(June):60-66, Ziebell et al. (2023) Brain Stimul. 16(5) : 1278-1288). However, significant challenges remain to perform an exhaustive, subject-specific mapping of FUS neuromodulation parameters to maximize therapeutic efficacy.SUMMARY OF THE INVENTION

[0003] Devices, systems, software, and methods are provided for treating neurological and neuropsychiatric disorders with ultrasound neuromodulation. In particular, a Bayesian optimization approach is used to tune ultrasound stimulation parameters to optimize the efficacy of treatment of neurological and neuropsychiatric disorders with ultrasound neuromodulation.

[0004] In one aspect, a method for treating a neurological or neuropsychiatric disorder in a subject is provided, the method comprising: (a) administering ultrasound neuromodulation to the subject using an initial selected set of ultrasound stimulation parameters; (b) measuring a response of the subject to said administering the ultrasound neuromodulation using the initial selected set of ultrasound stimulation parameters, wherein the measured response is used to calculate an efficacy score for treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation using the initial selected set of ultrasound stimulation parameters; (c) using the efficacy score from said administering the ultrasound neuromodulation using the initial selected setof ultrasound stimulation parameters to generate a Gaussian process regression model of an objective function to simulate responses of the subject to ultrasound neuromodulation administered using other ultrasound stimulation parameters; (d) performing Bayesian optimization with the Gaussian process regression model of the objective function using an acquisition function that selects a new set of ultrasound stimulation parameters with a goal of improving efficacy of treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation; (e) administering ultrasound neuromodulation to the subject using the new set of ultrasound stimulation parameters; (f) measuring a response of the subject to said administering the ultrasound neuromodulation using the new set of ultrasound stimulation parameters; (g) calculating an efficacy score for treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation using the new set of ultrasound stimulation parameters; (h) updating the Gaussian process regression model of the objective function with the efficacy score for the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation using the new set of ultrasound stimulation parameters, wherein the Gaussian process regression model of the objective function uses all the efficacy scores that have been calculated for the responses to all previous administrations of the ultrasound neuromodulation to the subject with previously selected ultrasound stimulation parameters; (i) performing Bayesian optimization with the updated Gaussian process regression model of the objective function, wherein the acquisition function selects another new set of ultrasound stimulation parameters with the goal of further improving the efficacy of the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation; (j) repeating steps (e)-(i) iteratively to further optimize the ultrasound stimulation parameters until a stopping criterion is met, wherein the acquisition function selects a final set of optimized ultrasound stimulation parameters; and (k) administering ultrasound neuromodulation to the subject using the final set of optimized ultrasound stimulation parameters.

[0005] In certain embodiments, the stopping criterion is determined by a strategy based on probabilistic regret bounds, wherein the probabilistic regret bound statistical test comprises: defining a regret threshold; estimating probability that the regret threshold can be surpassed by repeating steps (e)-(i) after each iteration of step (j); discontinuing the repeating of steps (e)-(i) when the estimated probability that the regret threshold can be surpassed falls below a chosen confidence level.

[0006] In certain embodiments, the stopping criterion is based on acquisition function convergence, wherein steps (e)-(i) are repeated until the acquisition function values converge below a predetermined threshold.

[0007] In certain embodiments, the stopping criterion is based on uncertainty reduction, wherein steps (e)-(i) are repeated until posterior variance around an estimated optimum is below a predetermined threshold.

[0008] In certain embodiments, the method further comprises performing cross-validation or bootstrapping, wherein the stopping criterion is based on stability of an identified optimum, wherein steps (e)-(i) are repeated until re-sampling shows stability of the identified optimum.

[0009] In certain embodiments, steps (e)-(i) are repeated until convergence to an optimized target brain region and ultrasound stimulation parameters that are effective for treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation.

[0010] In certain embodiments, the stopping criterion is that further Bayesian optimization of the ultrasound stimulation parameters no longer results in further improvement of the efficacy of the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation.

[0011] In certain embodiments, the stopping criterion is based on a predetermined optimization budget.

[0012] In certain embodiments, steps (e)-(i) are repeated up to 10 times, up to 20 times, up to 30 times, up to 40 times, up to 50 times, up to 75 times, up to 100 times, up to 125 times, or up to 150 times. In some embodiments, the method is repeated 10 to 100 times, 20 to 80 times, or 30 to 40 times, including any number of times within these ranges such as 10, 11 , 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, 50, 60, 70, 80, 90, or 100 times.

[0013] In certain embodiments, the acquisition function is a noisy expected improvement (qNEI) acquisition function, a max-value entropy search (MES) acquisition function, or ajoint entropy search (JES) acquisition function.

[0014] In certain embodiments, the method further comprises optimizing the parameters of the Gaussian process and of the acquisition function based on a model of the noise present in the data. In some embodiments, the Gaussian noise is modeled from a normal distribution N(0, o), wherein o is equal to a selected noise fraction.

[0015] In certain embodiments, the initial selected set of ultrasound stimulation parameters is selected using random sampling, Sobol sequence sampling, a pseudo-random distribution, or Voronoi parcellation.

[0016] In certain embodiments, the method further comprises repeating steps (a)-(k) at a later time to continually optimize the ultrasound stimulation parameters for the subject. In some embodiments, disease progression in the subject, a change in medication administered to the subject, or adaptation of the subject to the ultrasound neuromodulation have diminished the efficacy of the treatment of theneurological or neuropsychiatric disorder with the ultrasound neuromodulation over time. Optimization of the ultrasound stimulation parameters can be performed repeatedly to account for time-dependent factors, such as disease progression, changes in treatment, and adaptations to the stimulation, which may reduce therapeutic efficacy of the ultrasound neuromodulation with a particular parameter setting over time, wherein repeated or continual optimization of the ultrasound stimulation parameters may benefit the subject.

[0017] In certain embodiments, a model of the objective function estimated from a group of previously treated subjects with a given neurological or neuropsychiatric disorder is used to initialize the Gaussian process regression.

[0018] In certain embodiments, performing Bayesian optimization comprises optimizing hyperparameters selected from the acquisition function, seed queries (initial selected parameters), and an exploration / exploitation ratio.

[0019] In certain embodiments, performing Bayesian optimization comprises optimizing at least 2, at least 3, or at least 4 ultrasound stimulation parameters selected from a target brain region, a stimulation duration, an acoustic intensity, an acoustic pressure, an excitation voltage, a pulse repetition frequency, a pulse length, an ultrasound frequency, and a duty cycle. In some embodiments, the ultrasound stimulation parameters comprise the target brain region, the stimulation duration, and the excitation voltage. In some embodiments, the ultrasound stimulation parameters comprise the target brain region, the stimulation duration, the acoustic intensity, the acoustic pressure, the excitation voltage, the pulse repetition frequency, the pulse length, the ultrasound frequency, and the duty cycle.

[0020] In certain embodiments, the response is a body movement, a facial movement, a change in locomotor activity, a change in mood, or a change in brain electrical activity.

[0021] In certain embodiments, the body movement, the facial movement, or the change in locomotor activity is measured using an accelerometer, a gyroscope, or a video recording device that records images of the subject. In some embodiments, the accelerometer or the gyroscope is provided by a wearable device. In some embodiments, the wearable device is a smartwatch. In some embodiments, the video recording device is provided by a digital camera, a smartphone, a tablet, a laptop, or a camcorder.

[0022] In certain embodiments, brain electrical activity is measured by electroencephalography (EEG), stereoelectroencephalography (sEEG), electrocorticography (ECoG), magnetoencephalography (MEG), single photon emission computed tomography (SPECT), functional magnetic resonance imaging (fMRI), or functional ultrasound imaging (fUSI).

[0023] In certain embodiments, the neurological disorder is a movement disorder, and the ultrasound stimulation parameters are optimized to ameliorate a symptom of the movement disorder. In some embodiments, the symptom is bradykinesia, dyskinesia, essential tremor, or an abnormal or involuntary facial movement. In some embodiments, the response is measured using an accelerometer, a gyroscope, or a video recording device that records images of the subject. In some embodiments, the response is measured using a Movement Disorder Society-Sponsored Revision of the Unified Parkinson's Disease Rating Scale (MDS-UPDRS), a Hoehn and Yahr (HnY) scale, a Clinical Rating Scale for Tremor (CRST), a Parkinson’s Disease Composite Scale (PDCS), or a Schwab and England Activities of Daily Living (ADL) Scale for assessing treatment of a movement disorder.

[0024] In certain embodiments, the neuropsychiatric disorder is depression or anxiety, and the ultrasound stimulation parameters are optimized to ameliorate the depression or the anxiety. In some embodiments, the change in mood is measured using a Visual Analogue Scale for Depression (VAS- D), a Hamilton Depression Rating Scale (HAM-D), Montgomery-Asberg Depression Rating Scale (MADRS), a Beck Depression Inventories ( BD I) score, a Visual Analogue Scale for Anxiety (VAS-A) or a Beck Anxiety Inventories (BAI) score. In some embodiments, the change in mood is measured based on facial expression detection or facial emotion recognition analysis of a video recording of the subject’s face.

[0025] In certain embodiments, the neurological disorder is neuropathic pain, and the ultrasound stimulation parameters are optimized to ameliorate the neuropathic pain. In some embodiments, the response is measured using a numerical rating scale (NRS), a visual analog scale (VAS), or a categorical scale. In some embodiments, a Wong-Baker Faces Pain Scale, a FLACC Pain Scale, a CRIES Pain Scale, COMFORT Pain Scale, a McGill Pain Questionnaire, a Color Analog Pain Scale, Mankoski Pain Scale, a Brief Pain Inventory, or a Descriptor Differential Scale of Pain Intensity, or a combination thereof is used to evaluate pain.

[0026] In certain embodiments, the neurological disorder is a sleep-wake disorder, and the ultrasound stimulation parameters are optimized to ameliorate symptoms of the sleep-wake disorder. In some embodiments, the response is measured using a visual-analog scale (VAS), a psychomotor vigilance test (PVT), a Likert scale, a Stanford Sleepiness Scale (SSS), a maintenance of wakefulness test (MWT), an Epworth sleepiness scale (ESS), a multiple sleep latency test (MSLT), or an Athens insomnia scale, or by monitoring the subject using actigraphy, electroencephalography, or polysomnography.

[0027] In certain embodiments, the target brain region is a centromedial thalamus region. In some embodiments, administering the ultrasound neuromodulation increases arousal compared to in absence of said administering the ultrasound neuromodulation.

[0028] In certain embodiments, the ultrasound neuromodulation is administered with a portable ultrasound transducer.

[0029] In certain embodiments, the ultrasound neuromodulation is administered with a wearable ultrasound array.

[0030] In certain embodiments, the ultrasound neuromodulation is administered with a transcranial focused ultrasound (FUS) transducer, phased array transducer, capacitive micromachined ultrasound transducer (CMUT), piezoelectric micromachined ultrasonic transducer (pMLIT), or miniaturized half-concave transducer.

[0031] In another aspect, a computer-implemented method for programming an ultrasound transducer to treat a subject for a neurological or neuropsychiatric disorder with ultrasound neuromodulation is provided, the computer performing steps comprising: (a) instructing the ultrasound transducer to deliver ultrasound neuromodulation to the subject using an initial selected set of ultrasound stimulation parameters; (b) receiving experimental data from measuring a response of the subject to administering the ultrasound neuromodulation using the initial selected set of ultrasound stimulation parameters, wherein the measured response is used to calculate an efficacy score for treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation using the initial selected set of ultrasound stimulation parameters; (c) using the efficacy score from said administering the ultrasound neuromodulation using the initial selected set of ultrasound stimulation parameters to generate a Gaussian process regression model of an objective function to simulate responses of the subject to ultrasound neuromodulation administered using other ultrasound stimulation parameters; (d) performing Bayesian optimization with the Gaussian process regression model of the objective function using an acquisition function that selects a new set of ultrasound stimulation parameters with the goal of improving the efficacy of treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation; (e) instructing the ultrasound transducer to deliver ultrasound neuromodulation to the subject using the new set of ultrasound stimulation parameters; (f) receiving experimental data from measuring a response of the subject to administering the ultrasound neuromodulation using the new set of ultrasound stimulation parameters; (g) calculating an efficacy score for treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation using the new set of ultrasound stimulation parameters; (h) updating the Gaussian process regression model of the objective function with the efficacy score for the treatment of the neurological or neuropsychiatric disorder withthe ultrasound neuromodulation using the new set of ultrasound stimulation parameters, wherein the Gaussian process regression model of the objective function uses all the efficacy scores that have been calculated for the responses to all previous administrations of the ultrasound neuromodulation to the subject with previously selected ultrasound stimulation parameters; (i) performing Bayesian optimization with the updated Gaussian process regression model of the objective function, wherein the acquisition function selects another new set of ultrasound stimulation parameters with the goal of further improving the efficacy of the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation; (j) repeating steps (e)-(i) iteratively to further optimize the ultrasound stimulation parameters until a stopping criterion is met, wherein the acquisition function selects a final set of optimized ultrasound stimulation parameters; and (k) instructing the ultrasound transducer to deliver ultrasound neuromodulation to the subject using the final set of optimized ultrasound stimulation parameters. In certain embodiments, the stopping criterion is determined by a strategy based on probabilistic regret bounds, wherein the probabilistic regret bound statistical test comprises: defining a regret threshold; estimating probability that the regret threshold can be surpassed by repeating steps (e)-(i) after each iteration of step (j); discontinuing the repeating of steps (e)-(i) when the estimated probability that the regret threshold can be surpassed falls below a chosen confidence level. In certain embodiments, the stopping criterion is based on acquisition function convergence, wherein steps (e)-(i) are repeated until the acquisition function values converge below a pre-determined threshold. In certain embodiments, the stopping criterion is based on uncertainty reduction, wherein steps (e)-(i) are repeated until posterior variance around an estimated optimum is below a pre-determined threshold. In certain embodiments, the computer- implemented method further comprises performing cross-validation or bootstrapping, wherein the stopping criterion is based on stability of an identified optimum, wherein steps (e)-(i) are repeated until re-sampling shows stability of the identified optimum. In certain embodiments, steps (e)-(i) are repeated until convergence to an optimized target brain region and ultrasound stimulation parameters that are effective for treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation. In certain embodiments, the stopping criterion is that further Bayesian optimization of the ultrasound stimulation parameters no longer results in further improvement of the efficacy of the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation. In certain embodiments, the stopping criterion is based on a predetermined optimization budget. In certain embodiments, the acquisition function is a noisy expected improvement (qNE I) acquisition function, a max-value entropy search (MES) acquisition function, or a joint entropy search (JES) acquisition function. In certain embodiments, the computer-implemented method further comprises optimizing the parameters of the Gaussian process and of the acquisitionfunction based on a model of the noise present in the data. In some embodiments, the Gaussian noise is modeled from a normal distribution N(0, o), wherein o is equal to a selected noise fraction. In certain embodiments, the initial selected set of ultrasound stimulation parameters is selected using random sampling, Sobol sequence sampling, a pseudo-random distribution, or Voronoi parcellation.

[0032] In some embodiments, steps (e)-(i) of the computer-implemented method are repeated up to 10 times, up to 20 times, up to 30 times, up to 40 times, up to 50 times, up to 75 times, up to 100 times, up to 125 times, or up to 150 times. In some embodiments, steps (e)-(i) of the computer- implemented method are repeated 10 to 100 times, 20 to 80 times, or 30 to 40 times, including any number of times within these ranges such as 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, 50, 60, 70, 80, 90, or 100 times. In certain embodiments, the computer-implemented method further comprises repeating steps (a)-(k) at a later time to further optimize the ultrasound stimulation parameters for the subject at the later time.

[0033] In certain embodiments, the Bayesian optimization comprises optimizing hyperparameters selected from an acquisition function, seed queries, and an exploration / exploitation ratio. In certain embodiments, the Bayesian optimization comprises optimizing at least 2, at least 3, or at least 4 ultrasound stimulation parameters selected from a target brain region, a stimulation duration, an acoustic intensity, an acoustic pressure, an excitation voltage, a pulse repetition frequency, a pulse length, an ultrasound frequency, and a duty cycle. In some embodiments, the ultrasound stimulation parameters comprise the target brain region, the stimulation duration, and the excitation voltage. In some embodiments, the ultrasound stimulation parameters comprise the target brain region, the stimulation duration, the acoustic intensity, the acoustic pressure, the excitation voltage, the pulse repetition frequency, the pulse length, the ultrasound frequency, and the duty cycle.

[0034] In certain embodiments, the computer-implemented method further comprises displaying the optimized set of ultrasound stimulation parameters.

[0035] In certain embodiments, the computer-implemented method further comprises displaying a user interface presenting a questionnaire configured to receive input from the subject regarding selfreported results of the treatment with the ultrasound neuromodulation. In some embodiments, the questionnaire uses a Movement Disorder Society-Sponsored Revision of the Unified Parkinson's Disease Rating Scale (MDS-UPDRS), a Clinical Rating Scale for Tremor (CRST), a Hoehn and Yahr (HnY) scale, a Parkinson’s Disease Composite Scale (PDCS), or a Schwab and England Activities of Daily Living (ADL) Scale for assessing treatment of a movement disorder. In some embodiments, the questionnaire uses a Visual Analogue Scale for Depression (VAS-D), a Hamilton Depression Rating Scale (HAM-D), Montgomery-Asberg Depression Rating Scale (MADRS), a Beck DepressionInventories (BDI) score, a Visual Analogue Scale for Anxiety (VAS-A) or a Beck Anxiety Inventories (BAI) score for assessing treatment of depression or anxiety. In some embodiments, the questionnaire uses a numerical rating scale (NRS), a visual analog scale (VAS), or a categorical scale. In some embodiments, a Wong-Baker Faces Pain Scale, a FLACC Pain Scale, a CRIES Pain Scale, COMFORT Pain Scale, a McGill Pain Questionnaire, a Color Analog Pain Scale, Mankoski Pain Scale, a Brief Pain Inventory, or a Descriptor Differential Scale of Pain Intensity, or a combination thereof to evaluate pain. In some embodiments, the questionnaire uses a visual-analog scale (VAS), a psychomotor vigilance test (PVT), a Likert scale, a Stanford Sleepiness Scale (SSS), a maintenance of wakefulness test (MWT), an Epworth sleepiness scale (ESS), a multiple sleep latency test (MSLT), or an Athens insomnia scale to evaluate level of wakefulness, arousal, or sleepiness.

[0036] In another aspect, a non-transitory computer-readable medium is provided, the non-transitory computer-readable medium comprising program instructions that, when executed by a processor in a computer, causes the processor to perform a method described herein.

[0037] In another aspect, a kit comprising the non-transitory computer-readable medium described herein and instructions for determining symptom severity of a subject having a movement disorder is provided.

[0038] In another aspect, a system for treating a neurological or neuropsychiatric disorder with ultrasound neuromodulation is provided, the system comprising: an ultrasound transducer; and a processor programmed according to a computer-implemented method, described herein, to instruct the ultrasound transducer to deliver ultrasound neuromodulation to the subject using an optimized set of ultrasound stimulation parameters.

[0039] In certain embodiments, the system further comprises an accelerometer, a gyroscope, a video recording device, an electroencephalography (EEG) electrode, or a combination thereof. In some embodiments, the accelerometer, the gyroscope, or the combination thereof is provided by a wearable device. In some embodiments, the wearable device is a smartwatch. In some embodiments, the video recording device is provided by a digital camera, a smartphone, a tablet, a laptop, or a camcorder.

[0040] In certain embodiments, the ultrasound transducer of the system is portable. In some embodiments, the ultrasound transducer of the system is a wearable ultrasound array. In some embodiments, the ultrasound transducer of the system is a transcranial focused ultrasound (FUS) transducer, phased array transducer, capacitive micromachined ultrasound transducer (CMUT), piezoelectric micromachined ultrasonic transducer (pMUT), or miniaturized half-concave transducer.

[0041] In certain embodiments, the system further comprises a user interface comprising an input electronically coupled to the processor for instructing the ultrasound transducer to deliver ultrasound neuromodulation to the subject to treat the neurological or neuropsychiatric disorder. In some embodiments, the user interface is password protected and is operable by a health care practitioner.

[0042] In certain embodiments, the system further comprises a storage component for storing data, wherein the storage component is coupled to the processor.

[0043] In certain embodiments, the system further comprises a display for displaying the optimized set of ultrasound stimulation parameters.

[0044] In certain embodiments, the display further displays a user interface presenting a questionnaire configured to receive input from the subject regarding self-reported results of the treatment with the ultrasound neuromodulation.BRIEF DESCRIPTION OF THE DRAWINGS

[0045] FIGS. 1A-1 E. Bayesian optimization in LIFU neuromodulation. (FIGS. 1A-1C) Schematic representation of the Bayesian optimization (BO) loop. (FIG. 1 D) Synthetic ground-truth model based on our experimental mapping of CM-LIFU parameters in a rat model of arousal. The model was used to generate on-demand stimulation-response evaluations for the BO hyperparameter tuning. (FIG. 1 E) Noisy CM-LIFU models with three different noise levels, defined as fractions (0.25, 0.50, and 1 .00) of the standard deviation of the experimental CM-LIFU data distribution.

[0046] FIGS. 2A-2C. Bayesian optimization performance with different acquisition functions. (FIGS. 2A-2B) Median absolute error progression over 100 iterations (FIG. 2A) and absolute error summary statistics (median, IQR, range) at the different checkpoints (FIG. 2B) for all the BO acquisition functions in comparison to the GS and RGP methods. All the searches were repeated across 50 independent trials, each consisting of 100 iterations, in the low-, mid-, and high-noise conditions. (FIG. 2C) Number of trials (out of 50) that converged with an error lower than 1%, 5%, or 10% of the objective maximum for the best-performing acquisition functions (max-value entropy search, joint entropy search, and noisy expected improvement) in comparison to the GS and RGP methods.

[0047] FIGS. 3A-3C. Effect of the number and distribution of initialization points with the joint entropy search acquisition function. (FIG. 3A) Absolute error in the low-, mid-, and high-noise conditions at iteration n. 50 (inclusive of Ninit) as a function of the number and distribution of initialization points. The plots show the median, interquartile range, and range. (FIG. 3B) Absolute error progression in the high-noise condition. The dashed line indicates the iteration reported in (FIG. 3A). (FIG. 3C) Representative parameter walk showing the random distribution of Ninit = 15 initialization points, the parameter set evaluated at each BO iteration, and the current optimum at each iteration.

[0048] FIG. 4. Absolute error progression over 300 iterations in the low-, mid-, and high-noise conditions for the three best-performing acquisition functions (max-value entropy search, joint entropy search, and noisy expected improvement) in comparison to the GS and RGP methods.

[0049] FIG. 5. Number of trials (out of 50) that converged with an error lower than 1 %, 5%, or 10% of the objective maximum for all the acquisition functions tested in comparison to the GS and RGP methods.

[0050] FIG. 6. Absolute error with the log noisy expected improvement and max-value joint entropy search acquisition functions in the low-, mid-, and high-noise conditions at iteration n. 50 (inclusive of Ninit) as a function of the number and distribution of initialization points. The plots show the median, interquartile range, and range.

[0051] FIG. 7. Absolute error progression in the high-noise condition with the joint entropy search acquisition function. The dashed line indicates the iteration reported in FIG. 3A.DETAILED DESCRIPTION OF THE INVENTION

[0052] Devices, systems, software, and methods are provided for treating neurological and neuropsychiatric disorders with ultrasound neuromodulation. In particular, a Bayesian optimization approach is used to tune ultrasound stimulation parameters to optimize the efficacy of treatment of neurological and neuropsychiatric disorders with ultrasound neuromodulation.

[0053] Before the present devices, systems, software, and methods are described, it is to be understood that this invention is not limited to the particular devices, systems, software, and methods described, as such may, of course, vary. It is also to be understood that the terminology used herein is for the purpose of describing particular embodiments only, and is not intended to be limiting, since the scope of the present invention will be limited only by the appended claims.

[0054] Where a range of values is provided, it is understood that each intervening value, to the tenth of the unit of the lower limit unless the context clearly dictates otherwise, between the upper and lower limits of that range is also specifically disclosed. Each smaller range between any stated value or intervening value in a stated range and any other stated or intervening value in that stated range is encompassed within the invention. The upper and lower limits of these smaller ranges may independently be included or excluded in the range, and each range where either, neither or both limits are included in the smaller ranges is also encompassed within the invention, subject to any specifically excluded limit in the stated range. Where the stated range includes one or both of the limits, ranges excluding either or both of those included limits are also included in the invention.

[0055] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.Although any methods and materials similar or equivalent to those described herein can be used in the practice or testing of the present invention, some potential and preferred methods and materials are now described. All publications mentioned herein are incorporated herein by reference to disclose and describe the methods and / or materials in connection with which the publications are cited. It is understood that the present disclosure supersedes any disclosure of an incorporated publication to the extent there is a contradiction.

[0056] As will be apparent to those of skill in the art upon reading this disclosure, each of the individual embodiments described and illustrated herein has discrete components and features which may be readily separated from or combined with the features of any of the other several embodiments without departing from the scope or spirit of the present invention. Any recited method can be carried out in the order of events recited or in any other order which is logically possible.

[0057] It must be noted that as used herein and in the appended claims, the singular forms "a", "an", and "the" include plural referents unless the context clearly dictates otherwise. Thus, for example, reference to "a parameter" includes a plurality of such parameters and reference to "the response" includes reference to one or more responses, and so forth.

[0058] The publications discussed herein are provided solely for their disclosure prior to the filing date of the present application. Nothing herein is to be construed as an admission that the present invention is not entitled to antedate such publication by virtue of prior invention. Further, the dates of publication provided may be different from the actual publication dates which may need to be independently confirmed.Definitions

[0059] The term "about," particularly in reference to a given quantity, is meant to encompass deviations of plus or minus five percent.

[0060] The term “neuropsychiatric disorder” is used herein to refer to a condition that affects mood and / or behavior of a person suffering from the disorder. Neuropsychiatric disorders include major depressive disorder (MDD), generalized anxiety disorder (GAD), post-traumatic stress disorder (PTSD), addiction, anorexia, obsessive-compulsive disorder (OCD), bipolar disorder (BD) and chronic pain.

[0061] The term “affective disorders” is used herein to refer to a group of neuropsychiatric disorders that typically affect mood of the person suffering from such a disorder. Neuropsychiatric disorders that affect the mood are also referred to as mood disorders and are commonly associated with depression and / or anxiety. The main types of affective disorders are depression, bipolar disorder,and anxiety disorder. Symptoms vary by individual, but they typically affect mood. They can range from mild to severe.

[0062] As used herein the term “depression” refers to a mental state of morbid sadness, dejection, or melancholy.

[0063] As used herein the term “anxiety” refers to an uncomfortable and unjustified sense of apprehension that may be diffuse and unfocused and is often accompanied by physiological symptoms.

[0064] As used herein the term “bipolar disorder” refers to a type of affective disorder in which the person suffering from this disorder goes through periods of depression and periods of mania (feeling extremely positive and active).

[0065] As used herein the term “anxiety disorder” refers to a neuropsychiatric disorder characterized by feelings of nervousness, anxiety, and even fear. Anxiety disorders include social anxiety (anxiety caused by social situations), post-traumatic stress disorder (anxiety, fear, and flashbacks caused by a traumatic event), generalized anxiety disorder (anxiousness and fear in general, with no particular cause), panic disorder (anxiety that causes panic attacks), and obsessive-compulsive disorder (obsessive thoughts that cause anxiety and compulsive actions).

[0066] The term "neurological disorder” refers to any disorder of the nervous system. Neurological disorders include any condition or disease affecting the brain, spinal cord, or nerves including, but not limited to, apraxia, agnosia, amnesia, aphasia, dysarthria, peripheral neuropathy, trigeminal neuralgia, dysautonomia, multiple system atrophy, epilepsy, movement disorders such as Parkinson's disease, essential tremor, amyotrophic lateral sclerosis, and Tourette’s syndrome, multiple sclerosis, narcolepsy, speech disorders, headaches, including migraines, cluster headaches, and tension headaches, pain, including complex regional pain syndrome and fibromyalgia, delirium, dementia, Alzheimer's disease, coma, and stroke.

[0067] The term "movement disorder" refers to any type of neurological disorder that causes either increased movements or reduced or slow movements. Movement disorders include, but are not limited to, Parkinson's disease, parkinsonism, progressive supranuclear palsy, ataxia, cervical dystonia, chorea, dystonia, functional movement disorder, Huntington's disease, multiple system atrophy, myoclonus, tardive dyskinesia, Tourette syndrome, tremor, restless legs syndrome, and Wilson's disease. Symptoms may include, but art not limited to, tremor, involuntary movements, slowness of movement (bradykinesia), rigidity, postural instability, twisting movements, poor balance, irregularity of movements, stumbling, and difficulty with walking. In some cases, a movement disorder is caused by genetic and / or environmental factors, head trauma, infections,inflammation, metabolic disturbances, toxins, adverse reactions to medications, or stressful life events.

[0068] The term also includes movement disorders that causes abnormal facial movement, involuntary facial movement, or paralysis or partial paralysis of facial muscles. Facial movement disorders include, but are not limited to, Tardive dyskinesia, facial dystonia, hemifacial spasm, peripheral facial palsy, facial myokymia, facial bradykinesia, blepharospasm, Tourette syndrome, orofacial dyskinesia, brainstem tumor, peripheral neuropathy, chorea, multiple sclerosis, Guillain- Barre syndrome, Parkinson’s disease, congenital facial nerve palsy, acquired facial nerve palsy, Moebius syndrome, Poland syndrome, CHARGE syndrome, hemifacial microsomia, 22q1 1.2 deletion syndrome, facial paralysis or paresis, and psychogenic movement disorders. Symptoms of facial movement disorders may include, but are not limited to, rapid eye blinking, face twitching, face tremor, jaw tremor, lip tremor, lip-smacking, tongue twisting, tongue curling, tongue thrusting, lip puckering, facial drooping, or grimacing.

[0069] The term “sleep-wake disorder” is used herein to refer to any disease or condition that causes problems with poor quality of sleep, timing of sleep, and / or the amount of sleep of a subject, which may result in impaired daytime functioning. Sleep-wake disorders include, but are not limited to, hypersomnia, excessive daytime sleepiness, insomnia, sleep apnea, parasomnia, narcolepsy, restless leg syndrome, non-rapid eye movement (NREM) sleep arousal disorder, and rapid and eye movement (REM) sleep behavior disorder.

[0070] The term “hypersomnia” is used to refer herein to a sleep-wake disorder that causes excessive time spent sleeping or excessive sleepiness. Hypersomnia may be caused by idiopathic hypersomnia, narcolepsy, Kleine-Levin syndrome, Prader-Willi syndrome, Norrie disease, Niemann- Pick disease type C, myotonic dystrophy, fibromyalgia, chronic fatigue syndrome, lupus, rheumatoid arthritis, Morvan's syndrome, hypothyroidism, multiple sclerosis, encephalitis, Alzheimer's disease, Parkinson's disease, hydrocephalus, anemia, encephalitis lethargica, paraneoplastic syndrome, head trauma, or depression.

[0071] The term “insomnia” is used to refer herein to a sleep-wake disorder that causes sleeplessness, difficulty falling asleep, or difficulty staying asleep for a sufficient period of time. Insomnia may include, but is not limited to, sleep onset insomnia, middle-of-the-night awakening, early morning awakening, and / or poor sleep quality (e.g., not reaching stage 3 or delta sleep). Insomnia may be caused by psychological stress, chronic pain, heart failure, hyperthyroidism, sleep apnea, heartburn, restless leg syndrome, menopause, or amphetamines or other stimulants. In some cases, insomnia is caused by deep brain stimulation therapy (e.g., insomnia caused by DBS to ananterior limb of the internal capsule (ALIC) region such as used for treatment of obsessive- compulsive disorder or major depressive disorder).

[0072] The terms “individual”, “subject”, “recipient”, and “patient” are used interchangeably herein and refer to any mammalian subject for whom treatment or therapy is desired, particularly humans. "Mammal" for purposes of treatment refers to any animal classified as a mammal, including human and non-human mammals such as non-human primates, including chimpanzees and other apes and monkey species; laboratory animals such as mice, rats, rabbits, hamsters, guinea pigs, and chinchillas; domestic animals such as dogs and cats; and farm animals such as sheep, goats, pigs, horses and cows.

[0073] The term “user” as used herein refers to a person that interacts with a device and / or system disclosed herein for performing one or more steps of the presently disclosed methods. The user may be the patient receiving treatment for a neurological or neuropsychiatric disorder. The user may be a health care practitioner, such as the patient’s physician.

[0074] The terms "treatment", "treating", "treat" and the like are used herein to generally refer to obtaining a desired pharmacologic and / or physiologic effect. The effect can be prophylactic in terms of completely or partially preventing a disease or symptom(s) thereof and / or may be therapeutic in terms of a partial or complete stabilization or cure for a disease and / or adverse effect attributable to the disease. The term “treatment" encompasses any treatment of a disease in a mammal, particularly a human, and includes: (a) preventing the disease and / or symptom(s) from occurring in a subject who may be predisposed to the disease or symptom but has not yet been diagnosed as having it; (b) inhibiting the disease and / or symptom(s), i.e., arresting their development; or (c) relieving the disease symptom(s), i.e., causing regression of the disease and / or symptom(s). Those in need of treatment include those already inflicted (e.g., those with a neurological or neuropsychiatric disorder) as well as those in which prevention is desired those with a genetic predisposition to developing a neurological or neuropsychiatric disorder, those with increased susceptibility to developing a neurological or neuropsychiatric disorder, those suspected of having a neurological or neuropsychiatric disorder, etc.).

[0075] A therapeutic treatment is one in which the subject is inflicted prior to administration and a prophylactic treatment is one in which the subject is not inflicted prior to administration. In some embodiments, the subject has an increased likelihood of becoming inflicted or is suspected of being inflicted prior to treatment. In some embodiments, the subject is suspected of having an increased likelihood of becoming inflicted.

[0076] A "therapeutically effective dose" or “therapeutic dose” is an amount sufficient to effect desired clinical results (i.e., achieve therapeutic efficacy). A therapeutically effective dose can be administered in one or more administrations.

[0077] The term “responsive” as used herein means that the treatment is having the desired effect such as reducing symptom severity caused by a neurological or neuropsychiatric disorder. When the individual does not improve in response to the treatment, it may be desirable to seek a different therapy or treatment regime for the individual.Methods

[0078] The present disclosure provides methods for administering ultrasound neuromodulation to a subject who has a neurological or neuropsychiatric disorder. In particular, a Bayesian optimization model is used to optimize ultrasound stimulation parameters for an individual being treated for a neurological or neuropsychiatric disorder with ultrasound neuromodulation. The methods can be used in performing open-loop therapy to provide clinical guidance to clinicians or technicians for adjusting settings of an ultrasound transducer for performing ultrasound neuromodulation. Methods and systems are also provided for performing closed-loop therapy with an ultrasound transducer that automatically adjusts ultrasound stimulation settings and / or delivers ultrasound radiation to a selected target region of the brain of the subject using optimized ultrasound stimulation parameters. Various steps and aspects of the methods will now be described in greater detail below.

[0079] The method comprises: (a) administering ultrasound neuromodulation to a subject using an initial selected set of ultrasound stimulation parameters; (b) measuring a response of the subject to administering the ultrasound neuromodulation using the initial selected set of ultrasound stimulation parameters, wherein the measured response is used to calculate an efficacy score for treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation using the initial selected set of ultrasound stimulation parameters; (c) using the efficacy score from administering the ultrasound neuromodulation using the initial selected set of ultrasound stimulation parameters to generate a Gaussian process regression model of an objective function to simulate responses of the subject to ultrasound neuromodulation administered using other ultrasound stimulation parameters; (d) performing Bayesian optimization with the Gaussian process regression model of the objective function using an acquisition function that selects a new set of ultrasound stimulation parameters with the goal of improving efficacy of treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation; (e) administering ultrasound neuromodulation to the subject using the new set of ultrasound stimulation parameters; (f) measuring a response of the subject to administering the ultrasound neuromodulation using the new set of ultrasound stimulationparameters; (g) calculating an efficacy score for treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation using the new set of ultrasound stimulation parameters; (h) updating the Gaussian process regression model of the objective function with the efficacy score for the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation using the new set of ultrasound stimulation parameters, wherein the Gaussian process regression model of the objective function uses all the efficacy scores that have been calculated for the responses to all previous administrations of the ultrasound neuromodulation to the subject with previously selected ultrasound stimulation parameters; (i) performing Bayesian optimization with the updated Gaussian process regression model of the objective function, wherein the acquisition function selects another new set of ultrasound stimulation parameters with the goal of further improving the efficacy of the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation; (j) repeating steps (e)-(i) iteratively to further optimize the ultrasound stimulation parameters until a stopping criterion is met, wherein the acquisition function selects a final set of optimized ultrasound stimulation parameters; and (k) administering ultrasound neuromodulation to the subject using the final set of optimized ultrasound stimulation parameters. In certain embodiments, the initial selected set of ultrasound stimulation parameters is selected using random sampling, Sobol sequence sampling, a pseudo-random distribution, or Voronoi parcellation. In certain embodiments, performing Bayesian optimization comprises optimizing hyperparameters comprising the acquisition function, seed queries (initial ultrasound stimulation parameters tested), and an exploration / exploitation ratio.

[0080] The ultrasound stimulation parameters can be further optimized by performing further rounds of ultrasound neuromodulation with sets of optimized ultrasound stimulation parameters selected by the Bayesian optimization of parameters from a previous round. In some embodiments, steps (e)-(i) of the method are repeated up to 10 times, up to 20 times, up to 30 times, up to 40 times, up to 50 times, up to 75 times, up to 100 times, up to 125 times, or up to 150 times. In some embodiments, steps (e)-(i) of the method are repeated 10 to 100 times, 20 to 80 times, or 30 to 40 times, including any number of times within these ranges such as 10, 11 , 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, 50, 60, 70, 80, 90, or 100 times.

[0081] In some embodiments, steps (e)-(i) of the method are repeated iteratively until a stopping condition is reached. In certain embodiments, the stopping criterion is determined by a strategy based on probabilistic regret bounds, wherein the probabilistic regret bound statistical test comprises: defining a regret threshold; estimating probability that the regret threshold can be surpassed by repeating steps (e)-(i) after each iteration of step (j); discontinuing the repeating of steps (e)-(i) when the estimated probability that the regret threshold can be surpassed falls below achosen confidence level. In certain embodiments, the stopping criterion is based on acquisition function convergence, wherein steps (e)-(i) are repeated until the acquisition function values converge below a pre-determined threshold. In certain embodiments, the stopping criterion is based on uncertainty reduction, wherein steps (e)-(i) are repeated until posterior variance around an estimated optimum is below a pre-determined threshold. In certain embodiments, the method further comprises performing cross-validation or bootstrapping, wherein the stopping criterion is based on stability of an identified optimum, wherein steps (e)-(i) are repeated until re-sampling shows stability of the identified optimum. In certain embodiments, the stopping criterion is that further Bayesian optimization of the ultrasound stimulation parameters no longer results in further improvement of the efficacy of the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation. In certain embodiments, the stopping criterion is based on a predetermined optimization budget. In some embodiments, steps (e)-(i) of the method are repeated until convergence to an optimized target brain region and stimulation parameters that are effective for treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation. In some embodiments, steps (e)-(i) of the method are repeated until the Bayesian optimization of the set of ultrasound stimulation parameters no longer results in further improvement of the efficacy of the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation.

[0082] In some cases, time-dependent factors may alter what stimulation parameters are optimal over time. Optimization of the ultrasound stimulation parameters can be performed repeatedly to account for time-dependent factors, such as, but not limited to, disease progression in the subject, changes in treatment of the subject, and / or adaptation of the subject to the ultrasound neuromodulation, which may reduce therapeutic efficacy of the ultrasound neuromodulation at particular parameter settings over time. In such cases, reoptimization of the ultrasound stimulation parameters to account for changing conditions over time may benefit the subject. In certain embodiments, steps (a)-(k) are repeated at one or more timepoints to optimize the ultrasound stimulation parameters. In some embodiments, steps (a)-(k) are repeated twice a month, once a month, once every two months, once every three months, once every four months, once every five months, once every six months, or once a year.

[0083] In certain embodiments, performing Bayesian optimization comprises optimizing at least 2, at least 3, or at least 4 ultrasound stimulation parameters selected from a target brain region, a stimulation duration, an acoustic intensity, an acoustic pressure, an excitation voltage, a pulse repetition frequency, a pulse length, an ultrasound frequency, and a duty cycle. In some embodiments, the ultrasound stimulation parameters comprise the target brain region, the stimulation duration, and the excitation voltage. In some embodiments, the ultrasound stimulationparameters comprise the target brain region, the stimulation duration, the acoustic intensity, an acoustic pressure, the excitation voltage, the pulse repetition frequency, the pulse length, the ultrasound frequency, and the duty cycle.

[0084] In certain embodiments, the method further comprises optimizing the parameters of the Gaussian process and of the acquisition function based on a model of the noise present in the data. In some embodiments, the Gaussian noise is modeled from a normal distribution N(0, a), wherein o is equal to a selected noise fraction.

[0085] In certain embodiments, the response measured to provide an indication of the efficacy of the ultrasound neuromodulation for treating a neurological or neuropsychiatric disorder is a body movement, a facial movement, a change in locomotor activity, a change in mood, or a change in brain electrical activity, or a combination thereof. A body movement, facial movement, or change in locomotor activity can be measured, for example using an accelerometer, a gyroscope, or a video recording device that records images of the subject. An accelerometer and / or gyroscope can be provided by a wearable device such as a wrist-worn smartwatch. The video recording device may be any commercially available, off the shelf, video recording device. In certain embodiments, the video recording device is provided by a commercially available digital camera, smartphone, tablet, laptop, or camcorder. The video recording device is preferably mounted on a tripod to avoid any movement of the video recording device during video recordings of the subject. In some embodiments, movements of the subject are measured by monitoring the subject using multi-view video recordings of the subject.

[0086] Detection of brain electrical activity may be performed by any method known in the art. For example, functional brain imaging of neural activity may be carried out by electrical methods such as electroencephalography (EEG), stereoelectroencephalography (sEEG), electrocorticography (ECoG), magnetoencephalography (MEG), single photon emission computed tomography (SPECT), as well as metabolic and blood flow studies such as functional magnetic resonance imaging (fMRI), and positron emission tomography (PET).

[0087] In certain embodiments, the subject has a movement disorder, and the ultrasound stimulation parameters are optimized to ameliorate a symptom of the movement disorder such as, but not limited to, bradykinesia, dyskinesia, or an abnormal or involuntary facial movement. In some embodiments, efficacy of the treatment with ultrasound neuromodulation is evaluated using a Movement Disorder Society-Sponsored Revision of the Unified Parkinson's Disease Rating Scale (MDS-UPDRS) or a Hoehn and Yahr (HnY) scale, a Clinical Rating Scale for Tremor (CRST), a Parkinson’s Disease Composite Scale (PDCS), or a Schwab and England Activities of Daily Living (ADL) Scale. In some embodiments, the subject is monitored for motor symptoms using a video recording device or awearable monitor that can acquire accelerometry, gyroscope, and / or surface electromyographic (sEMG) data to monitor movement and evaluate the effectiveness of the treatment. A wrist-watch style wearable monitor such as the Parkinson’s KinetiGraph®, PKG® is available from PKG Health (San Francisco, CA), which can monitor movement continuously and detect various motor symptoms of a movement disorder such as dyskinesia, bradykinesia, tremor, daytime immobility, stiffness, slow movements, gait / walking, daytime somnolence, and sleep fragmentation. An Apple Watch is available from Apple Inc. (Cupertino, CA), which can be used, for example, for monitoring, bradykinesia, dyskinesia, tremors, gait, and arm movement.

[0088] In certain embodiments, the subject has depression or anxiety, and the ultrasound stimulation parameters are optimized to ameliorate the depression or the anxiety. In some embodiments, the change in mood is measured using a Visual Analogue Scale for Depression (VAS-D), a Hamilton Depression Rating Scale (HAM-D), Montgomery-Asberg Depression Rating Scale (MADRS), a Beck Depression Inventories (BDI) score, a Visual Analogue Scale for Anxiety (VAS-A) or a Beck Anxiety Inventories (BAI) score. In some embodiments, the change in mood is measured based on facial expression detection or facial emotion recognition analysis of a video recording of the subject’s face.

[0089] In certain embodiments, the subject has neuropathic pain, and the ultrasound stimulation parameters are optimized to ameliorate the neuropathic pain. In some embodiments, the response and efficacy of the treatment with ultrasound neuromodulation is evaluated using a numerical rating scale (NRS), a visual analog scale (VAS), or a categorical scale. In some embodiments, a Wong- Baker Faces Pain Scale, a FLACC Pain Scale, a CRIES Pain Scale, COMFORT Pain Scale, a McGill Pain Questionnaire, a Color Analog Pain Scale, Mankoski Pain Scale, a Brief Pain Inventory, or a Descriptor Differential Scale of Pain Intensity, or a combination thereof is used to evaluate pain.

[0090] In certain embodiments, the subject has a sleep-wake disorder, and the ultrasound stimulation parameters are optimized to ameliorate symptoms of the sleep-wake disorder. In some embodiments, the response and efficacy of the treatment with ultrasound neuromodulation is evaluated using a visual-analog scale (VAS), a psychomotor vigilance test (PVT), a Likert scale, a Stanford Sleepiness Scale (SSS), a maintenance of wakefulness test (MWT), an Epworth sleepiness scale (ESS), a multiple sleep latency test (MSLT), or an Athens insomnia scale. Patients, who are treated with ultrasound neuromodulation, may be asked to self-rate their level of energy (e.g., from “not at all energetic” to “very energetic”) or their level of wakefulness / sleepiness (e.g., feeling active, vital, alert, or wide awake, or no longer fighting sleep, sleep onset soon, or having dream-like thoughts). Alternatively or additionally, a bedside observer may rate arousal metrics, e.g., using a 5- point Likert scale ranging from -2 (i.e., episodes of eye closure, decreased responsiveness to the environment, slowed speech rate) to 2 (i.e., high level of engagement with the environment,completion of tasks, relatively rapid speech rate). In some embodiments, assessing effectiveness of the treatment of a sleep-wake disorder in the subject comprises monitoring the subject using actigraphy, electroencephalography, or polysomnography.

[0091] Ultrasound neuromodulation may be applied to one or more target brain regions. The optimal brain region will depend on the particular neurological or neuropsychiatric disorder that is treated and can be optimized by Bayesian optimization using a Gaussian process regression model, as described herein. In certain embodiments, the method comprises mapping the brain of the subject to optimize the target brain region where ultrasound neuromodulation is applied. The target brain region is optimized to maximize clinical responses to ultrasound neuromodulation to treat a symptom of a neurological or neuropsychiatric disorder. In some embodiments, the centromedial thalamus region, ventralis intermedius nucleus region, subthalamic nucleus region, globus pallidus internus region, subgenual cingulate region, subgenual anterior cingulate cortex region, nucleus accumbens, orbitofrontal cortex region, ventral capsule region, ventral striatum region, prefrontal cortex region, visual cortex region, motor cortex region, somatosensory cortex region, amygdala region, caudate nuclei regions, hypothalamus region, hippocampus region, midbrain region, right anterior cortex region, or other regions of the brain are mapped to determine an optimal brain target region to apply ultrasound neuromodulation.

[0092] Closed-loop therapy can be performed with an ultrasound transducer used in combination with measuring responses of the subject to the ultrasound neuromodulation and automatically optimizing the ultrasound stimulation parameters with Bayesian optimization using a Gaussian process regression model, as described herein. The initial parameters for applying the ultrasound neuromodulation to the brain may be determined empirically during treatment or may be pre-defined, such as, from a trial study with a subject. The ultrasound stimulation parameters that are optimized may include one or more of the target brain region(s), stimulation duration, acoustic intensity, acoustic pressure, excitation voltage, pulse repetition frequency, pulse length, ultrasound frequency, duty cycle, and the like.

[0093] In some embodiments, the ultrasound is applied in pulses. The pulse repetition frequency of the ultrasound neuromodulation is the rate at which an ultrasound transducer emits pulses. In certain embodiments, the method comprises applying ultrasound neuromodulation to the brain with a pulse repetition frequency in a range of 10 Hz to 3,000 Hz, 300 Hz to 1 ,500 Hz, 300 Hz to 3 kHz, or 10 Hz to 100 Hz, or any pulse repetition frequency within these ranges such as 10 Hz, 20 Hz, 30 Hz, 40 Hz, 50 Hz, 60 Hz, 70 Hz, 80 Hz, 90 Hz, 100 Hz, 120 Hz, 140 Hz, 160 Hz, 180 Hz, 200 Hz, 250 Hz, 300 Hz, 350 Hz, 400 Hz, 450 Hz, 500 Hz, 600 Hz, 700 Hz, 800 Hz, 900 Hz, 1 kHz, 2 kHz, 3 kHz, 4 kHz, 5 kHz, 6 kHz, 7 kHz, 8 kHz, 9 kHz, 10 kHz, 20 kHz, 30 kHz, 40 kHz, 50 kHz, 60 kHz, 70 kHz,80 kHz, 90 kHz, 100 kHz, 200 kHz, 300 kHz, 400 kHz, 500 kHz, 600 kHz, 700 kHz, 800 kHz, or 900 kHz. In some embodiments, non-integer pulse repetition frequencies are used (e.g. 100.2 Hz, 1500.5 Hz, etc.).

[0094] In certain embodiments, ultrasound neuromodulation is administered with an ultrasound frequency in a range of 20 kHz to 5.0 MHz, 5 MHz to 20 MHz, 200 kHz to 400 kHz, 0.7 MHz to 3.0 MHz, or 1.0 MHz to 2.9 MHz, including any ultrasound frequency within these ranges, such as 20 kHz, 30 kHz, 40 kHz, 50 kHz, 60 kHz, 70 kHz, 80 kHz, 90 kHz, 100 kHz, 110 kHz, 120 kHz, 130 kHz, 140 kHz, 150 kHz, 160 kHz, 170 kHz, 180 kHz, 190 kHz, 200 kHz, 210 kHz, 220 kHz, 230 kHz, 240 kHz, 250 kHz, 260 kHz, 270 kHz, 280 kHz, 290 kHz, 300 kHz, 320 kHz, 340 kHz, 360 kHz, 380 kHz, 400 kHz, 420 kHz, 440 kHz, 460 kHz, 480 kHz, 500 kHz, 1 .0 MHz, 1.1 MHz, 1 .2 MHz, 1 .3 MHz, 1 .4 MHz, 1.5 MHz, 1.6 MHz, 1.7 MHz, 1.8 MHz, 1.9 MHz, 2.0 MHz, 2.1 MHz, 2.2 MHz, 2.3 MHz, 2.4 MHz, 2.5 MHz, 2.6 MHz, 2.7 MHz, 2.8 MHz, 3.0 MHz, 3.2 MHz, 3.4 MHz, 3.6 MHz, 3.8 MHz, 4.0 MHz, 4.2 MHz, 4.4 MHz, 4.6 MHz, 4.8 MHz, 5.0 MHz, 6 MHz, 7 MHz, 8 MHz, 9 MHz, 10 MHz, 11 MHz, 12 MHz, 13 MHz, 14 MHz, 15 MHz, 16 MHz, 17 MHz, 18 MHz, 19 MHz, or 20 MHz.

[0095] In certain embodiments, the method comprises applying ultrasound neuromodulation to the brain with a stimulation duration in a range of 0.1 ms to 500 ms, 200 ms to 500 ms, 0.3 ms to 5 ms, or 0.4 ms to 2 ms, or any stimulation duration within these ranges such as 0.1 ms, 0.2 ms, 0.3 ms, 0.4 ms, 0.5 ms, 0.6 ms, 0.7 ms, 0.8 ms, 0.9 ms, 1 .0 ms, 1 .5 ms, 2.0 ms, 2.5 ms, 3.0 ms, 3.5 ms, 4.0, 4.5 ms, or 5 ms.

[0096] In certain embodiments, the method comprises applying ultrasound neuromodulation to the brain with a duty cycle in a range from 0.01 % to 100%, 1% to 20%, 10% to 80%, or 50% to 70%, including any ultrasound duty cycle within these ranges such as 0.01%, 0.1%, 1%, 2%, 3%, 4%, 5%, 6%, 7%, 8%, 9%, 10%, 11%, 12%, 13%, 14%, 15%, 16%, 17%, 18%, 19%, 20%, 25%, 30%, 35%, 40%, 45%, 50%, 55%, 60%, 65%, 70%, 75%, 80%, 90%, 95%, or 100%.

[0097] Acoustic intensity is the average rate of energy flow through a given area. The ultrasound neuromodulation is generally applied at an acoustic intensity below a level that generates heat. In certain embodiments, the method comprises applying ultrasound neuromodulation to the brain with an acoustic intensity less than 190 W / cm2. In certain embodiments, the method comprises applying ultrasound neuromodulation to the brain with an acoustic intensity in a range from 0.04 W / cm2to 5.0 W / cm2, 0.04 W / cm2to 0.6 W / cm2, 0.05 W / cm2to 0.5 W / cm2, or 0.3 W / cm2to 4.9 W / cm2, including any acoustic intensity within these ranges such as 0.04 W / cm2, 0.05 W / cm2, 0.06 W / cm2, 0.07 W / cm2, 0.08 W / cm2, 0.09 W / cm2, 0.1 W / cm2, 0.2 W / cm2, 0.3 W / cm2, 0.4 W / cm2, 0.5 W / cm2, 0.6 W / cm2, 0.7 W / cm2, 0.8 W / cm2, 0.9 W / cm2, 1 W / cm2, 1 .5 W / cm2, 2.0 W / cm2, 2.5 W / cm2, 3.0 W / cm2, 3.5 W / cm2, 4.0 W / cm2, 4.5 W / cm2, or 5.0 W / cm2.

[0098] In certain embodiments, the method comprises applying ultrasound neuromodulation to the brain with an acoustic pressure in a range from 100 kPa to 2 MPa, 0.2 MPa to 1 .8 MPa, 0.05 MPa to 0.15 MPa, or 0.5 MPa to 1.0 MPa, including any acoustic pressure within these ranges such as 100 kPa, 200 kPa, 300 kPa, 400 kPa, 500 kPa, 600 kPa, 700 kPa, 800 kPa, 900 kPa, 0.05 MPa, 0.06 MPa, 0.07 MPa, 0.08 MPa, 0.09 MPa, 0.1 MPa, 0.2 MPa, 0.3 MPa, 0.4 MPa, 0.5 MPa, 0.6 MPa, 0.7 MPa, 0.8 MPa, 0.9 MPa, 1 .0 MPa, 1.1 MPa, 1 .2 MPa, 1 .3 MPa, 1 .4 MPa, 1 .5 MPa, 1 .6 MPa, 1 .7 MPa, 1 .8 MPa, 1 .9 MPa, or 2.0 MPa.

[0099] In certain embodiments, the method comprises applying ultrasound neuromodulation to the brain with an excitation voltage in a range from 1 volt to 100 volts, 2 volts to 12 volts, 20 volts to 80 volts, or 30 volts to 90 volts, including any excitation voltage within these ranges such as 1 , 2, 3, 4, 5, 6, 7, 8, 9, 10 volts, 20 volts, 30 volts, 40 volts, 50 volts, 60 volts, 70 volts, 80 volts, 90 volts, or 100 volts.

[0100] The ultrasound neuromodulation may be applied for a stimulation period of 0.1 sec-1 month, with periods of rest (i.e., no ultrasound neuromodulation) possible in between. In certain cases, the period of ultrasound neuromodulation may be 0.1 sec-1 week, 1 sec-1 day, 10 sec-12 hours, 1 min- 6 hours, 10 min-1 hour, and so forth. In certain cases, the period of ultrasound neuromodulation may be 1 sec-1 min, 1 sec-30 sec, 1 sec-15 sec, 1 sec-10 sec, 1 sec-6 sec, 1 sec-3 sec, 1 sec-2 sec, or 6 sec- 10 sec. The period of rest in between each stimulation period may be 60 sec or less, 30 sec or less, 20 sec or less, or 10 sec. In some embodiments, ultrasound neuromodulation may be applied for a year or more, 2 years or more, 3 years or more, 5 years or more, or 10 years or more. In some embodiments, ultrasound neuromodulation may be continued indefinitely as part of a long-term ultrasound neuromodulation therapy regimen.

[0101] The ultrasound neuromodulation having the parameters as set forth above may be applied over a program duration of around 1 day or less, such as, 18 hours, 6 hours, 3 hours, 2 hours, 1 hour, 45 minutes, 30 minutes, 20 minutes, 10 minutes, or 5 minutes, or less, e.g., 1 minute - 5 minutes, 2 minutes - 10 minutes, 2 minutes - 20 minutes, 2 minutes - 30 minutes, 5 minutes - 10 minutes, 5 minutes - 30 minutes, or 5 minutes - 15 minutes, 10 minutes - 400 minutes, 25 minutes - 300 minutes, 50 minutes - 200 minutes, or 75 minutes - 150 minutes, which period would include the application of pulses and the intervening rest period. The program may be repeated at a desired program frequency to treat a neurological or neuropsychiatric disorder in a subject. As such, a treatment regimen may include a program for performing ultrasound neuromodulation at a desired program frequency and program duration. In some embodiments, the treatment regimen is controlled by a control unit in communication with an ultrasound transducer and a device for monitoring theresponse of the subject (e.g., accelerometer, gyroscope, video recording device, or EEG electrodes) in a closed-loop treatment regimen.

[0102] As noted above, the treatment may ameliorate a symptom of a neurological or neuropsychiatric disorder. Assessment of effectiveness of the treatment may be performed using any known method for evaluating symptoms. The method selected will depend on the particular neurological or neuropsychiatric disorder undergoing treatment with ultrasound neuromodulation.

[0103] In some embodiments, efficacy of the treatment of a movement disorder is evaluated using a Movement Disorder Society-Sponsored Revision of the Unified Parkinson's Disease Rating Scale (MDS-UPDRS) or a Hoehn and Yahr (HnY) scale, a Clinical Rating Scale for Tremor (CRST), a Parkinson’s Disease Composite Scale (PDCS), or a Schwab and England Activities of Daily Living (ADL) Scale. In some embodiments, the subject is monitored for motor symptoms using a wearable monitor that can acquire accelerometry, gyroscope, and / or surface electromyographic (sEMG) data to evaluate the effectiveness of the treatment. A wrist-watch style wearable monitor such as the Parkinson’s KinetiGraph®, PKG® is available from PKG Health (San Francisco, CA), which can monitor movement continuously and detect various motor symptoms of a movement disorder such as dyskinesia, bradykinesia, tremor, daytime immobility, stiffness, slow movements, gait / walking, daytime somnolence, and sleep fragmentation. An Apple Watch is available from Apple Inc. (Cupertino, CA), which can be used, for example, for monitoring, bradykinesia, dyskinesia, tremors, gait, and arm movement.

[0104] In some embodiments, efficacy of the treatment of depression or the anxiety is evaluated using a Visual Analogue Scale for Depression (VAS-D), a Hamilton Depression Rating Scale (HAM- D), Montgomery-Asberg Depression Rating Scale (MADRS), a Beck Depression Inventories (BDI) score, a Visual Analogue Scale for Anxiety (VAS-A) or a Beck Anxiety Inventories (BAI) score. In some embodiments, the change in mood is measured based on facial expression detection or facial emotion recognition analysis of a video recording of the subject’s face.

[0105] In some embodiments, efficacy of the treatment of neuropathic pain is evaluated using a numerical rating scale (NRS), a visual analog scale (VAS), or a categorical scale. In some embodiments, a Wong-Baker Faces Pain Scale, a FLAGG Pain Scale, a CRIES Pain Scale, COMFORT Pain Scale, a McGill Pain Questionnaire, a Color Analog Pain Scale, Mankoski Pain Scale, a Brief Pain Inventory, or a Descriptor Differential Scale of Pain Intensity, or a combination thereof is used to evaluate pain.

[0106] In some embodiments, assessing effectiveness of treatment of a sleep-wake disorder in the subject comprises using a visual-analog scale (VAS), a psychomotor vigilance test (PVT), a Likert scale, a Stanford Sleepiness Scale (SSS), a maintenance of wakefulness test (MWT), an Epworthsleepiness scale (ESS), a multiple sleep latency test (MSLT), or an Athens insomnia scale. Patients, who are treated with ultrasound neuromodulation, may be asked to self-rate their level of energy (e.g., from “not at all energetic” to “very energetic”) or their level of wakefulness / sleepiness (e.g., feeling active, vital, alert, or wide awake, or no longer fighting sleep, sleep onset soon, or having dream-like thoughts). Alternatively or additionally, a bedside observer may rate arousal metrics, e.g., using a 5-point Likert scale ranging from -2 (i.e., episodes of eye closure, decreased responsiveness to the environment, slowed speech rate) to 2 (i.e., high level of engagement with the environment, completion of tasks, relatively rapid speech rate). In some embodiments, assessing effectiveness of the treatment of a sleep-wake disorder in the subject comprises monitoring the subject using actigraphy, electroencephalography, or polysomnography.

[0107] In certain cases, effectiveness of treatment may be assessed by detecting brain electrical activity. Detection of brain activity may be performed by functional brain imaging. Functional brain imaging may be carried out by electrical methods such as electroencephalography (EEG), chronic subgaleal recordings, burrhole or cranially mounted neurostimulator electrode recording, electrocorticography (ECoG), magnetoencephalography (MEG), single photon emission computed tomography (SPECT), as well as metabolic and blood flow studies such as functional magnetic resonance imaging (fMRI), functional ultrasound imaging (fUSI), and positron emission tomography (PET). In some embodiments, electrical methods for assessing effectiveness of treatment may involve use of a neural recording electrode for measuring electrical signals, which may be placed at a region of the brain or in the skull, or extracranially. One or more regions of the brain may be implanted with an electrode and electrical signals measured for assessment of effectiveness of the treatment. Any suitable electrodes may be used for measurements and may include one or more surface electrodes (non-brain penetrating electrode(s)) or one or more depth electrodes (brain penetrating electrode(s)).

[0108] Assessment of effectiveness of treatment may be performed at any suitable time point after commencement of the ultrasound neuromodulation, for example, during open-loop or closed-loop therapy. Embodiments of the subject methods include assessing effectiveness of ultrasound neuromodulation in treating a neurological or neuropsychiatric disorder in a subject within seconds, minutes, hours, or days after the initial treatment regimen has been completed. In some instances, assessment may be performed at multiple time points. In some cases, more than one type of assessment may be performed at the different time points. In some embodiments, a symptom of a neurological or neuropsychiatric disorder is assessed prior to and after the application of ultrasound neuromodulation, wherein reduced severity of the symptom indicates successful treatment.

[0109] Upon completion of a treatment regimen, the patient may be assessed for effectiveness of the treatment and the treatment regimen may be repeated, if needed. In certain cases, the treatment regimen may be altered before repeating. For example, one or more of the target brain region, stimulation duration, acoustic intensity, acoustic pressure, excitation voltage, pulse repetition frequency, pulse length, ultrasound frequency, and duty cycle may be altered before starting a second treatment regimen.

[0110] Application of the method may include a prior step of selecting a patient for treatment based on need as determined by clinical assessment, which may include assessment of severity of symptoms of a neurological or neuropsychiatric disorder, physical condition, cognitive assessment, anatomical assessment, behavioral assessment and / or neurophysiological assessment. In certain cases, a subject may be further assessed to determine if ultrasound neuromodulation will completely or partially (e.g., at least 50%) relieve symptoms. Such a patient may undergo ultrasound neuromodulation on a temporary trial basis to determine if ultrasound neuromodulation decreases the severity of symptoms experienced by the patient. Such a patient may also be administered ultrasound neuromodulation with a series of test ultrasound stimulation parameters to identify a personalized set of optimized ultrasound stimulation parameters to determine therapeutic stimulation parameters for the patient and / or evaluate whether ultrasound neuromodulation will be effective for treating the neurological or neuropsychiatric disorder in the patient. In some embodiments, the methods of the present disclosure include measurement of a patient’s response to ultrasound neuromodulation administered with optimized ultrasound stimulation parameters selected based on Bayesian optimization using a Gaussian process regression model, as described herein.

[0111] A closed-loop method allows selection of optimized ultrasound stimulation parameters for ultrasound neuromodulation based upon real-time feedback signals from measuring the response of the subject to ultrasound neuromodulation. Closed-loop methods and systems allow for automation of treatment of the subject including real-time need-based modulation of the treatment regimen. Exemplary closed-loop methods and associated systems for treatment of symptoms of a neurological or neuropsychiatric disorder are further discussed in the Examples section. Closed-loop methods and systems for automated delivery of ultrasound neuromodulation with optimized ultrasound stimulation parameters are further described below.Closed-Loop Method for Automated Delivery of Ultrasound Neuromodulation

[0112] In certain embodiments, a control algorithm is used to automate the delivery of ultrasound neuromodulation to the brain for treatment of a neurological or neuropsychiatric disorder. According to certain embodiments, the method may include optimizing one or more programmed stimulationparameters according to an algorithm’s control law based on experimental data from measuring a response of the subject to treatment with the ultrasound neuromodulation; and delivering ultrasound neuromodulation to the brain with the optimized stimulation parameters in a manner effective to mitigate symptoms of a neurological or neuropsychiatric disorder in the subject.

[0113] As described in the foregoing sections, effectiveness of treatment of a symptom of a neurological or neuropsychiatric disorder may be assessed using an accelerometer, gyroscope, video recording device, or EEG electrodes. In an open-loop system, stimulation is delivered in a preprogrammed way or manually by a user but is not automatically controlled by real-time feedback from experimental data of a measured response of the subject to treatment with the ultrasound neuromodulation. The experimental data from measuring a response of the subject to treatment with the ultrasound neuromodulation may be analyzed by a computing means which may output recommendations for updating the ultrasound stimulation parameters. A user may then carry out the recommendations, such as changing a parameter of the ultrasound neuromodulation program prior to starting another treatment regimen. In a closed-loop system, by contrast, a computing means can automatically update ultrasound stimulation parameters based upon analysis of the experimental data from measuring a response of the subject to treatment with the ultrasound neuromodulation at particular stimulation settings and / or automatically deliver ultrasound neuromodulation to the brain according to the ultrasound neuromodulation program. In some embodiments, either an open-loop or a closed-loop system may be integrated with a mechanism for user intervention, for example by allowing user-override of open-loop or closed-loop stimulation programs to enact or prevent stimulation that would ordinarily occur, or to manually change parameters of such stimulation.

[0114] In some embodiments, the computing means for directing closed-loop stimulation may be a combination of hardware / software which may be connected wirelessly or by wire to the ultrasound transducer. The computing means may communicate with a control unit (also referred to as a control module) that controls the ultrasound transducer. In certain embodiments, the computing means may be connected to a device such as an accelerometer, gyroscope, video recording device, or EEG electrodes that measure the response to the ultrasound neuromodulation. The computing means may include a control algorithm that determines modification of ultrasound stimulation parameters based on real-time outputs of the device measuring the response to the ultrasound neuromodulation.

[0115] In some embodiments, a computer-implemented method for programming an ultrasound transducer to treat a subject for a neurological or neuropsychiatric disorder with ultrasound neuromodulation is provided, the computer performing steps comprising: (a) instructing the ultrasound transducer to deliver ultrasound neuromodulation to the subject using an initial selected set of ultrasound stimulation parameters; (b) receiving experimental data from measuring a responseof the subject to administering the ultrasound neuromodulation using the initial selected set of ultrasound stimulation parameters, wherein the measured response is used to calculate an efficacy score for treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation using the initial selected set of ultrasound stimulation parameters; (c) using the efficacy score from said administering the ultrasound neuromodulation using the initial selected set of ultrasound stimulation parameters to generate a Gaussian process regression model of an objective function to simulate responses of the subject to ultrasound neuromodulation administered using other ultrasound stimulation parameters; (d) performing Bayesian optimization with the Gaussian process regression model of the objective function using an acquisition function that selects a new set of ultrasound stimulation parameters with the goal of improving the efficacy of treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation; (e) instructing the ultrasound transducer to deliver ultrasound neuromodulation to the subject using the new set of ultrasound stimulation parameters; (f) receiving experimental data from measuring a response of the subject to administering the ultrasound neuromodulation using the new set of ultrasound stimulation parameters; (g) calculating an efficacy score for treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation using the new set of ultrasound stimulation parameters; (h) updating the Gaussian process regression model of the objective function with the efficacy score for the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation using the new set of ultrasound stimulation parameters, wherein the Gaussian process regression model of the objective function uses all the efficacy scores that have been calculated for the responses to all previous administrations of the ultrasound neuromodulation to the subject with previously selected ultrasound stimulation parameters; (i) performing Bayesian optimization with the updated Gaussian process regression model of the objective function, wherein the acquisition function selects another new set of ultrasound stimulation parameters with the goal of further improving the efficacy of the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation; (j) repeating steps (e)-(i) iteratively to further optimize the ultrasound stimulation parameters until a stopping criterion is met, wherein the acquisition function selects a final set of optimized ultrasound stimulation parameters; and (k) instructing the ultrasound transducer to deliver ultrasound neuromodulation to the subject using the final set of optimized ultrasound stimulation parameters. In certain embodiments, the initial selected set of ultrasound stimulation parameters is selected using random sampling, Sobol sequence sampling, a pseudorandom distribution, or Voronoi parcellation. In certain embodiments, a model of the objective function estimated from a group of previously treated subjects with a given neurological or neuropsychiatric disorder is used to initialize the Gaussian process regression.

[0116] In certain embodiments, the acquisition function is a noisy expected improvement (qNEI) acquisition function, a max-value entropy search (MES) acquisition function, or ajoint entropy search (JES) acquisition function. In certain embodiments, the method further comprises optimizing the parameters of the Gaussian process and of the acquisition function based on a model of the noise present in the data. In some embodiments, the Gaussian noise is modeled from a normal distribution N(0, a), wherein o is equal to a selected noise fraction.

[0117] In certain embodiments, the stopping criterion is determined by a strategy based on probabilistic regret bounds, wherein the probabilistic regret bound statistical test comprises: defining a regret threshold; estimating probability that the regret threshold can be surpassed by repeating steps (e)-(i) after each iteration of step (j); discontinuing the repeating of steps (e)-(i) when the estimated probability that the regret threshold can be surpassed falls below a chosen confidence level. In certain embodiments, the stopping criterion is based on acquisition function convergence, wherein steps (e)-(i) are repeated until the acquisition function values converge below a predetermined threshold. In certain embodiments, the stopping criterion is based on uncertainty reduction, wherein steps (e)-(i) are repeated until posterior variance around an estimated optimum is below a pre-determined threshold. In certain embodiments, the computer-implemented method further comprises performing cross-validation or bootstrapping, wherein the stopping criterion is based on stability of an identified optimum, wherein steps (e)-(i) are repeated until re-sampling shows stability of the identified optimum. In certain embodiments, steps (e)-(i) are repeated until convergence to an optimized target brain region and ultrasound stimulation parameters that are effective for treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation. In certain embodiments, the stopping criterion is that further Bayesian optimization of the ultrasound stimulation parameters no longer results in further improvement of the efficacy of the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation. In certain embodiments, the stopping criterion is based on a predetermined optimization budget.

[0118] In some embodiments, steps (e)-(i) of the computer-implemented method are repeated up to 10 times, up to 20 times, up to 30 times, up to 40 times, up to 50 times, up to 75 times, up to 100 times, up to 125 times, or up to 150 times. In some embodiments, steps (e)-(i) of the computer- implemented method are repeated 10 to 100 times, 20 to 80 times, or 30 to 40 times, including any number of times within these ranges such as 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, 50, 60, 70, 80, 90, or 100 times.

[0119] In certain embodiments, the computer-implemented method further comprises repeating steps (a)-(k) at one or more later times to reoptimize the ultrasound stimulation parameters for the subject (e.g., to account for time-dependent changes in what stimulation parameters are optimal).

[0120] In certain embodiments, the Bayesian optimization comprises optimizing hyperparameters selected from an acquisition function, seed queries, and an exploration / exploitation ratio. In certain embodiments, the Bayesian optimization comprises optimizing at least 2, at least 3, or at least 4 ultrasound stimulation parameters selected from a target brain region, a stimulation duration, an acoustic intensity, an acoustic pressure, an excitation voltage, a pulse repetition frequency, a pulse length, an ultrasound frequency, and a duty cycle. In some embodiments, the ultrasound stimulation parameters comprise the target brain region, the stimulation duration, and the excitation voltage. In some embodiments, the ultrasound stimulation parameters comprise the target brain region, the stimulation duration, the acoustic intensity, the acoustic pressure, the excitation voltage, the pulse repetition frequency, the pulse length, the ultrasound frequency, and the duty cycle.

[0121] In certain embodiments, the computer-implemented method further comprises storing a user profile for the subject comprising information regarding the experimental data from measuring the response(s) of the subject to treatment with the ultrasound neuromodulation. In certain embodiments, the computer-implemented method further comprises storing a user profile for the subject comprising information regarding the optimized ultrasound stimulation parameters for treating the subject based on the experimental data.

[0122] In certain embodiments, the computer implemented method further comprises displaying a user interface presenting a questionnaire configured to receive input from the subject regarding selfreported results of the treatment with the ultrasound neuromodulation. In certain embodiments, the questionnaire uses a Movement Disorder Society-Sponsored Revision of the Unified Parkinson's Disease Rating Scale (MDS-UPDRS), a a Clinical Rating Scale for Tremor (CRST), and Yahr (HnY) scale, a Parkinson’s Disease Composite Scale (PDCS), or a Schwab and England Activities of Daily Living (ADL) Scale for assessing treatment of a movement disorder. In certain embodiments, the questionnaire uses a Visual Analogue Scale for Depression (VAS-D), a Hamilton Depression Rating Scale (HAM-D), Montgomery-Asberg Depression Rating Scale (MADRS), a Beck Depression Inventories ( BD I) score, a Visual Analogue Scale for Anxiety (VAS-A) or a Beck Anxiety Inventories (BAI) score to assess treatment of depression or anxiety. In certain embodiments, the questionnaire uses a numerical rating scale (NRS), a visual analog scale (VAS), or a categorical scale. In some embodiments, a Wong-Baker Faces Pain Scale, a FLACC Pain Scale, a CRIES Pain Scale, COMFORT Pain Scale, a McGill Pain Questionnaire, a Color Analog Pain Scale, Mankoski Pain Scale, a Brief Pain Inventory, or a Descriptor Differential Scale of Pain Intensity, or a combination thereof to evaluate pain. In certain embodiments, the questionnaire uses a visual-analog scale (VAS), a psychomotor vigilance test (PVT), a Likert scale, a Stanford Sleepiness Scale (SSS), a maintenance of wakefulness test (MWT), an Epworth sleepiness scale (ESS), a multiple sleeplatency test (MSLT), or an Athens insomnia scale to evaluate level of wakefulness / arousal or sleepiness.

[0123] In certain embodiments, the computer implemented method further comprises displaying the optimized stimulation parameters for treating a symptom of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation.

[0124] The methods described herein can be implemented in digital electronic circuitry, or in computer software, firmware, or hardware. The disclosed and other embodiments can be implemented as one or more computer program products, i.e., one or more modules of computer program instructions encoded on a computer readable medium for execution by, or to control the operation of, a data processing apparatus. The computer readable medium can be a machine- readable storage device, a machine-readable storage substrate, a memory device, a composition of matter effecting a machine-readable propagated signal, or any combination thereof.

[0125] A computer program (also known as a program, software, software application, script, or code) can be written in any form of programming language, including compiled or interpreted languages, and it can be deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment. A computer program does not necessarily correspond to a file in a file system. A program can be stored in a portion of a file that holds other programs or data (e.g., one or more scripts stored in a markup language document), in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files that store one or more modules, sub programs, or portions of code). A computer program can be deployed to be executed on one computer or on multiple computers that are located at one site or distributed across multiple sites and interconnected by a communication network.

[0126] In a further aspect, the system for performing the computer-implemented method, as described, may include a computer containing a processor, a storage component (i.e., memory), a display component, and other components typically present in general purpose computers. The storage component stores information accessible by the processor, including instructions that may be executed by the processor and data that may be retrieved, manipulated or stored by the processor.

[0127] The processor and / or memory may be operably connected to a display device, for example, via a wired, such as a Universal Serial Bus (USB) connection, or wireless connection, such as a Bluetooth connection. Any convenient display device, such as a liquid crystal display (LCD), lightemitting diode (LED) display, plasma (PDP) display, quantum dot (QLED) display or cathode ray tube display device may be used. The display component may display information regarding whether an intended movement state or a stationary state is detected for the subject, information about brainactivity associated with intended movement, current stimulation parameters, or recommended changes to the stimulation parameters.

[0128] The storage component may be of any type capable of storing information accessible by the processor, such as a hard-drive, memory card, ROM, RAM, DVD, CD-ROM, USB Flash drive, write- capable, and read-only memories. The processor may be a general purpose processor, a graphics processor unit, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein.

[0129] A general purpose processor can be a microprocessor, but in the alternative, the processor can be a controller, microcontroller, or state machine, combinations of the same, or the like. A processor can also be implemented as a combination of computing devices, e.g., a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in conjunction with a DSP core, or any other such configuration. Although described herein primarily with respect to digital technology, a processor can also include primarily analog components. A computing environment can include any type of computer system, including, but not limited to, a computer system based on a microprocessor, a graphics processor unit, a mainframe computer, a digital signal processor, a portable computing device, a personal organizer, a device controller, and a computational engine within an appliance, to name a few.

[0130] The steps of a method, process, or algorithm described in connection with the embodiments disclosed herein can be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module, engine, and associated databases can reside in memory resources such as in RAM memory, FRAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, a removable disk, a CD-ROM, or any other form of non-transitory computer-readable storage medium, media, or physical computer storage known in the art. An exemplary storage medium can be coupled to the processor such that the processor can read information from, and write information to, the storage medium. In the alternative, the storage medium can be integral to the processor. The processor and the storage medium can reside in an ASIC. The ASIC can reside in a user terminal. In the alternative, the processor and the storage medium can reside as discrete components in a user terminal.

[0131] The instructions may be any set of instructions to be executed directly (such as machine code) or indirectly (such as scripts) by the processor. In that regard, the terms "instructions," "steps" and "programs" may be used interchangeably herein. The instructions may be stored in object code form for direct processing by the processor, or in any other computer language including scripts orcollections of independent source code modules that are interpreted on demand or compiled in advance.

[0132] Data may be retrieved, stored or modified by the processor in accordance with the instructions. For instance, although the system is not limited by any particular data structure, the data may be stored in computer registers, in a relational database as a table having a plurality of different fields and records, XML documents, or flat files. The data may also be formatted in any computer-readable format such as, but not limited to, binary values, ASCII or Unicode. Moreover, the data may comprise any information sufficient to identify the relevant information, such as numbers, descriptive text, proprietary codes, pointers, references to data stored in other memories (including other network locations) or information which is used by a function to calculate the relevant data.

[0133] In certain embodiments, the processor and storage component may comprise multiple processors and storage components that may or may not be stored within the same physical housing. For example, some of the instructions and data may be stored on removable CD-ROM and others within a read-only computer chip. Some or all of the instructions and data may be stored in a location physically remote from, yet still accessible by, the processor. Similarly, the processor may comprise a collection of processors which may or may not operate in parallel.

[0134] In some embodiments, the method can be performed using a cloud computing system. In these embodiments, the experimental data from measuring the response(s) of the subject to treatment with the ultrasound neuromodulation and the programming for optimizing ultrasound stimulation parameters can be exported to a cloud computer, which runs the program, and returns an output to the user.

[0135] Components of systems for carrying out the presently disclosed methods are further described in the examples below.Systems

[0136] The present disclosure also provides systems which find use, e.g., in practicing the subject methods. The system may be an open-loop or closed-loop system configured for performing the methods provided herein. Any suitable ultrasound transducer may be included in the system, including any commercially available programmable ultrasound transducer suitable for performing ultrasound neuromodulation to treat a neurological or neuropsychiatric disorder. In some embodiments, the ultrasound transducer is a wearable ultrasound array. In some embodiments, the ultrasound transducer is a transcranial focused ultrasound (FUS) transducer, phased array transducer, capacitive micromachined ultrasound transducer (CMUT), piezoelectric micromachinedultrasonic transducer (pMUT), or miniaturized half-concave transducer. Commercially available ultrasound transducers are available, for example, from FUS Instruments Inc. (Toronto, ON, Canada) and Phillips Engineering Solutions (Eindhoven, Netherlands).

[0137] In some embodiments, the system further comprises one or more devices to measure the response of the subject to treatment with ultrasound neuromodulation such as, but not limited to, an accelerometer, gyroscope, video recording device, or EEG electrodes, or a combination thereof. In some embodiments, the system comprises a wearable monitor that can acquire accelerometry, gyroscope and / or surface electromyographic (sEMG) data of the subject to detect movement of the subject. In some embodiments, the wearable monitor is a wrist-worn monitor such as a smartwatch. For example, a wrist-watch style wearable monitor such as the Parkinson’s KinetiGraph®, PKG®, available from PKG Health (San Francisco, GA), can be used to monitor movement continuously to detect various motor symptoms of a movement disorder such as bradykinesia, dyskinesia, tremor, daytime immobility, stiffness, slow movements, gait / walking, daytime somnolence, and sleep fragmentation. An Apple Watch is available from Apple Inc. (Cupertino, GA), which can be used, for example, for monitoring, bradykinesia, dyskinesia, tremors, gait, and arm movement. In addition, the system may further comprise one or more video recording devices capable of providing multiview video recordings of the subject’s body for monitoring movement and / or detecting facial expressions or facial emotion based on video recordings of the subject’s face.

[0138] In a closed-loop system, the system may include a computing means and control unit programmed to instruct an ultrasound transducer to deliver ultrasound neuromodulation to a target region of the brain of the subject in a manner effective to treat a neurological or neuropsychiatric disorder. In some embodiments, one or more programmed stimulation parameters are optimized according to an algorithm’s control law based on experimental data for measured responses of the subject to ultrasound neuromodulation, and optimized ultrasound neuromodulation is delivered to the brain by the ultrasound transducer in a manner effective to treat a symptom of the neurological or neuropsychiatric disorder. A closed loop system may apply ultrasound neuromodulation to the brain with optimized ultrasound stimulation parameters automatically upon receiving a communication from the control unit.

[0139] The processor of the closed-loop system may run programming for assessing the effectiveness of treatment and optimize an ultrasound stimulation parameter as needed without user intervention. Thus, the closed-loop system may not necessarily include a user interface for a user to instruct the ultrasound transducer to deliver ultrasound neuromodulation to a target region of the brain to treat a symptom of a neurological or neuropsychiatric disorder in the subject. However, in some embodiments, a user interface may be included in the closed-loop system which may be usedto confirm the recommendation of the closed loop system, or to override it, or to change the recommendation.

[0140] In certain embodiments, the system comprises a user interface comprising an input electronically coupled to a processor for instructing the ultrasound transducer to apply ultrasound neuromodulation to treat a symptom of a neurological or neuropsychiatric disorder in the subject. In some embodiments, the user interface is password protected and is operable by a health care practitioner. In some embodiments, the system comprises an operator workstation comprising a display, one or more input devices such as a keyboard and mouse, or the like, and a processor. The operator workstation may provide an operator interface that enables ultrasound stimulation parameters to be entered manually for operating the ultrasound transducer. In certain embodiments, the system displays optimized stimulation parameters for treating a symptom of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation, which can be entered manually for operating the ultrasound transducer.

[0141] In certain embodiments, the system displays a user interface presenting a questionnaire configured to receive input from the subject regarding self-reported results of the treatment with the ultrasound neuromodulation. In certain embodiments, the questionnaire uses a Movement Disorder Society-Sponsored Revision of the Unified Parkinson's Disease Rating Scale (MDS-UPDRS), a Hoehn and Yahr (HnY) scale, a Clinical Rating Scale for Tremor (CRST), a Parkinson’s Disease Composite Scale (PDCS), or a Schwab and England Activities of Daily Living (ADL) Scale for assessing treatment of a movement disorder. In certain embodiments, the questionnaire uses a Visual Analogue Scale for Depression (VAS-D), a Hamilton Depression Rating Scale (HAM-D), Montgomery-Asberg Depression Rating Scale (MADRS), a Beck Depression Inventories (BDI) score, a Visual Analogue Scale for Anxiety (VAS-A) or a Beck Anxiety Inventories (BAI) score to assess treatment of depression or anxiety. In certain embodiments, the questionnaire uses a numerical rating scale (NRS), a visual analog scale (VAS), or a categorical scale. In some embodiments, a Wong-Baker Faces Pain Scale, a FLACC Pain Scale, a CRIES Pain Scale, COMFORT Pain Scale, a McGill Pain Questionnaire, a Color Analog Pain Scale, Mankoski Pain Scale, a Brief Pain Inventory, or a Descriptor Differential Scale of Pain Intensity, or a combination thereof to evaluate pain. In certain embodiments, the questionnaire uses a visual-analog scale (VAS), a psychomotor vigilance test (PVT), a Likert scale, a Stanford Sleepiness Scale (SSS), a maintenance of wakefulness test (MWT), an Epworth sleepiness scale (ESS), a multiple sleep latency test (MSLT), or an Athens insomnia scale to evaluate level of wakefulness / arousal or sleepiness.

[0142] Components of systems for carrying out the presently disclosed methods are further described in the examples below.Utility

[0143] The methods and systems of the present disclosure find use in the treatment of a subject for a neurological or neuropsychiatric disorder by using ultrasound neuromodulation. Ultrasound stimulation parameters can be rapidly optimized in a personalized manner to treat a neurological or neuropsychiatric disorder with ultrasound neuromodulation. The subject methods can be used to treat neurological and neuropsychiatric disorders including, but not limited to, neuropsychiatric disorders such as major depressive disorder (MDD), generalized anxiety disorder (GAD), post- traumatic stress disorder (PTSD), addiction, anorexia, obsessive-compulsive disorder (OCD), bipolar disorder (BD) and chronic pain; and neurological disorders such as apraxia, agnosia, amnesia, aphasia, dysarthria, peripheral neuropathy, trigeminal neuralgia, dysautonomia, multiple system atrophy, epilepsy, movement disorders such as Parkinson's disease, essential tremor, amyotrophic lateral sclerosis, and Tourette’s syndrome, multiple sclerosis, narcolepsy, speech disorders, headaches, including migraines, cluster headaches, and tension headaches, pain, including complex regional pain syndrome and fibromyalgia, delirium, dementia, Alzheimer's disease, coma, and stroke; and sleep-wake disorders such as hypersomnia, excessive daytime sleepiness, insomnia, sleep apnea, parasomnia, narcolepsy, restless leg syndrome, non-rapid eye movement (NREM) sleep arousal disorder, and rapid and eye movement (REM) sleep behavior disorder.Examples of Non-Limiting Aspects of the Disclosure

[0144] Aspects, including embodiments, of the present subject matter described above may be beneficial alone or in combination with one or more other aspects or embodiments. Without limiting the foregoing description, certain non-limiting aspects of the disclosure numbered 1-77 are provided below. As will be apparent to those of skill in the art upon reading this disclosure, each of the individually numbered aspects may be used or combined with any of the preceding or following individually numbered aspects. This is intended to provide support for all such combinations of aspects and is not limited to combinations of aspects explicitly provided below.1 . A method for treating a neurological or neuropsychiatric disorder in a subject, the method comprising:(a) administering ultrasound neuromodulation to the subject using an initial selected set of ultrasound stimulation parameters;(b) measuring a response of the subject to said administering the ultrasound neuromodulation using the initial selected set of ultrasound stimulation parameters, wherein the measured response is used to calculate an efficacy score for treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation using the initial selected set of ultrasound stimulation parameters;(c) using the efficacy score from said administering the ultrasound neuromodulation using the initial selected set of ultrasound stimulation parameters to generate a Gaussian process regression model of an objective function to simulate responses of the subject to ultrasound neuromodulation administered using other ultrasound stimulation parameters;(d) performing Bayesian optimization with the Gaussian process regression model of the objective function using an acquisition function that selects a new set of ultrasound stimulation parameters with a goal of improving efficacy of treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation;(e) administering ultrasound neuromodulation to the subject using the new set of ultrasound stimulation parameters;(f) measuring a response of the subject to said administering the ultrasound neuromodulation using the new set of ultrasound stimulation parameters;(g) calculating an efficacy score for treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation using the new set of ultrasound stimulation parameters;(h) updating the Gaussian process regression model of the objective function with the efficacy score for the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation using the new set of ultrasound stimulation parameters, wherein the Gaussian process regression model of the objective function uses all the efficacy scores that have been calculated for the responses to all previous administrations of the ultrasound neuromodulation to the subject with previously selected ultrasound stimulation parameters;(i) performing Bayesian optimization with the updated Gaussian process regression model of the objective function, wherein the acquisition function selects another new set of ultrasound stimulation parameters with the goal of further improving the efficacy of the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation;(j) repeating steps (e)-(i) iteratively to further optimize the ultrasound stimulation parameters until a stopping criterion is met, wherein the acquisition function selects a final set of optimized ultrasound stimulation parameters; and(k) administering ultrasound neuromodulation to the subject using the final set of optimized ultrasound stimulation parameters.2. The method of aspect 1 , wherein the stopping criterion is determined by a strategy based on probabilistic regret bounds, wherein the probabilistic regret bound statistical test comprises: defining a regret threshold; estimating probability that the regret threshold can be surpassed by repeating steps (e)-(i) after each iteration of step (j); discontinuing the repeating of steps (e)-(i) when the estimated probability that the regret threshold can be surpassed falls below a chosen confidence level.3. The method of aspect 1 , wherein the stopping criterion is based on acquisition function convergence, wherein steps (e)-(i) are repeated until the acquisition function values converge below a pre-determined threshold.4. The method of aspect 1 , wherein the stopping criterion is based on uncertainty reduction, wherein steps (e)-(i) are repeated until posterior variance around an estimated optimum is below a pre-determined threshold.5. The method of aspect 1 , further comprising performing cross-validation or bootstrapping, wherein the stopping criterion is based on stability of an identified optimum, wherein steps (e)-(i) are repeated until re-sampling shows stability of the identified optimum.6. The method of any one of aspects 1-5, wherein steps (e)-(i) are repeated until convergence to an optimized target brain region and ultrasound stimulation parameters that are effective for treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation.7. The method of aspect 1 , wherein the stopping criterion is that further Bayesian optimization of the ultrasound stimulation parameters no longer results in further improvement of the efficacy of the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation.8. The method of aspect 1 , wherein the stopping criterion is based on a predetermined optimization budget.9. The method of any one of aspects 1-8, wherein the acquisition function is a noisy expected improvement (qNEI) acquisition function, a max-value entropy search (MES) acquisition function, or a joint entropy search (JES) acquisition function.10. The method of any one of aspects 1 -9, further comprising optimizing the parameters of the Gaussian process and of the acquisition function based on a model of the noise present in the data.11. The method of aspect 10, wherein Gaussian noise is modeled from a normal distribution N(0, a), wherein a is equal to a selected noise fraction.12. The method of any one of aspects 1-11 , wherein the initial selected set of ultrasound stimulation parameters is selected using random sampling, Sobol sequence sampling, a pseudorandom distribution, or Voronoi parcellation.13. The method of any one of aspects 1 -12, further comprising repeating steps (a)-(k) at a later time to further optimize the ultrasound stimulation parameters for the subject at the later time.14. The method of aspect 13, wherein disease progression in the subject, a change in medication administered to the subject, or adaptation of the subject to the ultrasound neuromodulation have diminished the efficacy of the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation over time.15. The method of aspect 14, wherein a model of the objective function estimated from a group of previously treated subjects with a given neurological or neuropsychiatric disorder is used to initialize the Gaussian process regression.16. The method of any one of aspects 1 -15, wherein said performing Bayesian optimization comprises optimizing hyperparameters comprising the acquisition function, seed queries, and an exploration / exploitation ratio.17. The method of any one of aspects 1 -16, wherein said performing Bayesian optimization comprises optimizing at least 2, at least 3, or at least 4 ultrasound stimulationparameters selected from a target brain region, a stimulation duration, an acoustic intensity, an acoustic pressure, an excitation voltage, a pulse repetition frequency, a pulse length, an ultrasound frequency, and a duty cycle.18. The method of aspect 17, wherein the ultrasound stimulation parameters comprise the target brain region, the stimulation duration, and the excitation voltage.19. The method of aspect 18, wherein the ultrasound stimulation parameters comprise the target brain region, the stimulation duration, the acoustic intensity, the acoustic pressure, the excitation voltage, the pulse repetition frequency, the pulse length, the ultrasound frequency, and the duty cycle.20. The method of any one of aspects 1 -19, wherein the response is a body movement, a facial movement, a change in locomotor activity, a change in mood, or a change in brain electrical activity.21 . The method of aspect 20, wherein the body movement, the facial movement, or the change in locomotor activity is measured using an accelerometer, a gyroscope, or a video recording device that records images of the subject.22. The method of aspect 21 , wherein the accelerometer or the gyroscope is provided by a wearable device.23. The method of aspect 22, wherein the wearable device is a smartwatch.24. The method of aspect 21 , wherein the video recording device is provided by a digital camera, a smartphone, a tablet, a laptop, or a camcorder.25. The method of aspect 20, wherein the brain electrical activity is measured by electroencephalography (EEG), stereoelectroencephalography (sEEG), electrocorticography (ECoG), magnetoencephalography (MEG), single photon emission computed tomography (SPECT), or functional magnetic resonance imaging (fMRI).26. The method of any one of aspects 1-25, wherein the neurological disorder is a movement disorder, and wherein the ultrasound stimulation parameters are optimized to ameliorate a symptom of the movement disorder.27. The method of aspect 26, wherein the symptom is bradykinesia, dyskinesia, or an abnormal or involuntary facial movement.28. The method of any one of aspects 1 -25, wherein the neuropsychiatric disorder is depression or anxiety, and wherein the ultrasound stimulation parameters are optimized to ameliorate the depression or the anxiety.29. The method of aspect 28, wherein the response is measured using a Visual Analogue Scale for Depression (VAS-D), a Hamilton Depression Rating Scale (HAM-D), Montgomery-Asberg Depression Rating Scale (MADRS), a Beck Depression Inventories (BDI) score, a Visual Analogue Scale for Anxiety (VAS-A) or a Beck Anxiety Inventories (BAI) score.30. The method of aspect 28, wherein the response is measured based on facial expression detection or facial emotion recognition analysis of a video recording of the subject’s face.31. The method of any one of aspects 1 -25, wherein the neurological disorder is neuropathic pain, and wherein the ultrasound stimulation parameters are optimized to ameliorate the neuropathic pain.32. The method of aspect 31 , wherein the response is measured using a numerical rating scale (NRS), a visual analog scale (VAS), or a categorical scale. In some embodiments, a Wong- Baker Faces Pain Scale, a FLACC Pain Scale, a CRIES Pain Scale, COMFORT Pain Scale, a McGill Pain Questionnaire, a Color Analog Pain Scale, Mankoski Pain Scale, a Brief Pain Inventory, or a Descriptor Differential Scale of Pain Intensity, or a combination thereof is used to evaluate pain.33. The method of any one of aspects 1-25, wherein the neurological disorder is a sleepwake disorder, and wherein the ultrasound stimulation parameters are optimized to ameliorate symptoms of the sleep-wake disorder.34. The method of aspect 33, wherein the response is measured using a visual-analog scale (VAS), a psychomotor vigilance test (PVT), a Likert scale, a Stanford Sleepiness Scale (SSS), a maintenance of wakefulness test (MWT), an Epworth sleepiness scale (ESS), a multiple sleep latency test (MSLT), or an Athens insomnia scale, or by monitoring the subject using actigraphy, electroencephalography, or polysomnography.35. The method of aspect 33, wherein said administering the ultrasound neuromodulation increases arousal compared to in absence of said administering the ultrasound neuromodulation.36. The method of any one of aspects 1 -35, wherein the ultrasound neuromodulation is administered with a portable ultrasound transducer or a wearable ultrasound array.37. The method of any one of aspects 1 -36, wherein the ultrasound neuromodulation is administered with a transcranial focused ultrasound (FUS) transducer, phased array transducer, capacitive micromachined ultrasound transducer (CMUT), piezoelectric micromachined ultrasonic transducer (pMUT), or miniaturized half-concave transducer.38. A computer-implemented method for programming an ultrasound transducer to treat a subject for a neurological or neuropsychiatric disorder with ultrasound neuromodulation, the computer performing steps comprising:(a) instructing the ultrasound transducer to deliver ultrasound neuromodulation to the subject using an initial selected set of ultrasound stimulation parameters;(b) receiving experimental data from measuring a response of the subject to administering the ultrasound neuromodulation using the initial selected set of ultrasound stimulation parameters, wherein the measured response is used to calculate an efficacy score for treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation using the initial selected set of ultrasound stimulation parameters;(c) using the efficacy score from said administering the ultrasound neuromodulation using the initial selected set of ultrasound stimulation parameters to generate a Gaussian process regression model of an objective function to simulate responses of the subject to ultrasound neuromodulation administered using other ultrasound stimulation parameters;(d) performing Bayesian optimization with the Gaussian process regression model of the objective function using an acquisition function that selects a new set of ultrasound stimulationparameters with the goal of improving the efficacy of treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation;(e) instructing the ultrasound transducer to deliver ultrasound neuromodulation to the subject using the new set of ultrasound stimulation parameters;(f) receiving experimental data from measuring a response of the subject to administering the ultrasound neuromodulation using the new set of ultrasound stimulation parameters;(g) calculating an efficacy score for treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation using the new set of ultrasound stimulation parameters;(h) updating the Gaussian process regression model of the objective function with the efficacy score for the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation using the new set of ultrasound stimulation parameters, wherein the Gaussian process regression model of the objective function uses all the efficacy scores that have been calculated for the responses to all previous administrations of the ultrasound neuromodulation to the subject with previously selected ultrasound stimulation parameters;(i) performing Bayesian optimization with the updated Gaussian process regression model of the objective function, wherein the acquisition function selects another new set of ultrasound stimulation parameters with the goal of further improving the efficacy of the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation;(j) repeating steps (e)-(i) iteratively to further optimize the ultrasound stimulation parameters until a stopping criterion is met, wherein the acquisition function selects a final set of optimized ultrasound stimulation parameters; and(k) instructing the ultrasound transducer to deliver ultrasound neuromodulation to the subject using the final set of optimized ultrasound stimulation parameters.39. The computer-implemented method of aspect 38, wherein the stopping criterion is determined by a strategy based on probabilistic regret bounds, wherein the probabilistic regret bound statistical test comprises: defining a regret threshold; estimating probability that the regret threshold can be surpassed by repeating steps (e)-(i) after each iteration of step (j); discontinuing the repeating of steps (e)-(i) when the estimated probability that the regret threshold can be surpassed falls below a chosen confidence level.40. The computer-implemented method of aspect 38, wherein the stopping criterion is based on acquisition function convergence, wherein steps (e)-(i) are repeated until the acquisition function values converge below a pre-determined threshold.41 . The computer-implemented method of aspect 38, wherein the stopping criterion is based on uncertainty reduction, wherein steps (e)-(i) are repeated until posterior variance around an estimated optimum is below a pre-determined threshold.42. The computer-implemented method of aspect 38, further comprising performing cross-validation or bootstrapping, wherein the stopping criterion is based on stability of an identified optimum, wherein steps (e)-(i) are repeated until re-sampling shows stability of the identified optimum.43. The computer-implemented method of any one of aspects 38-42, wherein steps (e)- (i) are repeated until convergence to an optimized target brain region and ultrasound stimulation parameters that are effective for treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation.44. The computer-implemented method of aspect 38, wherein the stopping criterion is that further Bayesian optimization of the ultrasound stimulation parameters no longer results in further improvement of the efficacy of the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation.45. The computer-implemented method of aspect 38, wherein the stopping criterion is based on a predetermined optimization budget.46. The computer-implemented method of any one of aspects 38-45, wherein the acquisition function is a noisy expected improvement (qNEI) acquisition function, a max-value entropy search (MES) acquisition function, or a joint entropy search (JES) acquisition function.47. The computer-implemented method of any one of aspects 38-46, further comprising optimizing the parameters of the Gaussian process and of the acquisition function based on a model of the noise present in the data.48. The computer-implemented method of aspect 47, wherein Gaussian noise is modeled from a normal distribution N(0, o), wherein o is equal to a selected noise fraction.49. The computer-implemented method of any one of aspects 38-48, wherein the initial selected set of ultrasound stimulation parameters is selected using random sampling, Sobol sequence sampling, a pseudo-random distribution, or Voronoi parcellation.50. The computer-implemented method of any one of aspects 38-49, further comprising repeating steps (a)-(k) at a later time to further optimize the ultrasound stimulation parameters for the subject at the later time.51 . The computer-implemented method of aspect 50, wherein disease progression in the subject, a change in medication administered to the subject, or adaptation of the subject to the ultrasound neuromodulation have diminished the efficacy of the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation over time.52. The computer-implemented method of aspect 51 , wherein a model of the objective function estimated from a group of previously treated subjects with a given neurological or neuropsychiatric disorder is used to initialize the Gaussian process regression.53. The computer-implemented method of any one of aspects 38-52, wherein said performing Bayesian optimization comprises optimizing hyperparameters comprising the acquisition function, seed queries, and an exploration / exploitation ratio.54. The computer-implemented method of any one of aspects 38-53, wherein the ultrasound stimulation parameters comprise at least 2, at least 3, or at least 4 ultrasound stimulation parameters selected from a target brain region, a stimulation duration, an acoustic intensity, an acoustic pressure, an excitation voltage, a pulse repetition frequency, a pulse length, an ultrasound frequency, and a duty cycle.55. The computer-implemented method of aspect 54, wherein the ultrasound stimulation parameters comprise the target brain region, the stimulation duration, and the excitation voltage.56. The computer-implemented method of aspect 55, wherein the ultrasound stimulation parameters comprise the target brain region, the stimulation duration, the acoustic intensity, the acoustic pressure, the excitation voltage, the pulse repetition frequency, the pulse length, the ultrasound frequency, and the duty cycle.57. The computer-implemented method of any one of aspects 38-56, wherein the response is a body movement, a facial movement, a change in locomotor activity, a change in mood, or a change in brain electrical activity.58. The computer-implemented method of any one of aspects 38-57, further comprising displaying the optimized set of ultrasound stimulation parameters.59. The computer-implemented method of any one of aspects 38-58, further comprising displaying a user interface presenting a questionnaire configured to receive input from the subject regarding self-reported results of treatment with the ultrasound neuromodulation.60. The computer-implemented method of aspect 59, wherein the questionnaire uses a Movement Disorder Society-Sponsored Revision of the Unified Parkinson's Disease Rating Scale (MDS-UPDRS), a Hoehn and Yahr (HnY) scale, a Clinical Rating Scale for Tremor (CRST), a Parkinson’s Disease Composite Scale (PDCS), or a Schwab and England Activities of Daily Living (ADL) Scale for assessing treatment of a movement disorder.61 . The computer-implemented method of aspect 59, wherein the questionnaire uses a Visual Analogue Scale for Depression (VAS-D), a Hamilton Depression Rating Scale (HAM-D), Montgomery-Asberg Depression Rating Scale (MADRS), a Beck Depression Inventories (BDI) score, a Visual Analogue Scale for Anxiety (VAS-A) or a Beck Anxiety Inventories (BAI) score for assessing treatment of depression or anxiety.62. The computer-implemented method of aspect 59, wherein the questionnaire uses a numerical rating scale (NRS), a visual analog scale (VAS), or a categorical scale. In some embodiments, a Wong-Baker Faces Pain Scale, a FLACC Pain Scale, a CRIES Pain Scale, COMFORT Pain Scale, a McGill Pain Questionnaire, a Color Analog Pain Scale, Mankoski Pain Scale, a Brief Pain Inventory, or a Descriptor Differential Scale of Pain Intensity, or a combination thereof to evaluate pain.63. The computer-implemented method of aspect 59, wherein the questionnaire uses a visual-analog scale (VAS), a psychomotor vigilance test (PVT), a Likert scale, a Stanford Sleepiness Scale (SSS), a maintenance of wakefulness test (MWT), an Epworth sleepiness scale (ESS), a multiple sleep latency test (MSLT), or an Athens insomnia scale to evaluate level of wakefulness, arousal, or sleepiness.64. A non-transitory computer-readable medium comprising program instructions that, when executed by a processor in a computer, causes the processor to perform the method of any one of aspects 38-63.65. A kit comprising the non-transitory computer-readable medium of aspect 64 and instructions for determining symptom severity of a subject having a movement disorder.66. A system for treating a neurological or neuropsychiatric disorder with ultrasound neuromodulation, the system comprising: an ultrasound transducer; and a processor programmed according to the computer-implemented method of any one of aspects 38-63 to instruct the ultrasound transducer to deliver ultrasound neuromodulation to the subject using an optimized set of ultrasound stimulation parameters.67. The system of aspect 66, further comprising an accelerometer, a gyroscope, a video recording device, an electroencephalography (EEG) electrode, or a combination thereof.68. The system of aspect 67, wherein the accelerometer, the gyroscope, or the combination thereof is provided by a wearable device.69. The system of aspect 68, wherein the wearable device is a smartwatch.70. The system of aspect 67, wherein the video recording device is provided by a digital camera, a smartphone, a tablet, a laptop, or a camcorder.71. The system of any one of aspects 66-70, wherein the ultrasound transducer is portable or a wearable ultrasound array.72. The system of any one of aspects 66-71 , wherein the ultrasound transducer is a transcranial focused ultrasound (FLIS) transducer, phased array transducer, capacitive micromachined ultrasound transducer (CMUT), piezoelectric micromachined ultrasonic transducer (pMUT), or miniaturized half-concave transducer.73. The system of any one of aspects 66-72, wherein the system further comprises a user interface comprising an input electronically coupled to the processor for instructing the ultrasound transducer to deliver ultrasound neuromodulation to the subject to treat the neurological or neuropsychiatric disorder.74. The system of aspect 73, wherein the user interface is password protected and is operable by a health care practitioner.75. The system of any one of aspects 66-74, further comprising a storage component for storing data, wherein the storage component is coupled to the processor.76. The system of any one of aspects 66-75, further comprising a display for displaying the optimized set of ultrasound stimulation parameters.77. The system of aspect 76, wherein the display further displays a user interface presenting a questionnaire configured to receive input from the subject regarding self-reported results of treatment with the ultrasound neuromodulation.

[0145] It will be apparent to one of ordinary skill in the art that various changes and modifications can be made without departing from the spirit or scope of the invention.EXPERIMENTAL

[0146] The following examples are put forth so as to provide those of ordinary skill in the art with a complete disclosure and description of how to make and use the present invention, and are not intended to limit the scope of what the inventors regard as their invention nor are they intended to represent that the experiments below are all or the only experiments performed. Efforts have been made to ensure accuracy with respect to numbers used (e.g. amounts, temperature, etc.) but some experimental errors and deviations should be accounted for. Unless indicated otherwise, parts areparts by weight, molecular weight is weight average molecular weight, temperature is in degrees Centigrade, and pressure is at or near atmospheric.

[0147] All publications and patent applications cited in this specification are herein incorporated by reference as if each individual publication or patent application were specifically and individually indicated to be incorporated by reference.

[0148] The present invention has been described in terms of particular embodiments found or proposed by the present inventor to comprise preferred modes for the practice of the invention. It will be appreciated by those of skill in the art that, in light of the present disclosure, numerous modifications and changes can be made in the particular embodiments exemplified without departing from the intended scope of the invention. For example, due to codon redundancy, changes can be made in the underlying DNA sequence without affecting the protein sequence. Moreover, due to biological functional equivalency considerations, changes can be made in protein structure without affecting the biological action in kind or amount. All such modifications are intended to be included within the scope of the appended claims.Example 1Efficient Data-Driven Mapping of Ultrasound Neuromodulation Parameters Using Bayesian OptimizationBackground

[0149] Low-intensity focused ultrasound (LIFU) neuromodulation has shown promising results both in preclinical animal models1-5and humans6-8but its parameter space has remained poorly mapped. Sonication parameters such as acoustic pressure, intensity, pulse repetition frequency, duty cycle, and stimulation duration have all been demonstrated to affect behavioral and neurophysiological responses to LIFU stimulation with nonlinear and non-monotonic dose-response relationships1 ,9-11. In addition, sonication protocols applied to different brain targets in the same subjects can evoke diverging effects12,13. Moreover, nonspecific responses produced by auditory or other peripheral activations can confound the interpretation of the LIFU-evoked effects14,15. Given the significant inter- and intra-subject variability and the high dimensionality of the LIFU parameter space, performing an exhaustive search for subject-specific parameter mapping would require a large number of stimulation-response evaluations, making it impractical or even unfeasible in patient populations. In addition, time-dependent factors like disease progression and neuroplastic adaptations to the stimulation may hinder therapeutic efficacy over time, requiring repeated optimizations over thecourse of the treatment16. Altogether, these challenges underscore a critical need for systematic and efficient parameter mapping strategies to accelerate the identification of therapeutically effective LIFU neuromodulation protocols.

[0150] To address this gap, we propose a method for subject-specific optimization of LIFU parameters based on Bayesian optimization (BO) with Gaussian process (GP) regression. BO enables effective, data-driven parameter mapping by maximizing (or minimizing) an unknown objective function with as few evaluations as possible17 18. Compared to other optimization approaches commonly used in experimental settings, such as grid search and random search, BO has been demonstrated to converge faster and to yield better, more robust solutions in a number of different applications19-21. BO has also shown promising results with other neurostimulation modalities including deep brain stimulation21-24, transcranial magnetic stimulation25, transcranial current stimulation26, vagus nerve stimulation27, and neuroprosthetics20'28-30. Therefore, we hypothesized that this approach could be adapted for effective mapping of LIFU neuromodulation parameters.

[0151] Here we determined the feasibility of BO for data-driven optimization of LIFU parameters. We trained the optimizer and tuned relevant BO hyperparameters using a synthetic data model that provides on-demand stimulation-response evaluations, based on experimental data of LIFU stimulation of the centromedial (CM) thalamus (CM-LIFU) in a rat model of arousal from our prior work1.ResultsBayesian optimization

[0152] A schematic representation of the BO loop is shown in FIG. 1 . In the initialization phase (FIG. 1A), Ninit parameter combinations X are evaluated, and a response Y(X) is recorded for each combination (e.g., a behavioral, neurophysiological, or neuroimaging readout). These initial evaluations are used to fit a GP to the parameter space to create an initial (prior) surrogate model of the objective function. The information provided by this model is then leveraged by the acquisition function to select the next candidate to test Xn(FIG. I B). A new LIFU stimulation is delivered using the Xnparameters, and a response ynis recorded. The surrogate objective model (posterior) is then updated via GP regression, the acquisition function selects the next parameter set based on the updated surrogate model, and the process is repeated iteratively until a stopping criterion is met. Typically, upon convergence of the BO loop, the maximum (minimum) of the GP model tends toward the maximum (minimum) of the true objective function (FIG. 1C).Ground-truth model

[0153] To benchmark the BO approach against a ground-truth objective function in a LIFU neuromodulation application, we built a model to generate synthetic, on-demand stimulationresponse evaluations based on our experimental mapping of CM-LIFU parameters in a rat model of arousal1. Experimental data were originally collected using a fixed, discrete grid that partially sampled the parameter space to evaluate the effect of the target (CM or control), the stimulation duration (1 - or 3-burst stimulation), and the stimulation intensity (2 to 12 V excitation, corresponding to a peak negative pressure between 0.13 and 0.97 MPa and a spatial-peak pulse-average intensity between 0.52 and 30.22 W / cm2). Other stimulation parameters were kept constant (see the Methods section for further details). This discrete parameter sampling is the standard approach in most experimental settings.

[0154] We generated a continuous representation of the ground-truth parameter space using a generalized additive model (GAM) with a gamma error distribution and a log link function. The model included 21 cubic spline basis functions and applied a smoothing penalty of 0.01. The resulting fits are shown in FIG. 1 D.

[0155] Experimental measurements in biological systems are typically affected by noise. T o simulate real-world conditions, we injected noise into the ground-truth model and tested three different noise levels, defined as fractions (0.25, 0.50, and 1 .00) of the standard deviation of the original ground truth distribution (FIG. 1 E). For each noise level, whenever the objective function was sampled during optimization, Gaussian noise was added to the returned value. This noise was drawn from a normal distribution N(0, a), with a equal to the selected noise fraction.Bayesian optimization hyperparameter tuning

[0156] We used the synthetic CM-LIFU model shown in FIGS. 1 D-1 E to tune relevant BO hyperparameters, including the acquisition function, the number N!niJo initialization points, and the initialization strategy (random vs. Sobol sequence sampling), under different noise levels. For each hyperparameter combination, we repeated the optimization process across 50 independent trials, each consisting of 100 or 300 iterations. We benchmarked the BO approach against two alternative optimization strategies: grid search (GS) and random search + GP (RGP). The former strategy was included to mimic a realistic parameter mapping scenario, as in our experimental setting1; the latter was used to isolate the specific contribution of the acquisition function. The errors are reported in units of normalized distance, as shown in FIGS. 1 D-1 E.Acquisition function

[0157] To identify the most suitable acquisition function for LIFU neuromodulation parameter tuning, we tested nine different functions (see the Methods section for further details): expected improvement (El) with= 0.1 and= 0.01 ; upper confidence bound (UCB) with A = 2 and A = 0.2; log expected improvement31(Log-El) with = 0.1 and = 0.01 ; noisy expected improvement32(qNEI); max-value entropy search33(MES); and joint entropy search34(JES). At each iteration, the error was computed as the difference between the global optimum of the ground-truth objective function model in FIG. 1 D (2.5533; measured in the following condition: CM; 3-burst stimulation; 9.8497 V excitation) and the objective evaluation at the argmax of the current surrogate model (for BO and RGP) or the argmax of the mean stimulation-response function (for GS).

[0158] In the low-noise condition and using 20 randomly distributed initialization points, all the BO acquisition functions outperformed both the GS and RGP methods, independent of the number of iterations (FIGS. 2A-2B). Specifically, BO yielded a median error of less than 0.0171 (0.67% of the objective max; IQR 0.0017-0.0572) after 25 iterations and less than 0.0134 (0.52% of the objective max; IQR 0.0015-0.0481 ) after 100 iterations (both maximum errors observed with the UCB acquisition function with A = 0.2). The RGP search achieved similar performance as the BO acquisition functions, with a median error of 0.0432 (1 .69% of the objective max; IQR 0.0059-0.135) after 100 iterations; however, the convergence was much slower than the BO searches. Interestingly, the median error for the GS method remained remarkably constant over the 100 iterations and consistently above 0.2803 (10.98% of the objective max). Furthermore, both GS and RGP searches displayed substantial variability across trials (FIG. 2B).

[0159] In the high-noise condition, all the acquisition functions showed slower speed of convergence; nonetheless, the BO method still outperformed both GS and RGP searches (FIGS. 2A-2B). GS yielded a median error of 0.4384 (17.17% of the objective max; IQR 0.1532-1.2249) after 25 iterations and 0.2803 (10.97% of the objective max; IQR 0.0531 -0.4674) after 100 iterations. Conversely, BO yielded a median error of less than 0.1817 (7.12% of the objective max; IQR 0.059- 1.0343) after 25 iterations and less than 0.1046 (4.1% of the objective max; IQR 0.0329-0.5908) after 100 iterations. As in the low-noise condition, the UCB acquisition function with A = 0.2 was the worst-performing one. The best-performing acquisition functions were MES, JES, and qNEI, which yielded a median error of 0.0604 (2.37% of the objective max; IQR 0.0132-0.2873), 0.0749 (2.93% of the objective max; IQR 0.0094-0.4123), and 0.1018 (3.99% of the objective max; IQR 0.0267- 0.2822), respectively, already after 25 iterations (see Supplementary Tables 1-3 for a summary of all the search statistics). Importantly, the median error for RGP at the same checkpoint was 0.6397 (25.05% of the objective max; IQR 0.1918-1.0596), confirming that the performance gain of the BOmethod over GS was the result of the GP regression combined with the acquisition function that steered the search in the maximally beneficial direction at each iteration.

[0160] To assess whether the performance of the GS method would eventually converge with that of BO, we extended the optimization budget to 300 iterations and tested the best-performing acquisition functions. Even after 300 iterations, the GS median error in the high-noise condition was 0.2079 (8.14% of the objective max; IQR 0.0531 -0.4106), which was substantially higher than the BO error observed after 25 iterations (FIG. 4).

[0161] To assess the reliability of the BO approach, we quantified the number of trials (out of 50) that produced an error below 1 %, 5%, or 10% of the maximum of the objective function (i.e., an absolute error lower than 0.0255, 0.1277, or 0.2553, respectively) at various checkpoints (FIG. 2C and FIG. 5). In the high-noise condition after 25 iterations, 32% (16 / 50), 34% (17 / 50), or 14% (7 / 50) of trials produced an error lower than 1% with the MES, JES, and qNEI acquisition functions, respectively, compared to 8% (4 / 50) of trails with GS and 10% (5 / 50) of trials with RGP. Even with a more lenient 10% error threshold, only 30% (15 / 50) of trials converged with GS and 28% (14 / 50) with RGP (versus 66% (33 / 50), 68% (34 / 50), and 58% (29 / 50) with the three BO acquisition functions). After 300 iterations, only 18% (9 / 50) of GS trials converged with an error lower than 1%, compared to 98-100% of trials converging with the three BO acquisition functions (see Supplementary Tables 4-6 for a summary of all trial convergence metrics).

[0162] Taken together, our results underscore the advantages of using a data-driven approach like BO, which does not rely on a fixed, a-priori determined sampling of the parameter space, over conventional grid search. In our simulated LIFU neuromodulation application, BO yields a more effective optimization of the parameter space, both in terms of final convergence error and number of iterations required. Our analysis also highlights that the results are greatly dependent on the levels of noise present in the data, and that acquisition functions designed for noisy settings perform better, as expected. Among them, the JES acquisition function offers the best tradeoff between optimization error and speed of convergence.Random initialization

[0163] To assess the impact of the number Nnt and the sampling strategy used for initialization, we repeated the BO search with the three best-performing acquisition functions (MES, JES, and qNEI) and measured the error in the low-, mid-, and high-noise conditions with M nit between 5 and 30 (FIG. 3 and FIG. 6), comparing random and Sobol sequence sampling. For the different Ninit values, we quantified the error at iteration 50 inclusive of the initialization points to normalize the total optimization budget spent across the different conditions (i.e., for the case with Nin>t = 5, we quantifiedthe error at BO iteration 45; for the case of Ninit = 10, we quantified the error at BO iteration 40; etc.). In the low- and mid-noise conditions, the median errors were 0.0054 (low-noise; 0.21% of the objective max; IQR 0.0020-0.0113) and 0.0079 (mid-noise; 0.31% of the objective max; IQR 0.0009- 0.0254), respectively, and were not significantly affected by either NM or the sampling strategy. However, in the high-noise trials the best performance was obtained with Ninit of 15 (median error: 0.0441 , 1 .73% of the objective max; IQR 0.0126-0.2205) or 20 (median error: 0.0761 , 2.98% of the objective max; IQR 0.0057-0.1987). This indicates that when less than 15 initialization points were used, the BO search struggled to find the optimal parameter combination due to the insufficient characterization of the parameter space at the start of the BO loop. On the other hand, NM greater than 20 did not leave enough budget for the BO to converge to the optimal parameter combination by iteration 50 (FIG. 7). Unexpectedly, sampling based on Sobol sequences did not improve the converge of the BO search, despite providing a more uniform distribution of the initialization points (FIGS. 3A-3B).MethodsExperimental dataTraining data

[0164] The experimental data used for the training set has been described previously1. Briefly, the experiment involved delivering LIFU in freely behaving rats in an open-field chamber using a wearable ultrasound array of 64 randomly distributed elements operating at 1 MHz. Bottom-up videos were continuously recorded during the entire LIFU stimulation session. Open-field locomotor activity was estimated by tracking the rat’s instantaneous center position35, and we quantified distance traveled and speed time-locked to the LIFU stimuli. LIFU stimuli were delivered either to the central medial thalamus (CM) or to an active control target located at the ventral end of the dorsal peduncular cortex. Given the CM’s established role in arousal regulation, our hypothesis was that stimulation of the CM would elicit a greater increase in spontaneous locomotor activity compared to the control target1.

[0165] During each experimental session, LIFU stimuli were delivered in a block design consisting of 42 stimulation blocks presented every 2 min. Each block featured a specific combination of three stimulation parameters: target (either the CM or the active control region); stimulation duration (one or three 5-s bursts consisting of 80 ms sinusoidal pulses, with a pulse repetition period of 480 ms and an inter-burst interval of 10 s, resulting in a 17% duty cycle); and stimulus intensity (controlledvia the excitation voltage of the Verasonics scanner, ranging from 2 to 12 V in 2 V increments; the pressure, intensity, and temperature associated with each excitation voltage were estimated using acoustic and thermal simulations1). The stimulation duration was fixed at the session level, and the order of intensity and target conditions within each session was randomized to avoid order effects. Time-locked distance and speed were calculated within two post-stimulation windows in the 0-15 s and 15-30 s intervals.

[0166] To construct the dataset used in the BO training procedure, cumulative distance values from n = 3 rats and a total of 15 sessions were averaged across blocks and sessions, resulting in a single cumulative distance value per parameter combination per animal.Bayesian optimization implementationSurrogate model

[0167] The surrogate model of the objective function plays an essential role in BO, as it allows creation of an estimate of the unknown objective function. While BO can, in principle, be implemented with various types of surrogate models, Gaussian processes are particularly well-suited due to their analytical tractability and ability to provide an estimate for both the mean and uncertainty. Gaussian process regression is a non-parametric regression technique that places a Gaussian process prior over the latent function and, upon conditioning on observed data, yields a posterior predictive distribution providing an estimate for both the mean and the uncertainty. Importantly, Gaussian processes are entirely defined by only a mean function and a covariance (or kernel) function. For the mean function, we used a uniform constant mean function. For the kernel function, we tested two different kernel types implemented in the BoTorch framework36: a radial basis function kernel with a LogNormal prior on lengthscales that scales with the dimensionality of the problem37; a Matern kernel with a gamma prior on lengthscales. The BoTorch implementation of Gaussian process regression also includes a likelihood object that models the observation noise as homoscedastic and treats noise level as a hyperparameter.Acquisition function

[0168] The acquisition function is another key component of the BO process that provides a heuristic estimate of the potential utility of sampling a given point in the parameter space, based on the current surrogate model of the objective function. The acquisition function aims to balance exploration and exploitation of the parameter space, i.e., the resources spent exploring new regions of the parameterspace vs. exploiting / validating regions likely to contain optimal solutions. We evaluated the following acquisition functions:

[0169] 1 ) Expected improvement (El): El is one of the most widely used acquisition functions in BO.El quantifies the expected gain over the current best observation and incorporates both the probability and the size of improvement. Its behavior can be tuned toward exploration or exploitation by scaling the variance term with a coefficient Higher values favor exploration, while lower values favor exploitation. We tested= 0.1 and= 0.01 .

[0170] 2) Upper confidence bound (UCB): UCB combines the predicted mean p(x) and standard deviation o(x) of the surrogate model, with an explicit exploration / exploitation trade-off controlled by a hyperparameter A. Larger A values favor exploration, while smaller favor emphasize exploitation.

[0171] 3) Log expected improvement (Log-El): This is a stable implementation of El designed to mitigate numerically issues associated with computing El31.

[0172] 4) Noisy expected improvement (gNEI): This acquisition function extends El to noisy settings by directly integrating over the posterior distribution of the acquisition function. This is then optimized using a quasi-Monte Carlo approximation based on Sobol sequences32.

[0173] 5) Max-value entropy search (MES): This acquisition function is based on the concept ofShannon entropy and aims to maximize expected information gain about the global optimum value33.

[0174] 6) Joint entropy search (JES): This acquisition function is also based on the concept ofShannon entropy and maximizes the expected information gain about the joint optimal probability density over both the input and output space34.Implementation

[0175] For each combination of hyperparameters (number of seed points, acquisition function, noise level), we performed 50 optimization trials, each with an optimization budget of 100-300 iterations. To ensure comparability across different optimization strategies, pseudorandom number generation was controlled at the trial level using shared random seeds.Bayesian optimization benchmark

[0176] To assess the gains in optimization performance provided by the BO approach, we benchmarked it against two alternative optimization strategies:

[0177] 1 ) Grid search (GS): Candidates are sampled randomly from a discrete grid of 6 intensity values (ranging from 2 to 12 V with 2 V intervals). To avoid introducing sampling bias, the grid is randomly shifted in each optimization trial by a value uniformly drawn from the interval [-1 ,1].

[0178] 2) Random search + GP (RGP): Candidates are selected via uniform sampling from a continuous parameter space, when applicable (in our case, the only continuous variable was intensity). A Gaussian process model is then regressed after each objective evaluation (though the surrogate model is not used to guide sampling since no acquisition function is included).Optimization metrics

[0179] The main metric of interest was the difference between the true global optimum and the optimum as estimated by the optimization algorithm. The definition of this last value varied across different strategies:

[0180] BO and Random search + GP: The objective was evaluated at the argmax of the current surrogate model.

[0181] Grid Search: For each grid coordinate, the average of all sampled values at that coordinate is computed. The objective was evaluated at the argmax on the list of values observed so far.

[0182] For each optimization strategy, this metric was tracked over iterations within each optimization trial. We visualized the progression of these metrics and computed summary statistics at various iteration checkpoints. Speed of convergence, defined as the first time the difference between the true global optimum and the estimated optimum fell below certain thresholds (1%, 5%, or 10% of the true global optimum), was also recorded.References

[0183] 1. Di lanni, T. et al. High-throughput ultrasound neuromodulation in awake and freely behaving rats. Brain Stimulation 16, 1743-1752 (2023).

[0184] 2. Murphy, K. R. et al. Optimized ultrasound neuromodulation for non-invasive control of behavior and physiology. Neuron S0896627324004938 (2024) doi:10.1016 / j. neuron.2024.07.002.

[0185] 3. Yu, K., Niu, X., Krook-Magnuson, E. & He, B. Intrinsic functional neuron-type selectivity of transcranial focused ultrasound neuromodulation. Nat Common 12, 2519 (2021).

[0186] 4. Fouragnan, E. F. et al. The macaque anterior cingulate cortex translates counterfactual choice value into actual behavioral change. Nat Neurosci 22, 797-808 (2019).

[0187] 5. Yang, Y. et al. Induction of a torpor-like hypothermic and hypometabolic state in rodents by ultrasound. Nat Metab 5, 789-803 (2023).

[0188] 6. Fan, J. M. et al. Thalamic transcranial ultrasound stimulation in treatment resistant depression. Brain Stimulation 17, 1001-1004 (2024).

[0189] 7. Yaakub, S. N. et al. Transcranial focused ultrasound-mediated neurochemical and functional connectivity changes in deep cortical regions in humans. Nat Commun 14, 5318 (2023).

[0190] 8. Legon, W. et al. Transcranial focused ultrasound modulates the activity of primary somatosensory cortex in humans. Nat Neurosci 17, 322-329 (2014).

[0191] 9. Manuel, T. J. et al. Ultrasound neuromodulation depends on pulse repetition frequency and can modulate inhibitory effects of TTX. Sci Rep 10, 15347 (2020).

[0192] 10. Yuan, Y., Wang, Z., Liu, M. & Shoham, S. Cortical hemodynamic responses induced by low-intensity transcranial ultrasound stimulation of mouse cortex. NeuroImage 211 , 116597 (2020).

[0193] 11. Estrada, H. et al. Holographic transcranial ultrasound neuromodulation enhances stimulation efficacy by cooperatively recruiting distributed brain circuits. Nat. Biomed. Eng (2025) doi : 10.1038 / S41551 -025-01449-x.

[0194] 12. Yaakub, S. N. et al. Transcranial focused ultrasound-mediated neurochemical and functional connectivity changes in deep cortical regions in humans. Nat Commun 14, 5318 (2023).

[0195] 13. Legon, W., Strohman, A., In, A. & Payne, B. Noninvasive neuromodulation of subregions of the human insula differentially affect pain processing and heart-rate variability: a within-subjects pseudo-randomized trial. PAIN (2024).

[0196] 14. Kop, B. R. et al. Auditory confounds can drive online effects of transcranial ultrasonic stimulation in humans. eLife 12, RP88762 (2023).

[0197] 15. Guo, H. et al. Ultrasound Produces Extensive Brain Activation via a CochlearPathway. Neuron 98, 1020-1030.e4 (2018).

[0198] 16. Limousin, P. & Foltynie, T. Long-term outcomes of deep brain stimulation in Parkinson disease. Nat Rev Neurol 15, 234-242 (2019).

[0199] 17. Frazier, P. I. A Tutorial on Bayesian Optimization. Preprint at doi.org / 10.48550 / arXiv.1807.02811 (2018).

[0200] 18. Shahriari, B., Swersky, K., Wang, Z., Adams, R. P. & De Freitas, N. Taking the HumanOut of the Loop: A Review of Bayesian Optimization. Proc. IEEE 104, 148-175 (2016).

[0201] 19. Shields, B. J. et al. Bayesian reaction optimization as a tool for chemical synthesis.Nature 590, 89-96 (2021 ).

[0202] 20. Bonizzato, M. et al. Autonomous optimization of neuroprosthetic stimulation parameters that drive the motor cortex and spinal cord outputs in rats and monkeys. Cell Reports Medicine 4, 101008 (2023).

[0203] 21 . Stieve, B. J., Richner, T. J., Krook-Magnuson, C., Netoff, T. I. & Krook-Magnuson, E.Optimization of closed-loop electrical stimulation enables robust cerebellar-directed seizure control. Brain 146, 91-108 (2023).

[0204] 22. Fleming, J. E. et al. From dawn till dusk: Time-adaptive bayesian optimization for neurostimulation. PLoS Comput Biol 19, e101 1674 (2023).

[0205] 23. Cole, E. R. et al. SAFE-OPT: a Bayesian optimization algorithm for learning optimal deep brain stimulation parameters with safety constraints. J. Neural Eng. 21 , 046054 (2024).

[0206] 24. Grado, L. L., Johnson, M. D. & Netoff, T. I. Bayesian adaptive dual control of deep brain stimulation in a computational model of Parkinson’s disease. PLoS Comput Biol 14, e1006606 (2018).

[0207] 25. Oliver, L. D. et al. Bayesian Optimization Of NeuroStimulation (BOONStim). BrainStimulation 18, 1 12-115 (2025).

[0208] 26. Lorenz, R. et al. Efficiently searching through large tACS parameter spaces using closed-loop Bayesian optimization. Brain Stimulation 12, 1484-1489 (2019).

[0209] 27. Wernisch, L. et al. Online Bayesian optimization of vagus nerve stimulation. J. NeuralEng. 21 , 026019 (2024).

[0210] 28. Aiello, G., Valle, G. & Raspopovic, S. Recalibration of neuromodulation parameters in neural implants with adaptive Bayesian optimization. J. Neural Eng. 20, 026037 (2023).

[0211] 29. Laferriere, S., Bonizzato, M., Cote, S. L., Dancause, N. & Lajoie, G. HierarchicalBayesian Optimization of Spatiotemporal Neurostimulations for Targeted Motor Outputs. IEEE Trans. Neural Syst. RehabiL Eng. 28, 1452-1460 (2020).

[0212] 30. Losanno, E. et al. Bayesian optimization of peripheral intraneural stimulation protocols to evoke distal limb movements. J. Neural Eng. 18, 066046 (2021 ).

[0213] 31. Ament, S., Daulton, S., Eriksson, D., Balandat, M. & Bakshy, E. UnexpectedImprovements to Expected Improvement for Bayesian Optimization. Preprint at doi.org / 10.48550 / arXiv.2310.20708 (2025).

[0214] 32. Letham, B., Karrer, B., Ottoni, G. & Bakshy, E. Constrained Bayesian Optimization with Noisy Experiments. Bayesian Anal. 14, (2019).

[0215] 33. Wang, Z. & Jegelka, S. Max-value Entropy Search for Efficient Bayesian Optimization.Preprint at doi.org / 10.48550 / arXiv.1703.01968 (2018).

[0216] 34. Hvarfner, C., Hutter, F. & Nardi, L. Joint Entropy Search for Maximally-InformedBayesian Optimization. Preprint at doi.org / 10.48550 / arXiv.2206.04771 (2023).

[0217] 35. Rodriguez, A. et al. ToxTrac: A fast and robust software for tracking organisms.Methods Ecol Evol 9, 460-464 (2018).

[0218] 36. Balandat, M. et al. BoTorch: A Framework for Efficient Monte-Carlo BayesianOptimization. Preprint at doi.org / 10.48550 / arXiv.1910.06403 (2020).

[0219] 37. Hvarfner, C., Hellsten, E. O. & Nardi, L. Vanilla Bayesian Optimization Performs Great in High Dimensions. Preprint at doi.org / 10.48550 / arXiv.2402.02229 (2024).Example 2Stopping Criteria for Bayesian OptimizationImportance of efficient stopping in real-world settings

[0220] In real-world applications of Bayesian Optimization (BO), particularly in domains involving patients or other costly and sensitive contexts, the optimization budget is extremely limited. Each evaluation in a neuromodulation setting comes with significant cost, time, and ethical considerations. It is therefore of paramount importance to use this budget efficiently, maximizing the chance of identifying high-performing parameters within as few iterations as possible.The challenge of deciding when to stop the optimization process

[0221] Ideally, BO would be stopped precisely at the point when the optimum has been reached. However, in practice this is rarely achievable: the surrogate model provides only an estimate of the underlying objective, and the true optimum cannot be known with certainty. Stopping too early risks missing better solutions, while stopping too late increases the optimization budget unnecessarily. As such, stopping criteria are critical to balance efficiency with performance.Simple Stopping Strategies

[0222] The simplest strategies for stopping Bayesian Optimization include:Fixed budget: predetermining a maximum number of evaluations and stopping when this budget is exhausted.Predetermined target: stopping when the optimization achieves a predefined threshold of performance

[0223] These strategies are straightforward but often inefficient, since they do not adapt to the actual progress of the optimization process.More Advanced Strategies

[0224] Several more adaptive and efficient strategies for stopping have been proposed in the literature. Some examples include:

[0225] Acquisition function convergence-based rules: stop when the acquisition function values converge below a pre-determined threshold over multiple iterations.

[0226] Uncertainty reduction: stop when the posterior variance around the estimated optimum becomes sufficiently small, indicating the model is confident in its prediction.

[0227] Cross-validation or bootstrapping approaches: stop when repeated re-sampling shows stability of the identified optimum.

[0228] These strategies provide a better efficiency compared to fixed budgets or thresholds, but they still have limitations, especially under noisy conditions.Probabilistic Regret Bound (PRB)

[0229] PRB explicitly formalizes the decision to stop as a statistical test: at each iteration, the method estimates the probability that further evaluations could reduce the regret below a given threshold. In practice, PRB works by:Defining a regret threshold that represents an acceptable gap to the true optimum, with a corresponding acceptable confidence level.Using the surrogate model to estimate the probability that this threshold can be surpassed with additional evaluations.Stopping when this probability falls below the chosen confidence level.PRB relies on Monte Carlo sampling of the surrogate posterior to estimate the distribution of regret across the domain. For each posterior sample, the best achievable value is compared with the current incumbent, yielding a probability of satisfying the regret threshold. This sampling-based approach accommodates noise, making PRB a flexible and robust stopping criterion for neuromodulation parameter optimization in real-world scenarios.

Claims

What is claimed is:1 . A method for treating a neurological or neuropsychiatric disorder in a subject, the method comprising:(a) administering ultrasound neuromodulation to the subject using an initial selected set of ultrasound stimulation parameters;(b) measuring a response of the subject to said administering the ultrasound neuromodulation using the initial selected set of ultrasound stimulation parameters, wherein the measured response is used to calculate an efficacy score for treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation using the initial selected set of ultrasound stimulation parameters;(c) using the efficacy score from said administering the ultrasound neuromodulation using the initial selected set of ultrasound stimulation parameters to generate a Gaussian process regression model of an objective function to simulate responses of the subject to ultrasound neuromodulation administered using other ultrasound stimulation parameters;(d) performing Bayesian optimization with the Gaussian process regression model of the objective function using an acquisition function that selects a new set of ultrasound stimulation parameters with a goal of improving efficacy of treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation;(e) administering ultrasound neuromodulation to the subject using the new set of ultrasound stimulation parameters;(f) measuring a response of the subject to said administering the ultrasound neuromodulation using the new set of ultrasound stimulation parameters;(g) calculating an efficacy score for treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation using the new set of ultrasound stimulation parameters;(h) updating the Gaussian process regression model of the objective function with the efficacy score for the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation using the new set of ultrasound stimulation parameters, wherein the Gaussian process regression model of the objective function uses all the efficacy scores that have been calculated for the responses to all previous administrations of the ultrasound neuromodulation to the subject with previously selected ultrasound stimulation parameters;(i) performing Bayesian optimization with the updated Gaussian process regression model of the objective function, wherein the acquisition function selects another new set of ultrasoundstimulation parameters with the goal of further improving the efficacy of the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation;(j) repeating steps (e)-(i) iteratively to further optimize the ultrasound stimulation parameters until a stopping criterion is met, wherein the acquisition function selects a final set of optimized ultrasound stimulation parameters; and(k) administering ultrasound neuromodulation to the subject using the final set of optimized ultrasound stimulation parameters.

2. The method of claim 1 , wherein the stopping criterion is determined by a strategy based on probabilistic regret bounds, wherein the probabilistic regret bound statistical test comprises: defining a regret threshold; estimating probability that the regret threshold can be surpassed by repeating steps (e)-(i) after each iteration of step (j); discontinuing the repeating of steps (e)-(i) when the estimated probability that the regret threshold can be surpassed falls below a chosen confidence level.

3. The method of claim 1 , wherein the stopping criterion is based on acquisition function convergence, wherein steps (e)-(i) are repeated until the acquisition function values converge below a pre-determined threshold.

4. The method of claim 1 , wherein the stopping criterion is based on uncertainty reduction, wherein steps (e)-(i) are repeated until posterior variance around an estimated optimum is below a pre-determined threshold.

5. The method of claim 1 , further comprising performing cross-validation or bootstrapping, wherein the stopping criterion is based on stability of an identified optimum, wherein steps (e)-(i) are repeated until re-sampling shows stability of the identified optimum.

6. The method of any one of claims 1 -5, wherein steps (e)-(i) are repeated until convergence to an optimized target brain region and ultrasound stimulation parameters that are effective for treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation.

7. The method of claim 1 , wherein the stopping criterion is that further Bayesian optimization of the ultrasound stimulation parameters no longer results in further improvement of the efficacy of the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation.

8. The method of claim 1 , wherein the stopping criterion is based on a predetermined optimization budget.

9. The method of any one of claims 1 -8, wherein the acquisition function is a noisy expected improvement (qNEI) acquisition function, a max-value entropy search (MES) acquisition function, or a joint entropy search (JES) acquisition function.

10. The method of any one of claims 1 -9, further comprising optimizing the parameters of the Gaussian process and of the acquisition function based on a model of the noise present in the data.

11. The method of claim 10, wherein Gaussian noise is modeled from a normal distribution N(0, o), wherein a is equal to a selected noise fraction.

12. The method of any one of claims 1 -11 , wherein the initial selected set of ultrasound stimulation parameters is selected using random sampling, Sobol sequence sampling, a pseudorandom distribution, or Voronoi parcellation.

13. The method of any one of claims 1 -12, further comprising repeating steps (a)-(k) at a later time to further optimize the ultrasound stimulation parameters for the subject at the later time.

14. The method of claim 13, wherein disease progression in the subject, a change in medication administered to the subject, or adaptation of the subject to the ultrasound neuromodulation have diminished the efficacy of the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation over time.

15. The method of claim 14, wherein a model of the objective function estimated from a group of previously treated subjects with a given neurological or neuropsychiatric disorder is used to initialize the Gaussian process regression.

16. The method of any one of claims 1 -15, wherein said performing Bayesian optimization comprises optimizing hyperparameters comprising the acquisition function, seed queries, and an exploration / exploitation ratio.

17. The method of any one of claims 1 -16, wherein said performing Bayesian optimization comprises optimizing at least 2, at least 3, or at least 4 ultrasound stimulation parameters selected from a target brain region, a stimulation duration, an acoustic intensity, an acoustic pressure, an excitation voltage, a pulse repetition frequency, a pulse length, an ultrasound frequency, and a duty cycle.

18. The method of claim 17, wherein the ultrasound stimulation parameters comprise the target brain region, the stimulation duration, and the excitation voltage.

19. The method of claim 18, wherein the ultrasound stimulation parameters comprise the target brain region, the stimulation duration, the acoustic intensity, the acoustic pressure, the excitation voltage, the pulse repetition frequency, the pulse length, the ultrasound frequency, and the duty cycle.

20. The method of any one of claims 1 -19, wherein the response is a body movement, a facial movement, a change in locomotor activity, a change in mood, or a change in brain electrical activity.21 . The method of claim 20, wherein the body movement, the facial movement, or the change in locomotor activity is measured using an accelerometer, a gyroscope, or a video recording device that records images of the subject.

22. The method of claim 21 , wherein the accelerometer or the gyroscope is provided by a wearable device.

23. The method of claim 22, wherein the wearable device is a smartwatch.

24. The method of claim 21 , wherein the video recording device is provided by a digital camera, a smartphone, a tablet, a laptop, or a camcorder.

25. The method of claim 20, wherein the brain electrical activity is measured by electroencephalography (EEG), stereoelectroencephalography (sEEG), electrocorticography (ECoG), magnetoencephalography (MEG), single photon emission computed tomography (SPECT), or functional magnetic resonance imaging (fMRI).

26. The method of any one of claims 1 -25, wherein the neurological disorder is a movement disorder, and wherein the ultrasound stimulation parameters are optimized to ameliorate a symptom of the movement disorder.

27. The method of claim 26, wherein the symptom is bradykinesia, dyskinesia, or an abnormal or involuntary facial movement.

28. The method of any one of claims 1-25, wherein the neuropsychiatric disorder is depression or anxiety, and wherein the ultrasound stimulation parameters are optimized to ameliorate the depression or the anxiety.

29. The method of claim 28, wherein the response is measured using a Visual Analogue Scale for Depression (VAS-D), a Hamilton Depression Rating Scale (HAM-D), Montgomery-Asberg Depression Rating Scale (MADRS), a Beck Depression Inventories (BDI) score, a Visual Analogue Scale for Anxiety (VAS-A) or a Beck Anxiety Inventories (BAI) score.

30. The method of claim 28, wherein the response is measured based on facial expression detection or facial emotion recognition analysis of a video recording of the subject’s face.

31. The method of any one of claims 1 -25, wherein the neurological disorder is neuropathic pain, and wherein the ultrasound stimulation parameters are optimized to ameliorate the neuropathic pain.

32. The method of claim 31 , wherein the response is measured using a numerical rating scale (NRS), a visual analog scale (VAS), or a categorical scale. In some embodiments, a Wong- Baker Faces Pain Scale, a FLACC Pain Scale, a CRIES Pain Scale, COMFORT Pain Scale, a McGill Pain Questionnaire, a Color Analog Pain Scale, Mankoski Pain Scale, a Brief Pain Inventory, or a Descriptor Differential Scale of Pain Intensity, or a combination thereof is used to evaluate pain.

33. The method of any one of claims 1 -25, wherein the neurological disorder is a sleepwake disorder, and wherein the ultrasound stimulation parameters are optimized to ameliorate symptoms of the sleep-wake disorder.

34. The method of claim 33, wherein the response is measured using a visual-analog scale (VAS), a psychomotor vigilance test (PVT), a Likert scale, a Stanford Sleepiness Scale (SSS), a maintenance of wakefulness test (MWT), an Epworth sleepiness scale (ESS), a multiple sleep latency test (MSLT), or an Athens insomnia scale, or by monitoring the subject using actigraphy, electroencephalography, or polysomnography.

35. The method of claim 33, wherein said administering the ultrasound neuromodulation increases arousal compared to in absence of said administering the ultrasound neuromodulation.

36. The method of any one of claims 1 -35, wherein the ultrasound neuromodulation is administered with a portable ultrasound transducer or a wearable ultrasound array.

37. The method of any one of claims 1 -36, wherein the ultrasound neuromodulation is administered with a transcranial focused ultrasound (FUS) transducer, phased array transducer, capacitive micromachined ultrasound transducer (CMUT), piezoelectric micromachined ultrasonic transducer (pMUT), or miniaturized half-concave transducer.

38. A computer-implemented method for programming an ultrasound transducer to treat a subject for a neurological or neuropsychiatric disorder with ultrasound neuromodulation, the computer performing steps comprising:(a) instructing the ultrasound transducer to deliver ultrasound neuromodulation to the subject using an initial selected set of ultrasound stimulation parameters;(b) receiving experimental data from measuring a response of the subject to administering the ultrasound neuromodulation using the initial selected set of ultrasound stimulation parameters, wherein the measured response is used to calculate an efficacy score for treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation using the initial selected set of ultrasound stimulation parameters;(c) using the efficacy score from said administering the ultrasound neuromodulation using the initial selected set of ultrasound stimulation parameters to generate a Gaussian processregression model of an objective function to simulate responses of the subject to ultrasound neuromodulation administered using other ultrasound stimulation parameters;(d) performing Bayesian optimization with the Gaussian process regression model of the objective function using an acquisition function that selects a new set of ultrasound stimulation parameters with the goal of improving the efficacy of treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation;(e) instructing the ultrasound transducer to deliver ultrasound neuromodulation to the subject using the new set of ultrasound stimulation parameters;(f) receiving experimental data from measuring a response of the subject to administering the ultrasound neuromodulation using the new set of ultrasound stimulation parameters;(g) calculating an efficacy score for treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation using the new set of ultrasound stimulation parameters;(h) updating the Gaussian process regression model of the objective function with the efficacy score for the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation using the new set of ultrasound stimulation parameters, wherein the Gaussian process regression model of the objective function uses all the efficacy scores that have been calculated for the responses to all previous administrations of the ultrasound neuromodulation to the subject with previously selected ultrasound stimulation parameters;(i) performing Bayesian optimization with the updated Gaussian process regression model of the objective function, wherein the acquisition function selects another new set of ultrasound stimulation parameters with the goal of further improving the efficacy of the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation;(j) repeating steps (e)-(i) iteratively to further optimize the ultrasound stimulation parameters until a stopping criterion is met, wherein the acquisition function selects a final set of optimized ultrasound stimulation parameters; and(k) instructing the ultrasound transducer to deliver ultrasound neuromodulation to the subject using the final set of optimized ultrasound stimulation parameters.

39. The computer-implemented method of claim 38, wherein the stopping criterion is determined by a strategy based on probabilistic regret bounds, wherein the probabilistic regret bound statistical test comprises: defining a regret threshold; estimating probability that the regret threshold can be surpassed by repeating steps (e)-(i) after each iteration of step (j);discontinuing the repeating of steps (e)-(i) when the estimated probability that the regret threshold can be surpassed falls below a chosen confidence level.

40. The computer-implemented method of claim 38, wherein the stopping criterion is based on acquisition function convergence, wherein steps (e)-(i) are repeated until the acquisition function values converge below a pre-determined threshold.41 . The computer-implemented method of claim 38, wherein the stopping criterion is based on uncertainty reduction, wherein steps (e)-(i) are repeated until posterior variance around an estimated optimum is below a pre-determined threshold.

42. The computer-implemented method of claim 38, further comprising performing cross- validation or bootstrapping, wherein the stopping criterion is based on stability of an identified optimum, wherein steps (e)-(i) are repeated until re-sampling shows stability of the identified optimum.

43. The computer-implemented method of any one of claims 38-42, wherein steps (e)-(i) are repeated until convergence to an optimized target brain region and ultrasound stimulation parameters that are effective for treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation.

44. The computer-implemented method of claim 38, wherein the stopping criterion is that further Bayesian optimization of the ultrasound stimulation parameters no longer results in further improvement of the efficacy of the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation.

45. The computer-implemented method of claim 38, wherein the stopping criterion is based on a predetermined optimization budget.

46. The computer-implemented method of any one of claims 38-45, wherein the acquisition function is a noisy expected improvement (qNEI) acquisition function, a max-value entropy search (MES) acquisition function, or a joint entropy search (JES) acquisition function.

47. The computer-implemented method of any one of claims 38-46, further comprising optimizing the parameters of the Gaussian process and of the acquisition function based on a model of the noise present in the data.

48. The computer-implemented method of claim 47, wherein Gaussian noise is modeled from a normal distribution N(0, o), wherein a is equal to a selected noise fraction.

49. The computer-implemented method of any one of claims 38-48, wherein the initial selected set of ultrasound stimulation parameters is selected using random sampling, Sobol sequence sampling, a pseudo-random distribution, or Voronoi parcellation.

50. The computer-implemented method of any one of claims 38-49, further comprising repeating steps (a)-(k) at a later time to further optimize the ultrasound stimulation parameters for the subject at the later time.51 . The computer-implemented method of claim 50, wherein disease progression in the subject, a change in medication administered to the subject, or adaptation of the subject to the ultrasound neuromodulation have diminished the efficacy of the treatment of the neurological or neuropsychiatric disorder with the ultrasound neuromodulation over time.

52. The computer-implemented method of claim 51 , wherein a model of the objective function estimated from a group of previously treated subjects with a given neurological or neuropsychiatric disorder is used to initialize the Gaussian process regression.

53. The computer-implemented method of any one of claims 38-52, wherein said performing Bayesian optimization comprises optimizing hyperparameters comprising the acquisition function, seed queries, and an exploration / exploitation ratio.

54. The computer-implemented method of any one of claims 38-53, wherein the ultrasound stimulation parameters comprise at least 2, at least 3, or at least 4 ultrasound stimulation parameters selected from a target brain region, a stimulation duration, an acoustic intensity, an acoustic pressure, an excitation voltage, a pulse repetition frequency, a pulse length, an ultrasound frequency, and a duty cycle.

55. The computer-implemented method of claim 54, wherein the ultrasound stimulation parameters comprise the target brain region, the stimulation duration, and the excitation voltage.

56. The computer-implemented method of claim 55, wherein the ultrasound stimulation parameters comprise the target brain region, the stimulation duration, the acoustic intensity, the acoustic pressure, the excitation voltage, the pulse repetition frequency, the pulse length, the ultrasound frequency, and the duty cycle.

57. The computer-implemented method of any one of claims 38-56, wherein the response is a body movement, a facial movement, a change in locomotor activity, a change in mood, or a change in brain electrical activity.

58. The computer-implemented method of any one of claims 38-57, further comprising displaying the optimized set of ultrasound stimulation parameters.

59. The computer-implemented method of any one of claims 38-58, further comprising displaying a user interface presenting a questionnaire configured to receive input from the subject regarding self-reported results of treatment with the ultrasound neuromodulation.

60. The computer-implemented method of claim 59, wherein the questionnaire uses a Movement Disorder Society-Sponsored Revision of the Unified Parkinson's Disease Rating Scale (MDS-UPDRS), a Hoehn and Yahr (HnY) scale, a Clinical Rating Scale for Tremor (CRST), a Parkinson’s Disease Composite Scale (PDCS), or a Schwab and England Activities of Daily Living (ADL) Scale for assessing treatment of a movement disorder.61 . The computer-implemented method of claim 59, wherein the questionnaire uses a Visual Analogue Scale for Depression (VAS-D), a Hamilton Depression Rating Scale (HAM-D), Montgomery-Asberg Depression Rating Scale (MADRS), a Beck Depression Inventories (BDI) score, a Visual Analogue Scale for Anxiety (VAS-A) or a Beck Anxiety Inventories (BAI) score for assessing treatment of depression or anxiety.

62. The computer-implemented method of claim 59, wherein the questionnaire uses a numerical rating scale (NRS), a visual analog scale (VAS), or a categorical scale. In someembodiments, a Wong-Baker Faces Pain Scale, a FLACC Pain Scale, a CRIES Pain Scale, COMFORT Pain Scale, a McGill Pain Questionnaire, a Color Analog Pain Scale, Mankoski Pain Scale, a Brief Pain Inventory, or a Descriptor Differential Scale of Pain Intensity, or a combination thereof to evaluate pain.

63. The computer-implemented method of claim 59, wherein the questionnaire uses a visual-analog scale (VAS), a psychomotor vigilance test (PVT), a Likert scale, a Stanford Sleepiness Scale (SSS), a maintenance of wakefulness test (MWT), an Epworth sleepiness scale (ESS), a multiple sleep latency test (MSLT), or an Athens insomnia scale to evaluate level of wakefulness, arousal, or sleepiness.

64. A non-transitory computer-readable medium comprising program instructions that, when executed by a processor in a computer, causes the processor to perform the method of any one of claims 38-63.

65. A kit comprising the non-transitory computer-readable medium of claim 64 and instructions for determining symptom severity of a subject having a movement disorder.

66. A system for treating a neurological or neuropsychiatric disorder with ultrasound neuromodulation, the system comprising: an ultrasound transducer; and a processor programmed according to the computer-implemented method of any one of claims 38-63 to instruct the ultrasound transducer to deliver ultrasound neuromodulation to the subject using an optimized set of ultrasound stimulation parameters.

67. The system of claim 66, further comprising an accelerometer, a gyroscope, a video recording device, an electroencephalography (EEG) electrode, or a combination thereof.

68. The system of claim 67, wherein the accelerometer, the gyroscope, or the combination thereof is provided by a wearable device.

69. The system of claim 68, wherein the wearable device is a smartwatch.

70. The system of claim 67, wherein the video recording device is provided by a digital camera, a smartphone, a tablet, a laptop, or a camcorder.71 . The system of any one of claims 66-70, wherein the ultrasound transducer is portable or a wearable ultrasound array.

72. The system of any one of claims 66-71 , wherein the ultrasound transducer is a transcranial focused ultrasound (FUS) transducer, phased array transducer, capacitive micromachined ultrasound transducer (CMUT), piezoelectric micromachined ultrasonic transducer (pMUT), or miniaturized half-concave transducer.

73. The system of any one of claims 66-72, wherein the system further comprises a user interface comprising an input electronically coupled to the processor for instructing the ultrasound transducer to deliver ultrasound neuromodulation to the subject to treat the neurological or neuropsychiatric disorder.

74. The system of claim 73, wherein the user interface is password protected and is operable by a health care practitioner.

75. The system of any one of claims 66-74, further comprising a storage component for storing data, wherein the storage component is coupled to the processor.

76. The system of any one of claims 66-75, further comprising a display for displaying the optimized set of ultrasound stimulation parameters.

77. The system of claim 76, wherein the display further displays a user interface presenting a questionnaire configured to receive input from the subject regarding self-reported results of treatment with the ultrasound neuromodulation.

Images