Systems and methods for determining targets for noninvasive brain therapies based on optical images of the head

Abstract

A method and a system are provided for determining a location of a brain target for neuromodulation therapy without requiring MRI imaging of the head. The system includes a sensor or a plurality of sensors configured to acquire data describing a geometry of a head of a subject. A processor is coupled to the sensor and is configured to receive the data describing a geometry of the head of the subject gathered by the sensor, determine 3D coordinates of head-surface contour points from the acquired data, apply a predictive model to the determined 3D coordinates of head-surface contour points, receive 3D coordinates of a located brain target as output from the model, and output 3D coordinates of the located brain target for subsequent delivery of neuromodulatory brain therapy. The model is pre-trained on a dataset of head and brain images and configured to associate head-surface coordinates with 3D coordinates of at least one deep brain target.

Description

SYSTEMS AND METHODS FOR DETERMINING TARGETS FOR NONINVASIVE BRAIN THERAPIES BASED ON OPTICAL IMAGES OF THE HEADCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to United States Provisional Patent Application No. 63 / 741,387 filed on January 2, 2025, and titled “MRI FREE TARGETING OF DEEP BRAIN STRUCTURES IN THE ANTERIOR CINGULATE CORTEX USING FACIAL LANDMARKS,” United States Provisional Patent Application No. 63 / 789,031 filed April 15, 2025, and titled “METHODS AND SYSTEMS FOR DETERMINING TARGETS FOR NONINVASIVE BRAIN THERAPIES BASED ON OPTICAL IMAGES OF THE HEAD,” and United States Provisional Patent Application No. 63 / 884,849 filed September 19, 2025, and title “MRI-FREE TARGETING OF DEEP BRAIN REGIONS USING THE 10-20 SYSTEM,” each of which are hereby incorporated by reference in their entiretiesTECHNICAL FIELD

[0002] The present disclosure relates to systems and methods for determining targets for noninvasive brain therapies based on optical images of the head to perform noninvasive brain stimulation that does not require MRI or CT images of the headBRIEF DESCRIPTION OF THE DRAWINGS

[0003] The embodiments disclosed herein will become more fully apparent from the following description and appended claims, taken in conjunction with the accompanying drawings. The drawings depict only typical embodiments, which embodiments will be described with additional specificity and detail in connection with the drawings in which:

[0004] FIG. 1 A illustrates a targeting system for targeting and delivering therapeutic ultrasound to a target region in a head of a subject, according to embodiments described herein.

[0005] FIG. 1B illustrates a targeting system for targeting and delivering therapeutic ultrasound to a target region in a head of a subject, according to embodiments described herein.

[0006] FIG. 2 illustrates a schematic drawing of an ultrasound device capable of targeting and delivering ultrasonic energy to a brain target region in a head of a subject, according to embodiments described herein.

[0007] FIG. 3 illustrates a top view of an optical guidance system for extracting contour data of the head of a subject, according to embodiments described herein.

[0008] FIG. 4A illustrates a lateral view of a head of a subject and showcases a corresponding anterior-posterior midline, according to embodiments described herein.

[0009] FIG. 4B illustrates a frontal view of a head of a subject and showcases a corresponding left-right transverse midline, according to embodiments described herein.

[0010] FIG. 5 illustrates an embodiment of a representation of contour data of a head of a subject, according to embodiments described herein

[0011] FIG. 6 illustrates an embodiment of target coordinates corresponding to three-dimensional locations of deep brain regions as produced by the targeting model, according to embodiments described herein.14910-4258-31682DETAILED DESCRIPTION

[0012] Mental and neurological disorders affect nearly one fifth of the world's population. Approximately one third of patients across mental and neurological conditions are treatmentresistant. Neuromodulation has the potential to provide a targeted reset of the malfunctioning circuits, but current state-of-the-art approaches, detailed below, have significant limitations. These limitations leave millions of patients in the United States and worldwide not adequately treated.

[0013] These disorders typically involve neural networks situated deep in the brain, including limbic, basal ganglia, memory, and brain stem networks. Progress in treatments of these has been hampered by the lack of tools to effectively and safely modulate and reset these circuits. Deep brain stimulation (DBS) has shown promise in providing a selective reset of the involved deep brain circuits, but the surgical implantation of stimulating leads is associated with high costs and risks, including brain hemorrhage, infection, and in some cases, death.

[0014] Deep brain neuromodulation demands precise spatial guidance. MRI-based neuronavigation provides this accuracy but imposes structural limits on scalability. MRI access, cost, and workflow complexity create barriers for broad deployment, repeated interventions, and outpatient or home-based neuromodulation These constraints prevent widespread use of precision targeting despite the therapeutic need.

[0015] Accordingly, systems and methods for brain target guidance that do not require MRI or CT images of the head are provided herein. The approach includes optical acquisition of headsurface geometry and ensuing, model-based prediction of internal coordinates. Three input strategies are supported: dense head-contour geometry, 10-20 EEG landmarks (midline-only or expanded sets), and sparse facial fiducials. Each strategy supplies 3D surface coordinates to a targeting model trained on about 430 MRI volumes with scalp segmentation and ground-truth target coordinates for deep brain regions.

[0016] Optical systems capture the required surface coordinates using single- or multi-view imaging and / or optical tracking. Extracted points include dense contours, canonical 10-20 landmarks, or facial fiducials, which are then converted to 3D geometry and fed into MRI-trained models that map surface structure to deep-brain targets in AC-PC space. Resulting coordinates drive ultrasound or other noninvasive neuromodulation devices, with therapeutic focal volumes selected to exceed residual targeting error.

[0017] In some embodiments, a facial-fiducial approach uses nine craniofacial points to drive a global linear template transform. Such an approach can yield about 5.7 mm error for deep brain targets such as the anterior cingulate and commissures

[0018] In other embodiments, a 10-20-based approach can be used. In some cases, a midline-only model trained is trained on seven 10-20 points to achieve about 5 mm accuracy and retain stability under several millimeters of landmark jitter. In still other embodiments, an expanded 25-landmark 10-20 representation further improves precision. Linear regression can in such an approach produces about 3.9 ± 1.7 mm error for cingulate subregions and about 4 924910-4258-31682± 2.5 mm globally. In some cases, such a model outperforms both template-registration methods and neural network variants.

[0019] In some embodiments, the provided approach includes a model trained to fit a template brain to anatomical landmarks of the head. Using facial anatomical landmarks, the approach can achieve about 5.8 ± 3.0 mm accuracy for deep brain targets including the cingulate cortex and the anterior and posterior commissures.

[0020] The phrases “coupled to” and “in communication with” refer to any form of interaction between two or more entities, including mechanical, electrical, magnetic, electromagnetic, fluid, and thermal interaction. Two components may be coupled to or in communication with each other even though they are not in direct contact with each other. For example, two components may be coupled to or in communication with each other through an intermediate component.

[0021] Embodiments may be understood by reference to the drawings, wherein like parts are designated by like numerals throughout. It will be understood by one of ordinary skill in the art having the benefit of this disclosure that the components of the embodiments, as generally described and illustrated in the figures herein, could be arranged and designed in a wide variety of different configurations Thus, the following, more detailed description of various embodiments, as represented in the figures, is not intended to limit the scope of the disclosure but is merely representative of various embodiments. While the various aspects of the embodiments are presented in drawings, the drawings are not necessarily drawn to scale unless specifically indicated.

[0022] It will be appreciated that various features are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure. Many of these features may be used alone and / or in combination with one another.

[0023] In many embodiments, an ultrasound system (e.g., ultrasound-based neuromodulation system) for at-home or outpatient use is disclosed. An ultrasound system may comprise a controller and one or more transducers operably coupled to the controller and configured to selectively generate ultrasonic waves at a low frequency (e.g., about 100 kHz to about 650 kHz) effective to stimulate a target region of a body of a subject without ablating the target region. The ultrasonic waves generated by the one or more transducers have a focal width that is at least about three millimeters and / or larger than the target region of the body of the subject. In some embodiments, the focal width may be determined solely in relation to the size of the target area, embodiments where there is no absolute minimize size of the focal width are within the scope of this disclosure. For example, embodiments wherein the target region is at least two times larger than the target region, regardless of the absolute size of the focal width are within the scope of this disclosure. Similar focal widths at least three times, at least four times, at least 1.5 times, or greater than 5 times larger than the target region are all within the scope of this disclosure

[0024] As used herein, “stimulation” of the brain target region (or other target regions) may include a modulation of activity of excitable cells, such as neurons, glial cells, pancreatic cells, or other cell types that are responsive to the mechanical pressure waves associated with34910-4258-31682ultrasound. As used herein, “stimulation” is broad enough to include delivery of mechanical pressure waves at any degree, energy level, or amount configured to induce a therapeutic response from the target cell.

[0025] In some embodiments, deep brain target coordinates are expressed in the anterior-commissure / posterior-commissure (AC-PC) coordinate system. In this reference frame, the X-axis is defined as the normal vector of the midline plane, the Y-axis spans the posterior commissure to the anterior commissure, and the Z-axis is taken as the cross-product with positive orientation along the ventral-dorsal direction. Expressing predictions in AC-PC coordinates enables integration with neuromodulation systems and allows consistent comparison of targeting error across subjects.

[0026] FIG. 1 A illustrates a targeting system 160 for targeting and delivering therapeutic ultrasound to a target region in a head of a subject, according to embodiments described herein. The application is not limited to the specific targeting system outlined in FIG. 1 A as other targeting systems with more or less components than illustrated in FIG. 1A may be used. As seen in the illustrated embodiment, an optical guidance system 120 can be applied to a head 10 of a subject to extract contour data 162, or data reflective of a 3D geometric shape of the head 10 of the subject. FIG. 1B illustrates another embodiment of the targeting system 160 for targeting and delivering therapeutic energy using a neuromodulation device 100’ to a target region in the head 10 of the subject, according to embodiments described herein. In the embodiment of FIG 1B, the optical guidance system 120 uses head landmarks 162’, discussed below.

[0027] In some cases, contour data 162 extracted by optical guidance system 120 can be or include a dense head-contour geometry, 10-20 EEG anatomical landmarks (limited or extended), or a sparse set of facial fiducials. For instance, in some embodiments, the optical guidance system 120 includes one or more image sensors (e.g., such as image sensor 130 as will be described with respect to FIG. 3) that is / are configured to acquire one (or more) two-dimensional image(s) of the head 10. In embodiments where a single image is gathered, anatomical landmarks or surface features can be identified within the acquired images, and corresponding landmark coordinates are derived in an image-based reference frame.

[0028] Although in some implementations only a single two-dimensional image may be taken, in alternate embodiments, the optical guidance system 120 can acquire a plurality of two-dimensional images, e.g., from different viewpoints using an image sensor (e.g., such as image sensor 130 as will be described with respect to FIG. 3) and reconstruct three-dimensional landmark coordinates or surface geometry via photogrammetric reconstruction. The resulting three-dimensional representation is then stored as contour data 162.

[0029] In some embodiments, the optical guidance system 120 can include or be operably coupled to an optical digitizing device, or an optical tracker (e g , in place of an image sensor) configured to directly measure three-dimensional coordinates of head landmarks. The optical digitizer device can track the spatial position of a probe or marker relative to a calibrated44910-4258-31682reference frame, thus enabling direct acquisition of landmark coordinates without image reconstruction.

[0030] Accordingly, acquisition of head-surface coordinates may be performed using multiple alternative modalities, such as (i) two-dimensional imaging in which landmark coordinates are extracted from one or more 2D images; (ii) multi-view imaging in which a plurality of 2D images are combined to reconstruct three-dimensional landmark coordinates or surface geometry using photogrammetry; and (iii) direct three-dimensional digitization using an optical tracking or digitizing device. Any of these acquisition modalities may be used with dense contours, facial fiducials, and / or EEG landmark sets.

[0031] In some embodiments, optical guidance system 120 can be omitted and / or replaced and contour data 162 is obtained via manual measurement of anatomical landmarks (e.g., in some cases the 10-20 EEG anatomical landmarks, whether they be limited or extended) or the sparse set of facial fiducials (e.g., using an optical tracking device). For example, in some cases, an operator may manually measure 10-20 or 10-10 EEG landmarks (midline-only or expanded sets) or craniofacial fiducials using tape-based proportional measurements, calipers, and / or a tracked probe. The resulting contour data 162 (e.g., three-dimensional coordinates) may be entered into the targeting model 128.

[0032] In some embodiments, contour data 162 can include or be a dense three-dimensional head-surface contour of subject’s head 10. The head-surface contour can represent the global cranial geometry and may include hundreds or even thousands of sampled surface points acquired over the scalp. Such dense contour geometry provides a high-information representation of head shape to enable accurate inference of internal brain coordinates, without reliance on sparser sets of predefined anatomical landmarks.

[0033] Building on the above, in some embodiments, contour data 162 are acquired and used without identification of discrete anatomical fiducials. For example, contour data 162 may comprise a head silhouette, a dense point cloud, or a sampled three-dimensional scalp surface representing global cranial geometry. In such embodiments, the contour data encode curvature, dimensions, and spatial relationships across the head that are sufficient to support MRI-free estimation of intracranial target coordinates, either via regression-based models or surfaceregistration to a template. Thus the system may forgo manual or automated labeling of individual landmarks.

[0034] In alternate embodiments, contour data 162 can include a sparser set of anatomically defined landmarks, including, for instance, the nasion, left and right tragus, and medial or lateral canthi of the eyes, and so on. These landmarks cam be extracted by optical guidance system 120 and provide a condensed geometric representation sufficient to constrain spatial mappings between external head geometry and deep brain targets.

[0035] In some embodiments, facial landmarks can be selected based on anatomical salience, ease of identification, and reproducibility across subjects. Accordingly, landmarks such as the nasion, tragus, and ocular canthi can provide stable geometric references that constrain head scale, orientation, and facial-to-cranial alignment, while minimizing operator-dependent error or54910-4258-31682variability. Such landmarks thus enable signal-rich geometric representations that remain informative for MRI-free targeting of deep brain regions

[0036] In further embodiments, the contour data 162 includes anatomical landmarks defined according to the international 10-20 or 10-10 EEG system. For instance, in some cases, optical guidance system 120 may extract a midline-only subset or an expanded multi-point representation distributed across the scalp. These standardized landmarks are provided to targeting model 128 as structured surface-coordinate inputs.

[0037] From the contour data 162, a pre-trained, targeting model 128 can be used to predict or determine target coordinates 164 for specific, deep brain target regions. These target coordinates 164 can then be delivered to a controller of an ultrasound device 100 to deliver ultrasonic therapy to such deep brain targets using transcranial focused ultrasound. However, the specification is not limited to ultrasound device 100. As illustrated in FIG. 1 B, the neuromodulation device 100’ may be used to delivery various types of therapy. In some embodiments, transcranial focused ultrasound, Transcranial Magnetic Stimulation (TMS) or Transcranial Current Stimulation (tACS, tDCS) may be used instead of ultrasound.

[0038] In some cases, the optical guidance system 120 can gather contour data 162 from a camera or other image sensor. In this strategy, contour data 162 can include dense headcontour geometry and / or anatomical landmarks Otherwise stated, the global shape of the head can be captured using one or more captured images (or an optical tracking device).

[0039] The targeting model 128 is trained on a large MRI library to learn statistical associations between external contour geometry and internal target coordinates. The targeting model 128 can thus be used to generate target coordinates 164 via direct regression from optical landmarks to deep brain coordinates or fitting the optical contour to a standard MRI template and inferring targets from the template. Optical imaging of the head 10 can thus replace MRI by first extracting a three-dimensional surface shape and consequently mapping that contour to internal coordinates.

[0040] In certain embodiments, the targeting model 128 is trained on a large MRI library to learn the association between external contour geometry and internal target locations. The mapping can be linear, nonlinear, or implemented with a neural network.

[0041] In some embodiments, the targeting model 128 is trained and evaluated using subjectwise cross-validation to approximate performance on unseen heads. For example, in one configuration, a library of approximately 430 T1-weighted MRI volumes is used, and the model is trained and tested using cross-validation which leaves a subject out. For each held-out subject, the model is fit on all remaining subjects and then used to predict that subject’s deep-brain target coordinates from the corresponding head-surface landmarks. Targeting accuracy is quantified as the Euclidean distance between the predicted coordinates and ground-truth intracranial coordinates derived from MRI-based atlas registration Reported errors correspond to the average over all subjects and all repetitions of this procedure.

[0042] In alternate embodiments, the optical guidance system 120 can extract contour data 162 corresponding to anatomical landmarks according to the 10-20 system. In this strategy, the64910-4258-31682optical guidance system extracts anatomical landmarks consistent with the 10-20 or 10-10 EEG coordinate system. In some cases, this may include seven midline points and / or an expanded set of twenty-five canonical scalp points derived from MRI-based percentile distances.

[0043] In this configuration, the targeting model 128 can be a regression model (e.g., linear, nonlinear, or neural network based trained on a large MRI library of approximately 430 subjects. For instance, in this case, the targeting model 128 directly maps 10-20 surface coordinates to deep brain region target coordinates 164. When trained on the full twenty-five points, linear regression achieves roughly 3.9 mm accuracy for anterior cingulate subregions and approximately 4.9 mm accuracy globally. These results show that routine 10-20 measurements can support sub-4-mm MRI-free guidance.

[0044] In alternate embodiments, the optical guidance system 120 can extract contour data 162 corresponding to a sparser set of facial fiducials. For instance, in some cases, a landmarkbased, linear template method can replace MRI by relying on anatomical landmarks (e.g., 9 facial points) such as the nasion, left inner canthus (LIC), right inner canthus (RIC), left outer canthus (LOC), right outer canthus (ROC), left preauricular point (LPA), right preauricular point (RPA), left external acoustic meatus (LEM), right external acoustic meatus (REM).

[0045] In this configuration, the targeting model 128 can be a single, global, linear transform that computes the optimal scale, rotation, and translation required to align a template head to the subject. Applying the transform to the template brain can deliver predicted deep-brain target coordinates 164 without subject-specific MRI.

[0046] This method can achieve approximately 5.7 mm accuracy relative to MRI-based nonlinear registration, with five fiducials performing comparably to nine. Validated targets include at least the anterior cingulate cortex (ACC), the anterior commissure (AC), and the posterior commissure (PC). The transformed template brain can thus yield deep-brain coordinates without subject-specific imaging.

[0047] In additional or alternative embodiments, instead of predicting intracranial coordinates directly, the system can fit an acquired head contour to a standard MRI head template using a surface-matching transform. Once the head contour is aligned to the template surface, deepbrain targets are inferred directly from the standard template MRI These targets may include subregions of the anterior cingulate cortex, the subcallosal cingulate cortex, and the anterior corpus callosum. This template-fitting strategy provides an alternative MRI-free targeting pathway when full-surface contour data are available, enabling deep-brain coordinate estimation without reliance on landmark-only regression models.

[0048] In some embodiments, the fitting between the optical head contour and the MRI template head surface uses a nonlinear deformation field rather than a global similarity transform. The nonlinear mapping aligns local curvature features of the optical contour to corresponding regions of the template surface, after which target coordinates are inferred from the deformed template brain.

[0049] After target coordinates 164 are produced via the targeting model 128, a transducer frame is positioned so as to deliver therapeutic stimulation to a desired deep brain target region.74910-4258-31682

[0050] In additional or alternative embodiments, the predictive model can be or include a linear regression model configured to compute a mapping from head-surface coordinate inputs to intracranial target coordinates as outputs. The model can receive a representation of the three-dimensional head-surface contour points and applies a linear transformation, optionally estimated using a Moore-Penrose pseudo-inverse, to generate predicted coordinates of one or more deep brain targets.

[0051] In some embodiments, the linear model is trained using a dataset of magnetic resonance images for a plurality of subjects, wherein each MRI volume includes both scalpsurface geometry and ground-truth locations for deep brain structures. Training can include solving for regression coefficients that minimize a prediction error across the training set. These may be optionally subject to regularization and / or penalties. The resulting linear mapping thus can encode statistical associations between external cranial geometry and internal brain coordinates.

[0052] In some embodiments, training can include augmentation of head-surface inputs by adding random perturbations to landmark coordinates. Such an augmentation can improve a robustness of the linear model to uncertainty from surface-point extraction and measurement noise arising from optical imaging conditions. The trained model can thus retain stable performance under several millimeters of coordinate jitter

[0053] In some embodiments, separate linear regression models can be trained for distinct deep-brain targets. In others, a single multivariate linear model can simultaneously output multiple intracranial target coordinates. Outputs may include subregions of the anterior cingulate cortex, subcallosal cingulate cortex, commissural structures, or any other deep structure included in the training library.

[0054] Otherwise stated, in some embodiments, a single predictive model (e.g., a linear, nonlinear, or deep learning model) that was trained to associate the head landmarks to specific targets in the brain is configured to output coordinates for multiple deep brain targets from a single set of head-surface inputs. In alternate embodiments separate models are trained for respective targets or target subregions. This enables selection among alternative neuromodulatory targets without requiring additional imaging or re-acquisition of head-surface geometry.

[0055] In additional or alternative embodiments, the predictive model can be or include a neural network configured to map three-dimensional head-surface coordinate inputs to three-dimensional coordinates of a deep brain target as outputs. In some cases, the neural network includes one or more hidden layers, nonlinear activation functions, and / or fully connected or convolutional layers. For instance, in some applications, the network can receive a structured representation of surface geometry, including anatomical landmarks, 10-20 positions, or dense contour points as inputs

[0056] In particular embodiments, the neural network is trained on a dataset of MRI volumes with paired scalp-surface and deep-brain coordinate annotations. Ground-truth coordinates are obtained by registering each MRI volume to a standard brain template, and head-surface84910-4258-31682coordinates are derived from corresponding scalp segmentations. During training, the network parameters are then optimized to minimize a prediction error between the network output and the ground-truth coordinates.

[0057] In such template-based embodiments, a standard brain template (e.g., such as an average MNI-based brain template) in which anatomical structures and candidate neuromodulatory targets have known coordinates can be fitted into the contour or the set of head landmarks. Thus, once the subject-specific transform aligning the head contour to the template is determined, coordinates for any target defined within the template can be generated without training a specific or more narrowed landmark-to-target regression model. Otherwise stated, because the brain locations within the standard brain are known, fitting this brain within the head landmarks readily provides the coordinates of all brain targets, without having to train a model.

[0058] The training process for the network may include regularization techniques such as dropout, weight decay, or batch normalization. Training may also include data augmentation, such as random perturbations of head-surface coordinates in spatial dimensions, thus improving resilience to optical-sensor noise and landmark extraction error Augmentation may further include random scaling, rotation, or translation consistent with plausible variations in cranial geometry.

[0059] In some embodiments, the neural network can be configured to output multiple intracranial targets from a single set of head-surface inputs For example, in some instances, the network may simultaneously output predicted coordinates for several anterior cingulate subregions, commissural reference points, or additional deep nuclei used for neuromodulatory targeting.

[0060] Performance of the neural network may be evaluated using subject-wise cross-validation, leave-one-out validation, or train-validation-test splits across the MRI dataset.Metrics may include Euclidean error of predicted target coordinates, mean absolute error, and distance to other defined structures. Validation procedures can quantify generalization performance on unseen subjects and establish accuracy relative to MRI-based methods.

[0061] FIG. 2 illustrates a schematic drawing of an ultrasound device 100 capable of targeting and delivering ultrasonic energy to a brain target region 20a in a head 10 of a subject, according to embodiments described herein. In some embodiments, the ultrasound device 100 may be coupled to the head 10 at multiple points or regions. For example, the transducers 102 may be secured to an adjustable frame 103 that is configured to support the transducers 102 in a position that allows the head 10 of the subject to be disposed between the two transducers 102. The adjustable frame 103 may be configured to position the head 10 and / or the transducers 102 such that the transducers 102 provide maximal intensity of the ultrasonic waves 115 at the intended brain target region 20a

[0062] Ultrasound is emitted from one or two transducers toward the target such that the transducer and ultrasound beam axis 106 is perpendicular to the head symmetry plane 107 To account for modest inaccuracies in targeting, it can be additionally desirable that the therapeutic94910-4258-31682focal volume 119 of the tissue activated by the ultrasound is substantially (e.g., 2-3 times) broader than the volume occupied by the brain target region 20a.

[0063] The ultrasonic waves 115 may be delivered into the head 10 of the subject from the transducers 102 using a coupling medium 104. The coupling medium 104 may comprise any material that conducts the ultrasonic waves 115, such as a cryogel. In some embodiments described in greater detail below, the transducers 102 are configured to adjust to the head 10 of the subject such that the transducers 102 and / or the coupling medium 104 contact the head 10 of the subject. In some embodiments, the transducers 102 are selectively steerable and the controller 101 includes a steering control configured to steer the transducers 102 to direct the ultrasonic waves 115 at the brain target region 20a when the head 10 of the subject is positioned between the transducers 102. Thus, the ultrasound device 100 may provide the ability for an operator to steer the ultrasonic waves 115 into the brain target region 20a using the controller 101 (e.g., an electronic controller) coupled to the ultrasound device 100. In some embodiments, ultrasound aberrations by the head may be compensated for using an ultrasound through-transmit procedure described by Riis, et al. in the publication of “Controlled noninvasive modulation of deep brain regions in humans,” Communications Engineering, 3(1), 13 (2024), which is hereby incorporated by reference in its entirety.

[0064] The positioning of the transducers 102 allow the ultrasound device 100 and related methods of use to deliver ultrasonic waves 115 from one or more transducers 102 into specified deep brain target regions (e.g., the brain target region 20a) of the subject. The targeting of the ultrasonic waves 115 into specific brain regions (e.g., the brain target region 20a) for a given condition or disorder may be mediated using fixed transducer holders, such that the ultrasonic waves 115 are aimed specifically into the desired brain target region 20a.

[0065] In some embodiments, an ultrasound system of this disclosure may be configured to treat a condition of the brain including at least one of cognitive decline or Alzheimer’s disease, and the target region 20a may include one or more of a region of the brain associated with memory functions, a hippocampus of the brain, an entorhinal cortex of the brain, an amygdala of the brain, or a nucleus basalis of Meynert of the brain. In some embodiments, an ultrasound system of this disclosure may be configured to treat a condition of the brain including depression, and the target region 20a may include one or more of a cingulate cortex of the brain or a subcallosal cingulate cortex of the brain. In some embodiments, an ultrasound system of this disclosure may be configured to treat a condition of the brain including chronic pain, and the target region 20a may include one or more of an anterior cingulate cortex of the brain, a medial cingulate cortex of the brain, a subcallosal cingulate cortex, a ventral posterolateral nucleus, or a ventral posteromedial nucleus. In some embodiments, an ultrasound system of this disclosure may be configured to treat a condition of the brain including addiction, and the target region 20a may include one or more of a nucleus accumbens of the brain, a subcallosal cingulate cortex of the brain, or an anterior cingulate cortex of the brain. In some embodiments, an ultrasound system of this disclosure may be configured to treat a condition of the brain including food104910-4258-31682cravings, and the target region 20a may include one or more of a nucleus accumbens of the brain or a nucleus accumbens shell of the brain.

[0066] In some embodiments, an ultrasound system of this disclosure may be configured to treat a condition of the brain including anxiety, and the target region 20a may include one or more of an amygdala of the brain or a stria terminalis of the brain In some embodiments, an ultrasound system of this disclosure may be configured to treat a condition of the brain including post-traumatic stress brain disorder, and the target region 20a may include one or more of an amygdala of the brain or a bed nucleus of a stria terminalis of the brain.

[0067] The controller 101 may be configured to provide voltages of specific waveforms to the transducers 102. In some embodiments, the voltages and wave forms may be defined by the required stimulation parameters for the brain target region. Generally, low-intensity ultrasound should be safe and thus the stimulation parameters would ideally comply with the FDA 510k guidelines on safe ultrasound exposure, i e., not exceeding peak intensity of 190 W / cm2 and not exceeding time-average intensity of 0.72 W / cm2. The controller 101 can drive either a single channel (for single-element transducer) or multiple channels (for ultrasound arrays).

[0068] The controller 101 may be configured to implement any of the example methods disclosed herein. Moreover, the controller 101 may be configured to coordinate or otherwise direct the transducers 102 to emit the ultrasonic waves 115 at a selected frequency. The controller 101 may include at least one computing device configured to perform one or more of the acts described herein The at least one computing device of the controller 101 can include one or more servers, one or more computers (e.g., desk-top computer, lap-top computer), or one or more mobile computing devices (e.g., smartphone, tablet, etc.). The computing device of the controller 101 can comprise at least one processor, memory, a storage device, an input / output (“I / O”) device / interface, and a communication interface. Additional or alternative components may be used in some examples. Further, in some examples, the controller 101 or the computing device can include fewer components.

[0069] In some examples, the processor(s) of the controller 101 includes hardware for executing instructions (e.g., instructions for carrying out one or more portions of any of the methods disclosed herein), such as those making up a computer program. For example, to execute instructions, the processor(s) may retrieve (or fetch) the instructions from an internal register, an internal cache, the memory, or a storage device and decode and execute them. In particular examples, processor(s) of the controller 101 may include one or more internal caches for data. As an example, the processor(s) of the controller 101 may include one or more instruction caches, one or more data caches, and one or more translation lookaside buffers (TLBs). Instructions in the instruction caches may be copies of instructions in memory or storage device. In some examples, the processor of the controller 101 may be configured (e.g., include programming stored thereon or executed thereby) to carry out one or more portions of any of the example methods or acts disclosed herein. In some examples, the processor of the controller 101 is configured to perform any of the acts disclosed herein or cause one or more portions of the computing device or the controller 101 to perform at least one of the acts114910-4258-31682disclosed herein. Such configuration can include one or more operational programs (e.g., computer program products) that are executable by the at least one processor of the controller 101.

[0070] The at least one computing device (e.g., a server) of the controller 101 may include at least one memory storage medium (e g , memory and / or storage device). The computing device of the controller 101 may include memory, which is operably coupled to the processor(s) of the controller 101. The memory may be used for storing data, metadata, and programs for execution by the processor(s). The memory of the controller 101 may include one or more of volatile and non-volatile memories, such as Random Access Memory (RAM), Read-Only Memory (ROM), a solid-state disk (SSD), Flash, Phase Change Memory (PCM), or other types of data storage. The memory of the controller 101 may be internal or distributed memory

[0071] The computing device of the controller 101 may include the storage device having storage for storing data or instructions. The storage device may be operably coupled to the at least one processor. In some examples, the storage device of the controller can comprise a non-transitory memory storage medium, such as any of those described above. The storage device (e g., non-transitory storage medium) of the controller 101 may include a hard disk drive (HDD), a floppy disk drive, flash memory, an optical disc, a magneto-optical disc, magnetic tape, or a Universal Serial Bus (USB) drive or a combination of two or more of these. The storage device of the controller 101 may include removable or non-removable (or fixed) media. The storage device of the controller 101 may be internal or external to the computing device. In some examples, the storage device of the controller 101 may include non-volatile solid-state memory. In some examples, the storage device of the controller 101 may include read-only memory (ROM). Where appropriate, this ROM may be mask-programmed ROM, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), electrically alterable ROM (EAROM), or flash memory or a combination of two or more of these. In some examples, one or more portions of the memory and / or the storage device (e.g., memory storage medium(s)) may store one or more databases thereon.

[0072] The computing device of the controller 101 also may include one or more I / O devices / interfaces, which are provided to allow a user to provide input to, receive output from, and otherwise transfer data to and from the computing device. These I / O devices / interfaces of the controller 101 may include a mouse, keypad or a keyboard, a touch screen, camera, optical scanner, network interface, web-based access, modem, a port, other known I / O devices, or a combination of such I / O devices / interfaces. The touch screen may be activated with a stylus or a finger. The I / O devices / interfaces of the controller 101 may include one or more devices for presenting output to a user, including, but not limited to, a graphics engine, a display (e.g , a display screen or monitor), one or more output drivers (e.g., display drivers), one or more audio speakers, and one or more audio drivers

[0073] The computing device of the controller 101 also may include a communication interface. The communication interface may include hardware, software, or both. The communication interface of the controller 101 may provide one or more interfaces for124910-4258-31682communication (such as, for example, packet-based communication) between the computing device and one or more additional computing devices or one or more networks. For example, communication interface of the controller 101 may include a network interface controller (NIC) or network adapter for communicating with an Ethernet or other wire-based network or a wireless NIC (WNIC) or wireless adapter for communicating with a wireless network, such as a WIFI. Any suitable network and any suitable communication interface of the controller 101 may be used. For example, the computing device of the controller 101 may communicate with an ad hoc network, a personal area network (PAN), a local area network (LAN), a wide area network (WAN), a metropolitan area network (MAN), or one or more portions of the Internet or a combination of two or more of these. One or more portions of one or more of these networks may be wired or wireless. As an example, one or more portions of controller 101 may communicate with a wireless PAN (WPAN) (such as, for example, a BLUETOOTH WPAN), a WIFI network, a WI-MAX network, a cellular telephone network (such as, for example, a Global System for Mobile Communications (GSM) network), or other suitable wireless network or a combination thereof. The computing device of the controller 101 may include any suitable communication interface for any of these networks, where appropriate.

[0074] The computing device of the controller 101 may include a bus. The bus can include hardware, software, or both that couples components of computing device of the controller 101 to each other For example, the bus of the controller 101 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a front-side bus (FSB), a HYPERTRANSPORT (HT) interconnect, an Industry Standard Architecture (ISA) bus, an INFINIBAND interconnect, a low-pin-count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCIe) bus, a serial advanced technology attachment (SATA) bus, a Video Electronics Standards Association local (VLB) bus, or another suitable bus or a combination thereof.

[0075] In many embodiments, the frequency of the ultrasonic waves 115 generated by the transducers 102 may be about 100 kHz to about 650 kHz, about 100 kHz to about 300 kHz, about 200 kHz to about 650 kHz, about 100 kHz to about 200 kHz, about 124 kHz to about 250 kHz, about 200 kHz to about 300 kHz, about 250 kHz to about 350 kHz, about 300 kHz to about 650 kHz, less than about 500 kHz, less than about 450 kHz, less than about 650 kHz, less than about 350 kHz, less than about 300 kHz, less than about 250 kHz, less than about 200 kHz, or less than about 124 kHz.

[0076] An ultrasound system of this disclosure may be configured to treat a variety of conditions or disorders in the brain. The transducers 102 may be positioned or selectively positionable to treat the variety of conditions or disorders in the brain. For example, an ultrasound device 100 and the controller 101 may be configured to treat a condition of the brain including at least one of cognitive decline or Alzheimer’s disease, and the target region 20a may include one or more of a region of the brain associated with memory functions, a hippocampus of the brain, an entorhinal cortex of the brain, an amygdala of the brain, or a nucleus basalis of134910-4258-31682Meynert of the brain. In some embodiments, an ultrasound device 100 and the controller 101 may be configured to treat a condition of the brain including depression, and the target region 20a may include one or more of a cingulate cortex of the brain or a subcallosal cingulate cortex of the brain In some embodiments, an ultrasound device 100 and the controller 101 may be configured to treat a condition of the brain including chronic pain, and the target region 20a may include one or more of an anterior cingulate cortex of the brain, a medial cingulate cortex of the brain, a subcallosal cingulate cortex, a ventral posterolateral nucleus, or a ventral posteromedial nucleus. In some embodiments, an ultrasound device 100 and the controller 101 may be configured to treat a condition of the brain including addiction, and the target region 20a may include one or more of a nucleus accumbens of the brain, a subcallosal cingulate cortex of the brain, or an anterior cingulate cortex of the brain. In some embodiments, an ultrasound device 100 and the controller 101 may be configured to treat a condition of the brain including food cravings, and the target region 20a may include one or more of a nucleus accumbens of the brain or a nucleus accumbens shell of the brain. In some embodiments, an ultrasound device 100 and the controller 101 may be configured to treat a condition of the brain including anxiety, and the target region 20a may include one or more of an amygdala of the brain or a stria terminalis of the brain. In some embodiments, an ultrasound device 100 and the controller 101 may be configured to treat a condition of the brain including post-traumatic stress brain disorder, and the target region 20a may include one or more of an amygdala of the brain or a bed nucleus of a stria terminalis of the brain.

[0077] In some embodiments, the transducers 102 may be disposed substantially parallel to each other on either side of the frame 103. Further the transducers 102 may be aligned with each other across the frame 103. The frame 103 may be configured as a substantially rigid member, configured to maintain the relative positions of the transducers 102.

[0078] The transducers 102 may be positioned or selectively positionable to treat the variety of conditions or disorders in the brain For example, an ultrasound device 100 and the controller 101 may be configured to treat a condition of the brain including at least one of cognitive decline or Alzheimer’s disease, and the target region 20a may include one or more of a region of the brain associated with memory functions, a hippocampus of the brain, an entorhinal cortex of the brain, an amygdala of the brain, or a nucleus basalis of Meynert of the brain. In some embodiments, an ultrasound system including the ultrasound device 100 and the controller 101 may be configured to treat a condition of the brain including depression, and the target region 20a may include one or more of a cingulate cortex of the brain or a subcallosal cingulate cortex of the brain In some embodiments, an ultrasound system including the ultrasound device 100 and the controller 101 may be configured to treat a condition of the brain including chronic pain, and the target region 20a may include one or more of an anterior cingulate cortex of the brain, a medial cingulate cortex of the brain, a subcallosal cingulate cortex, a ventral posterolateral nucleus, or a ventral posteromedial nucleus. In some embodiments, an ultrasound system including the ultrasound device 100 and the controller 101 may be configured to treat a condition of the brain including addiction, and the target region 20a may include one or more of a nucleus144910-4258-31682accumbens of the brain, a subcallosal cingulate cortex of the brain, or an anterior cingulate cortex of the brain. In some embodiments, an ultrasound system including the ultrasound device 100 and the controller 101 may be configured to treat a condition of the brain including food cravings, and the brain target region 20a may include one or more of a nucleus accumbens of the brain or a nucleus accumbens shell of the brain In some embodiments, an ultrasound system including the ultrasound device 100 and the controller 101 may be configured to treat a condition of the brain including anxiety, and the brain target region 20a may include one or more of an amygdala of the brain or a stria terminalis of the brain. In some embodiments, an ultrasound system including the ultrasound device 100 and the controller 101 may be configured to treat a condition of the brain including post-traumatic stress brain disorder, and the brain target region 20a may include one or more of an amygdala of the brain or a bed nucleus of a stria terminalis of the brain.

[0079] In some cases, the transducers 102 are positioned symmetrically with respect to the mid-sagittal plane of the head of the subject. The ultrasound device 100 is configured to correct for the ultrasound aberration of the head 10 and coupling medium 104. The system can compensate for the attenuation of ultrasound by the respective segment of the head 10 and the coupling medium 104. Such compensation takes into account all obstacles positioned between transducers. These include the skull, the scalp, the ultrasound coupling, any air pockets between the scalp and the transducers, and inner parts of the head including the dura and the brain.

[0080] In some embodiments, the system can compensate for attenuation of ultrasound using a Relative Through-Transmit (RTT) procedure that determines transmission coefficients between each transducer and a target region of the brain. Such an RTT procedure can include obtaining reference signal amplitude measurements while the system (e.g., the transducers) is immersed in a liquid medium (e.g., degassed water), and then obtaining corresponding measurements with the transducers positioned on the head of the subject. Based on a comparison between the measured amplitudes (e.g., between the reference and in-vivo measurements) the system or controller can compute a scaling factor and adjust driving voltages to the transducers to compensate for the attenuation of ultrasonic energy.

[0081] Thus, in some cases, the system is configured to compensate for skull-induced aberrations to maintain a focal precision at the intended deep brain region. Thus, in some cases, the system is configured to compensate for skull-induced aberrations to maintain a focal precision at the intended deep brain region using an ultrasound through-transmit procedure.

[0082] In many embodiments, a method of activating the two transducers to generate ultrasonic waves effective to stimulate the target region of the brain of the subject without ablating the target region of the brain of the subject comprises activating the two transducers to generate ultrasonic waves at a frequency of about 200 kHz to about 650 kHz effective to stimulate the target region of the brain of the subject without ablating the target region of the brain of the subject.154910-4258-31682

[0083] In some embodiments of the method, the target region may comprise a region of the brain that is associated with at least one of cognitive decline or Alzheimer’s disease. In these and other embodiments of the method, the target region may include one or more of a region of the brain associated with memory functions, a hippocampus of the brain, an entorhinal cortex of the brain, an amygdala of the brain, or a nucleus basalis of Meynert of the brain.

[0084] In some embodiments of the method, the target region may comprise a region of the brain associated with depression. In these and other embodiments of the method, the target region may include one or more of a cingulate cortex of the brain or a subcallosal cingulate cortex of the brain.

[0085] In some embodiments of the method, the target region may comprise a region of the brain associated with chronic pain. In these and other embodiments of the method, the target region may include one or more of an anterior cingulate cortex of the brain, a medial cingulate cortex of the brain, a subcallosal cingulate cortex, a ventral posterolateral nucleus, or a ventral posteromedial nucleus.

[0086] In some embodiments of the method, the target region may comprise a region of the brain associated with addiction. In these and other embodiments of the method, the target region may include one or more of a nucleus accumbens of the brain, a subcallosal cingulate cortex of the brain, or an anterior cingulate cortex of the brain.

[0087] In some embodiments of the method, the target region may comprise a region of the brain associated with food craving. In these and other embodiments, the target region may include one or more of a nucleus accumbens of the brain or a nucleus accumbens shell of the brain.

[0088] In some embodiments of the method, the target region may comprise a region of the brain associated with anxiety. In these and other embodiments of the method, the target region may include one or more of an amygdala of the brain or a stria terminals of the brain.

[0089] In some embodiments of the method, the target region may comprise a region of the brain associated with post-traumatic stress brain disorder. In these and other embodiments of the method, the target region may include one or more of an amygdala of the brain or a bed nucleus of a stria terminalis of the brain.

[0090] In some embodiments of this method of treating a condition of a brain of a subject, the condition of the brain may include at least one of cognitive decline or Alzheimer’s disease and the target region may include one or more of a region of the brain associated with memory functions, a hippocampus of the brain, an entorhinal cortex of the brain, an amygdala of the brain, or a nucleus basalis of Meynert of the brain.

[0091] FIG. 3 illustrates a perspective view of an optical guidance system 120 for extracting contour data 162 of the head 10 of a subject, according to embodiments described herein. As seen in the illustrated embodiment, optical guidance system 120 can be used to extract dense contour data of the head, anatomical landmarks corresponding to the 10-20 system, and / or a sparser set of facial fiducials, as have been previously described. An image sensor 130 can be used to acquire such data describing a geometry of a head of a subject164910-4258-31682

[0092] As seen in the illustrated embodiment, an image sensor 130 (e.g., a camera) can be positioned at a fixed distance from the head 10. An adjustable frame 103 holding the ultrasonic transducer or transducers is attached to the head at defined anatomical landmarks, such as the nasion and the inion. The image sensor 130 images the contour of the head 10. When the transducer and ultrasound beam axis 106 is perpendicular to the head symmetry plane 107, it is possible to take one or more images of the side of the head to obtain targeting information (e.g., 3D coordinates of head-surface contour points) Otherwise stated, image sensor 130 is positioned to capture an optical image of the head of a subject while a head-mounted frame with ultrasonic transducers is affixed to anatomical reference points of the head.

[0093] In some embodiments, the optical guidance system 120 can gather all necessary headsurface geometry from a single two-dimensional image. In some embodiments such an image is captured when the beam axis of the transducer and / or image sensor is perpendicular to the subject’s midline sagittal plane. Under this configuration, the optical guidance system 120 can isolate the head silhouette, extracts contour coordinates, and reconstructs three-dimensional surface geometry corresponding to the head of the subject.

[0094] In alternate embodiments, the system repeatedly captures optical images after each adjustment of the transducer frame. The targeting model 128 can then recompute target coordinates from updated head-surface geometry, and the controller incrementally adjusts transducer position until the computed beam path intersects the desired deep-brain target within a desired tolerance

[0095] An image of a ruler 122 or visual marker, of a known dimension, can be placed onto the frame and / or device so as to be used to convert the camera pixels into real-world distances The image sensor 130 is placed at a sufficient distance so that the ruler 122 positioned on the frame spans a similar number of camera pixels as if the ruler 122 was placed on the head contour 124. The ruler 122 on the frame has a known dimension and may include QR-coded patterns or other calibration structures to facilitate pixel-to-distance conversion and to support de-warping when the image is captured at a non-orthogonal angle. The image sensor 130 is positioned near the transducer beam axis 106 to minimize distortion.

[0096] In some cases the image sensor 130 is placed at a distance of at least 5-10 times the thickness of the head. The center of the image sensor 130 can be placed close to the axis of the ultrasound transducers and the ultrasound beam axis 106 to minimize image distortion. In some cases, a single side-view image is sufficient to capture all relevant geometry of the head.

[0097] In some cases, the image sensor 130 is a two-dimensional optical camera or an optical tracking system capable of capturing the full head contour. In some cases, the head of the subject is fitted with a swim cap, EEG cap, or equivalent so as to compress the hair, so that a camera (e.g., image sensor 130) can image the contour of the head 10. A frame holding the ultrasonic transducer or transducers 102 is attached to the head to defined anatomical landmarks, such as the nasion and the inion.

[0098] The camera acquires at least a single side-view image when the ultrasound beam axis 106 is perpendicular to the head’s midline sagittal plane, in some cases enabling complete174910-4258-31682extraction of head contour data from one image. In alternative configurations where the transducer assembly is oblique to the symmetry plane, the system acquires multiple images to triangulate the contour.

[0099] Once an image is captured, the controller 101 and / or system at large may extract head-surface contour points from the image data. Depending on the embodiment, these may include dense head-geometry or head-surface contour points, anatomical landmarks corresponding to the 10-20 system landmarks, or facial fiducials.

[0100] The processor can then convert pixel coordinates into 3D coordinates using triangulation or optical-tracking techniques. In some embodiments, the optical guidance system 120 may incorporate QR-coded rulers to correct image de-warping when images are taken at an angle. In some embodiments, dense samples of the head surface are collected using an optical tracking system to produce a full 3D head contour.

[0101] In some cases, the system may instead use an optical tracking device (e.g., OptiTrack) to track and / or acquire the 3D positions of facial fiducials or to discretize the head contour into a dense point cloud. The optical tracker can also capture the position of transducer-mounted markers concurrently with head-surface points, thus enabling iterative adjustment and / or validation of transducer placement. For instance, in some cases, an optical tracker and / or image sensor can re-image the head of the subject, re-extract the 3D coordinates of the contour points and of the transducer, and provide guidance to adjust a position of the transducer, image sensor, and / or optical tracker so as to better align consequently emitted therapeutic ultrasound with estimated 3D coordinates of a deep brain target.

[0102] In some embodiments, the targeting may be accomplished sans identification of discrete anatomical fiducials. For instance, in some cases, the optical guidance system 120 may acquire a dense point cloud representing the head contour, and a surface-registration algorithm aligns this point cloud to a standard MRI head template The transform obtained from this surface matching is applied to the associated template brain, and deep brain target coordinates are inferred directly from the aligned template anatomy. In some cases, such a contour-only approach can eliminate need or use of manual or automated labeling of individual fiducial points.

[0103] The acquired contour data 162 (in any form) can then be used as inputs to model(s) trained on MRI libraries. In embodiments based on large MRI datasets (e.g., 430+ MRIs with scalp segmentation, standardization, and quality control), the head contour geometric features are statistically associated with corresponding deep-brain coordinates, including subregions of the anterior cingulate cortex, subcallosal cingulate cortex, anterior commissure, posterior commissure, and other regions of therapeutic relevance. These associations were previously quantified to achieve mean target-registration errors on the order of 4-6 mm when only optical or facial-landmark geometry is available

[0104] Additionally, the optical guidance system 120 can be used to properly position the ultrasound device and / or adjustable frame 103. For instance, in some cases, the optical guidance system 120 extracts the geometric information needed to orient the ultrasound184910-4258-31682transducer(s) such that the beam axis is aligned with the estimated deep brain target coordinates. The therapeutic focal volume 119 may be selected to be two to three times broader than the expected 5-6 mm optical-based targeting error, ensuring engagement of the target even with modest geometric uncertainty.

[0105] FIGS. 4A and 4B illustrates an embodiment of a representation of contour data 162, visible as extracted head-contour geometry (which may be provided as a dense silhouette), a set of discretized contour points, or a structured set of anatomical fiducial coordinates obtained from optical imaging and / or optical tracking. In some cases, the contour data 162 can include corresponding contour points 125 disposed along the 10-20 midlines of the head as derived from the 10-20 EEG system. Accordingly, FIG. 4A illustrates a lateral view of a head of a subject and showcases a corresponding anterior-posterior midline, according to embodiments described herein. FIG. 4B illustrates a frontal view of a head of a subject and showcases a corresponding left-right transverse midline, according to embodiments described herein.

[0106] As seen in the illustrated embodiments, the contour points 125 may include the seven midline points (for each midline). In some implementations, only the facial fiducials (e.g., the nasion, left inner canthus (LIC), right inner canthus (RIC), left outer canthus (LOC), right outer canthus (ROC), left preauricular point (LPA), right preauricular point (RPA), left external acoustic meatus (LEM), right external acoustic meatus (REM) may be extracted). In extended implementations, additional canonical 10-20 or 10-10 landmarks can be used. For instance, along the head symmetry axis, these can include the points nasion 132 and inion 134, as well as points FPi(e.g., point 142B), Fz (e.g., point 140B), Cz (e.g., point 136), Pz (e.g., point 140C), and one of the occipital lobes (e.g., point 140A). Along the sagittal midline, these can include the RPA 138A, and LPA 138B, as wells as Cz (e.g., point 142B), Cz (e.g., point 142B), C3 (e.g., point 142B). The points Ts, T4, and A1, Az (not shown in FIGS. 4A and 4B) may be included as well.

[0107] Points 123A and 123B represent the right and left (respectively) exterior points of the ultrasound device placed on the head, while point 140D represents the foremost point.

[0108] These discretized head contour points 125 can serve as predictors for deep brain target regions, such as the anterior cingulate cortex, the subcallosal cingulate cortex, and the anterior corpus callosum. The discretized head contour points 125 derived from the anterior-posterior and left-right midlines can be used as inputs to a targeting model 128 trained on a large MRI library containing coordinate mappings from scalp contour data to deep brain regions. Otherwise stated, the discretized head contour points 125 (e.g., including anatomical landmarks) function as predictors for deep structures such as the anterior cingulate cortex 144, the subcallosal cingulate cortex 146, and the anterior corpus callosum 148 (as seen in FIG. 6). The learned relationship between midline surface points and deep structures may be represented using linear regression, nonlinear regression, or neural-network models. The targeting model 128 is trained using a large MRI library that supplies ground-truth pairings between scalp landmarks and internal brain coordinates.

[0109] As previously discussed, the learned mapping from 10-20 surface points to deep targets may be implemented as a linear regression model, a nonlinear regression model, or a194910-4258-31682neural network with one or multiple layers. The targeting model 128 is trained on a library of MRI images of the head. For instance, in one embodiment, a library of 431 MRI images of the head was used to build models that associated the 7 midline coordinates of the 10-20 system to 8 subregions of the anterior cingulate cortex. One model was built for each subregion, and the results were averaged together A linear model, and nonlinear deep learning models (1 or 2 neuronal layers) were used The performance of the models was evaluated using 10-fold cross-validation, i. e , the models were tested on an independent test set. The data are shown as mean ± s d., across the 8 subregions.

[0110] To evaluate the robustness of the models with respect to inaccuracy in the determination of the anatomical landmarks, the abscissa provides the standard deviation of jitter injected into all 3 coordinates of each of the 7 midline points. Such a method was shown to provide a guidance accuracy of about 5mm This is sufficient when the therapeutic focal volume of the tissue activated by the ultrasound is about twice that value. The accuracy is relatively insensitive to small errors in estimation of the head contour coordinates (up to about 4mm error, standard deviation) For larger errors, more complex models (nonlinear, multiple layers) perform better than simpler models.

[0111] In some embodiments, the contour points 125 are provided to a predictive targeting model 128 trained on a dataset of MRI scans. The targeting model 128 can associate headsurface geometry with deep-brain coordinates, using MRI-derived ground-truth targets. Training datasets typically include approximately 430 MRI subjects (mean age 31.3 ± 11.0 years; nasioninion distance 357.1 ± 16.9 mm; LPA-RPA distance 320.3 ± 194 mm; head circumference 562.8 ± 20.9 mm; 53.2% female). Subjects with severe defacing or known neurological conditions are excluded. MRI volumes are preprocessed using BrainSuite for artifact removal and nonuniformity correction, and linear and nonlinear registration tools from FSL (flirt, fnirt) are used to map a template brain (e.g., the 696-parcel Yale Brain Atlas or the MNI152 template) to each subject. These transformations define the true coordinates of deep-brain targets, including multiple anterior cingulate subregions (subgenual BA25; pregenual; mid-cingulate parcels s24, p24, a24, 33) and the anterior and posterior commissures.

[0112] In some cases, model architectures may include linear regression (e.g., Moore-Penrose pseudo-inverse), nonlinear regression, or multilayer neural networks. In prior MRI-free targeting studies, linear models outperform neural-network variants for many brain regions, achieving mean errors of 3.9 ± 1.7 mm for cingulate targets and 4.9 ± 2.5 mm globally.Comparable optical-capture models achieve approximately 5.75 ± 2.98 mm average error across deep-brain targets, with robustness to landmark jitter up to 4 mm in all coordinate dimensions.

[0113] Accordingly, the optical guidance system 120 can produce 3D target coordinates suitable for neuromodulatory interventions, including transcranial focused ultrasound, temporally interfering electric fields, and other methods requiring accurate MRI-free deep-target estimation For example, after the predicted or determined 3D target coordinates corresponding to a deep brain target region (e.g., deep brain target region 20a as seen in FIG. 2) can then be transmitted to a controller of a head-mounted device (e g , such as controller 101 and / or ultrasound device204910-4258-31682100 as seen in FIG. 2) including at least one ultrasound transducer. The device can then emit ultrasonic energy toward the output 3D coordinates of the located deep brain target so as to deliver neuromodulatory brain-therapy to the located deep brain target. The delivery of the ultrasonic energy to the located deep brain target may be subsequent to the outputting the 3D coordinates of the located deep brain target. In some cases, the therapeutic focal volume may be engineered to be at least twice the measured targeting error to ensure reliable engagement of the intended deep structure despite modest geometric variability.

[0114] FIG. 5 illustrates an embodiment of a representation of contour data 162, corresponding to three-dimensional coordinates of canonical 10-20 EEG positions 150, including intermediate and / or anatomical landmarks supplementing midline positions, according to embodiments described herein. These expanded 10-20 EEG positions 150 provide a denser sampling of the scalp surface and increase geometric constraint for MRI-free targeting. These points can include some, or all of the 10-20 landmarks, including FP, Fs, Cs, Ps, F?, T3, Ts, O1 , A1 any of the midline points previously discussed, or even of the points corresponding to the opposite side of the head not seen in FIG. 5 (e.g., FP2, F4, Fs, C4, T4, O2, A2, Te). Within both FIG. 5 and FIG. 6, the percentages seen between the landmarks (e.g , 10%, 20%, etc.) can correspond to a percentage of the total distance (front-back or right-left) of the skull. As seen in the illustrated embodiment, these anatomical points may be extracted from optical imaging or optical tracking systems and serve as inputs to a targeting model 128 trained to predict deepbrain coordinates.

[0115] In this expanded 10-20 configuration, the pre-trained, targeting model 128 receives the full set of intermediate and midline EEG landmarks as inputs and produces predicted target coordinates corresponding to deep brain regions of interest.

[0116] FIG. 6 illustrates an embodiment of target coordinates 164 corresponding to three-dimensional locations of deep brain regions 20a as produced by the targeting model 128, according to embodiments described herein. Once a set of head-surface coordinates (e.g., contour data 162) is acquired (e.g., from dense contour geometry, facial fiducials, midline-only 10-20 landmarks, or expanded 10-20 intermediate positions), the targeting model 128 maps these surface inputs to predicted intracranial coordinates such as the anterior cingulate cortex 144, the subcallosal cingulate cortex 146, and the anterior corpus callosum 148.

[0117] Given a set of head-surface coordinates derived from optical imaging, the model outputs predicted coordinates for targets such as multiple subregions of the anterior cingulate cortex (e.g., subgenual, pregenual, anterior mid-cingulate), the subcallosal cingulate cortex, and the anterior and posterior commissures These predicted coordinates 164 are expressed in a standardized intracranial reference frame, such as AC-PC space, enabling direct integration with neuronavigation or ultrasound-positioning controllers.

[0118] Accordingly, the optical guidance system 120 can produce 3D target coordinates suitable for neuromodulatory interventions that benefit from individualized MRI-free targeting, including transcranial focused ultrasound, transcranial magnetic stimulation (TMS), and transcranial electrical stimulation (tACS and / or tDCS). For example, after predicted or214910-4258-31682determined 3D target coordinates corresponding to a deep brain target region are generated, the coordinates may be transmitted to a controller of a head-mounted or external stimulation device for subsequent delivery of therapy. In some embodiments, a therapeutic focal volume, induced field distribution, or stimulation montage is selected to be tolerant to residual targeting error, ensuring engagement of the intended target despite modest geometric variability.

[0119] In some embodiments, the noninvasive neuromodulatory therapy comprises transcranial magnetic stimulation (TMS). The predicted target coordinates may be used to guide placement and orientation of a stimulation coil relative to the head such that an induced electromagnetic field preferentially affects a cortical region corresponding to the predicted target, or a cortical site selected to modulate a deeper brain circuit associated with the predicted deep target.

[0120] In some embodiments, the noninvasive neuromodulatory therapy comprises transcranial electrical stimulation, including direct current stimulation (tDCS) and / or alternating current stimulation (tACS). The predicted target coordinates may be used to inform electrode placement, polarity, and stimulation montage such that current flow is biased toward a target region determined from head-surface geometry, without reliance on MRI-derived head models.

[0121] Any methods disclosed herein comprise one or more steps or actions for performing the described method. The method steps and / or actions may be interchanged with one another. In other words, unless a specific order of steps or actions is required for proper operation of the embodiment, the order and / or use of specific steps and / or actions may be modified.

[0122] References to approximations are made throughout this specification, such as by use of the term “substantially.” For each such reference, it is to be understood that, in some embodiments, the value, feature, or characteristic may be specified without approximation. For example, where qualifiers such as “about” and “substantially” are used, these terms include within their scope the qualified words in the absence of their qualifiers.

[0123] Similarly, in the above description of embodiments, various features are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure. This method of disclosure, however, is not to be interpreted as reflecting an intention that any claim require more features than those expressly recited in that claim Rather, as the following claims reflect, inventive aspects lie in a combination of fewer than all features of any single foregoing disclosed embodiment.

[0124] The claims following this written disclosure are hereby expressly incorporated into the present written disclosure, with each claim standing on its own as a separate embodiment. This disclosure includes all permutations of the independent claims with their dependent claims. Moreover, additional embodiments capable of derivation from the independent and dependent claims that follow are also expressly incorporated into the present written description

[0125] Without further elaboration, it is believed that one skilled in the art can use the preceding description to utilize the invention to its fullest extent. The claims and embodiments disclosed herein are to be construed as merely illustrative and exemplary, and not a limitation of224910-4258-31682the scope of the present disclosure in any way. It will be apparent to those having ordinary skill in the art, with the aid of the present disclosure, that changes may be made to the details of the above-described embodiments without departing from the underlying principles of the disclosure herein. In other words, various modifications and improvements of the embodiments specifically disclosed in the description above are within the scope of the appended claims. Moreover, the order of the steps or actions of the methods disclosed herein may be changed by those skilled in the art without departing from the scope of the present disclosure. In other words, unless a specific order of steps or actions is required for proper operation of the embodiment, the order or use of specific steps or actions may be modified. The scope of the invention is therefore defined by the following claims and their equivalents.234910-4258-31682

Claims

CLAIMSWhat is claimed is:

1. A system for determining a location of a deep brain target, the system comprising:a sensor configured to acquire data describing a geometry of a head of a subject; and a processor coupled to the sensor, the processor configured to:receive the data describing a geometry of the head of the subject gathered by the sensor;determine 3D coordinates of head-surface contour points from the acquired data; apply a predictive model to the determined 3D coordinates of head-surface contour points, the model having been previously trained on a dataset of head and brain images and configured to associate head-surface coordinates with 3D coordinates of at least one deep brain target;receive 3D coordinates of a located deep brain target as output from the model; andoutput 3D coordinates of the located deep brain target for subsequent delivery of neuromodulatory brain-therapy.

2. The system of claim 1, wherein the head-surface contour points comprises anatomical landmarks, 10-20 scalp positions, or a dense head-surface point cloud extracted from the acquired data.

3. The system of claim 1 or 2, further comprising:a head-mounted device comprising at least one ultrasound transducer configured to emit ultrasonic energy toward the output 3D coordinates of the located deep brain target so as to deliver neuromodulatory brain-therapy to the located deep brain target.

4. The system of any one of claims 1 to 3, wherein a focal volume of tissue activated by the ultrasonic energy is at least twice as large as a targeting error associated with the predictive model.

5. The system of any one of claims 1 to 4, wherein the sensor comprises an optical image sensor configured to acquire one or more images of a 2D contour of the head of the subject.

6. The system of claim 5, wherein a single 2D image of the head contour provides sufficient geometric information for locating the deep-brain target7. The system of claim 5, wherein the acquired data comprises a plurality of two-dimensional images and determining 3D coordinates of head-surface contour points from the acquired data comprises reconstructing 3D coordinates of head-surface contour points or surface geometry via photogrammetric reconstruction based on the plurality of two-dimensional images.

8. The system of any one of claims 1 to 7, wherein the sensor is positioned at a distance of at least five times a thickness of the head of the subject; andwherein the sensor is positioned such that a beam axis of the sensor is substantially perpendicular to a midline sagittal plane of the head of the subject.

9. The system of any one of claims 1 to 4, wherein the sensor comprises an optical tracking system configured to track 3D positions of the head-surface contour points.244910-4258-3168210. The system of any one of claims 1 to 9, wherein the predictive model comprises a linear regression model or a neural network mapping from the 3D positions of the head-surface contour points to the 3D coordinates of the located deep brain target.

11. The system of any one of claims 1 to 10, wherein the dataset of head and brain images comprises a library of magnetic resonance (MRI) images of a plurality of subjects associating the 3D positions of head-surface contour points and deep brain targets.

12. A method for determining a location of a deep brain target, the method comprising:positioning an image sensor at a fixed distance from a head of a subject; capturing, via the image sensor, one or more images of the head of the subject; extracting, from the captured images, 3D coordinates of contour points on the head; providing the 3D coordinates of contour points to a predictive model trained on a library of magnetic resonance (MRI) images, wherein the predictive model is configured to map from the 3D coordinates of the contour points to 3D coordinates of at least one deep-brain target; receiving estimated 3D coordinates of at least one deep-brain target from the predictive model;positioning a brain-therapy transducer onto the head of the subject;emitting a therapeutic ultrasound to the estimated 3D coordinates of at least one deepbrain target.

13. The method of claim 12, wherein the one or more images further include a calibration marker of known dimension, and extracting the 3D coordinates comprises determining pixel-to-distance scale from a visual ruler or the calibration marker14. The method of claim 12 or 13, further comprising:re-imaging the head with the image sensor;re-extracting the 3D coordinates of the contour points and of the transducer; and adjusting a position of the transducer to align the emitted therapeutic transcranial ultrasound with the estimated 3D coordinates of the deep brain target.

15. The method of any one of claims 12 to 14, wherein applying the predictive model further comprises fitting the extracted 3D coordinates of the 3D coordinates of the contour points to a standard brain template and inferring the 3D coordinates of the deep brain target from the standard brain template.

16. A system for determining a location of a brain target comprising:a sensor configured to acquire data describing a geometry of a head of a subject; a processor coupled to the sensor and configured to:extract, from the acquired data, a plurality of head-surface contour points representing a three-dimensional surface of the head;align the extracted contour points to a standard MRI head template using a surface-matching transform;determine, from the aligned MRI template, 3D coordinates of a deep brain target of the subject; and254910-4258-31682output the determined 3D coordinates for use in delivery of noninvasive neuromodulatory brain therapy.

17. The system of claim 16, further comprising a head-mounted frame including two or more calibration markers of known 3D geometry, wherein the sensor acquires data describing the calibration markers together with the geometry of a head of a subject.

18. The system of claim 16, wherein the processor outputs the determined 3D coordinates in an anterior-commissure / posterior-commissure (AC-PC) coordinate system.

19. The system of claim 16, wherein the noninvasive brain-therapy comprises transcranial focused ultrasound directed to a brain region associated with at least one of: depression, anxiety, chronic pain, addiction, cognitive decline, epilepsy, post-traumatic stress disorder, or a movement disorder20. The system of claim 16, wherein the noninvasive brain-therapy comprises transcranial magnetic stimulation delivered by a stimulation coil positioned based on the determined 3D coordinates.

21. The system of claim 16, wherein the noninvasive brain-therapy comprises transcranial electrical stimulation including at least one of transcranial direct current stimulation or transcranial alternating current stimulation delivered using electrodes positioned based on the determined 3D coordinates.

22. A method for determining a location of a brain target for delivery of noninvasive brain therapy, the method comprising:receiving data describing a geometry of a head of a subject from one or more sensors; computing 3D coordinates of a plurality of head-surface contour points from the received data;applying a model to the computed 3D coordinates, the model having been previously trained on a dataset of head and brain images and configured to associate head-surface coordinates with 3D coordinates of at least one deep brain target; andreceiving the 3D coordinates of the at least one deep brain target as output from the model;outputting the 3D coordinates of the at least one deep brain target for delivery of noninvasive brain-therapy.264910-4258-31682

Images