Markerless tracking and placement of transcranial magnetic stimulation coils for therapeutic and diagnostic procedures

Abstract

A method for identifying a location and a position of a treatment coil used for transcranial magnetic stimulation (TMS). The method may comprise receiving a plurality of images from one or more cameras. The plurality of images may comprise images of the treatment coil and of a part of a live subject. The method may comprise determining a surface of the treatment coil within an image by using a neural network and or surface detection. The method may comprise determining the location of the treatment coil based on the distance and direction to the detected surface(s), determining an orientation of the treatment coil based on an orientation of the plurality of normal vectors, and displaying, via a graphical user interface (GUI) on a display, the location and the orientation of the treatment coil relative to a target location on the head of the human subject.

Description

NNI 0507 DCVS PCTMARKERLESS TRACKING AND PLACEMENT OF TRANSCRANIAL MAGNETIC STIMULATION COILS FOR THERAPEUTIC AND DIAGNOSTIC PROCEDURESCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of Provisional U.S. Patent Application No. 63 / 691,581, filed September 6, 2024, the entire disclosure of which is hereby incorporated by reference herein in its entirety.BACKGROUND

[0002] A number of medical ailments are treated or treatable through the application of electrical stimulation to an afflicted portion of a human subject’s body. Examples of electrical stimulation may include magnetic or inductive stimulation, which may make use of a changing magnetic field, and electric or capacitive stimulation in which an electric field may be applied to the tissue. Neurons, muscle, and tissue cells are forms of biological circuitry capable of carrying electrical signals and responding to electrical stimuli. For example, when an electrical conductor is passed through a magnetic field, an electric field is induced causing current to flow in the conductor.Because various parts of the body may act as a conductor, when a changing magnetic field is applied to the portion of the body, an electric field is created causing current to flow. In the context of biological tissue, for example, the resultant flow of electric current stimulates the tissue by causing neurons in the tissue to depolarize. Also, in the context of muscles, for example, muscles associated with the stimulated neurons contract. In essence, the flow of electrical current allows the body to stimulate typical and often desired chemical reactions.

[0003] Electrical stimulation has many beneficial and therapeutic biological effects. For example, the use of magnetic stimulation is effective in rehabilitating injured or paralyzed muscle groups. Another area in which magnetic stimulation is proving effective is treatment of the spine. The spinal cord is difficult to access directly because vertebrae surround it. Magnetic stimulation may be used to block the transmission of pain via nerves in the back (e.g., those responsible for lower back pain).NNI 0507 DCVS PCTFurther, unlike the other medical procedures that stimulate the body, electrical stimulation may be non-invasive. For example, using magnetic fields to generate current in the body produces stimulation by passing the magnetic field through the skin of a human subject.

[0004] Magnetic stimulation also has proven effective in stimulating regions of the brain, which is composed predominantly of neurological tissue. One area of particular therapeutic interest is the treatment of neuropsychiatric disorders. It is believed that more than 28 million people in the United States alone suffer from some type of neuropsychiatric disorder. These include specific conditions such as depression, schizophrenia, mania, obsessive-compulsive disorder, panic disorders, just to name a few. One particular condition, depression, is the often referred to as the “common cold” of psychiatric disorders, believed to affect 19 million people in the United States alone, and possibly 340 million people worldwide. Modem medicine offers depression human subjects a number of treatment options, including several classes of anti -depressant medications like selective serotonin reuptake inhibitors (SSRI), MAIs, tricyclics, lithium, and electroconvulsive therapy (ECT). Yet many human subjects remain without satisfactory relief from the symptoms of depression.

[0005] Repetitive transcranial magnetic stimulation (') has been shown to have anti -depressant effects for human subjects, even those that do not respond to the traditional methods and medications. For example, a subconvulsive stimulation may be applied to the prefrontal cortex in a repetitive manner, causing a depolarization of cortical neuron membranes. The membranes are depolarized by the induction of small electric fields, usually in excess of 1 volt per centimeter (V / cm). These small electric fields result from a rapidly changing magnetic field applied non- invasively.

[0006] Therapeutic and / or diagnostic procedures, such as TMS for example, may require a technician to locate a treatment location (e.g., or target location that is used to determine the treatment location) before performing the therapeutic and / or diagnostic procedure. This process can be time consuming and burdensome. For example, the technician may be required to manually collect multiple points on a patient one-by-one to generate a model of the patient, and after the model is generated, locate the treatment location on the model.NNI 0507 DCVS PCTSUMMARY

[0007] Described herein are methods and apparatuses for identifying a location and a position of a treatment coil used for transcranial magnetic stimulation (TMS) and / or the location and position of a human subject. The foregoing summary is described with reference to a method, but it should be appreciated that the method may be performed by one or more systems as described herein. For example, a system that is configured to identify a location and a position of a treatment coil used for TMS may include any combination of a treatment coil, one or more cameras (e.g., depth cameras), one or more processors, and memory.

[0008] The method for identifying a location and a position of a treatment coil used for TMS may include any combination of the following. The method may comprise receiving a plurality of images from one or more cameras. The plurality of images may comprise images of the treatment coil and of a head of a human subject. The method may comprise determining a surface of the treatment coil within an image of the plurality of images based on a direction of a plurality of vectors within the image. The method may comprise determining the location of the treatment coil based on the plurality of vectors. The method may comprise determining an orientation of the treatment coil based on an orientation of the plurality of vectors. The method may comprise displaying, via a graphical user interface (GUT) on a display, the location and the orientation of the treatment coil relative to a target location on the head of the human subject.

[0009] The treatment system may display, via the GUI on the display, an indication of a displacement error of the treatment coil relative to the target location. The treatment system may receive signals from a contact sensor located on the treatment coil. The treatment system may determine whether the treatment coil is in contact with the head of the human subject based on the signals received from the contact sensor. The treatment system may display, via the GUI on the display, an indication of whether the treatment coil is in contact with the head of the human subject.

[0010] The treatment system may display, via the GUI on the display, an indication of a direction or a distance that the position of the treatment coil needs to be moved to reach the target location. The indication of the direction or the distance may comprise an arrow that is overlaid on top of a graphical representation of the head of the human subject. The treatment system may display, via the GUI on the display, an indication that the position of the treatment coil has been moved to the target location.NNI 0507 DCVS PCT

[0011] The surface of the treatment coil may be determined within the image based on a first subset of the plurality of vectors being in the same direction. The treatment system may display, via the GUI on the display, an object that highlights the position of the treatment coil. The treatment system may track movement of the treatment coil based on additional images received via the one or more cameras.

[0012] The location of the treatment coil may be determined based on a relationship between one or more geometric features of the treatment coil and an induced electric field of the treatment coil. The one or more cameras comprise one or more RGB-d depth cameras.

[0013] A distance between a landmark and a patient marker or identifiable patient features in one or more respective medical images may be used to guide the location of the treatment coil. The one or more medical images may comprise magnetic resonance imaging (MRI) images, positron emission tomography (PET) scan images, single-photon emission computed tomography (SPECT) scan images, and / or ultrasound images.

[0014] The target location may be determined based on a known location in a fitted head model. Displaying, via the GUI on the display, the indication that the position of the treatment coil has been moved to the target location may be done by changing a color of a location indicator.

[0015] A system (e g., set of computers, mathematical model, images, cameras, memory, user input, instructions, and / or data) may train a neural network to detect a location and an orientation of a treatment coil relative to a head of a human subject. The training may comprise receiving a plurality of images from one or more cameras. The plurality of images may comprise images of the treatment coil. The training may comprise locating the coil within one or more images. The training may comprise determining a surface of the treatment coil within the plurality of images. The training may comprise identifying a plurality of vectors associated with the surface of the treatment coil. The training may comprise training a machine learning model to identify the surface of the treatment coil using training data. The training data may comprise the plurality of images, an identification of the surface of the treatment coil within the plurality of images, and the plurality of vectors associated with the surface of the treatment coil.

[0016] A treatment system may receive a second plurality of images associated with a second treatment coil. The treatment system may determine a surface of the second treatment coil using the trained machine learning model based on the second plurality of images.NNI 0507 DCVS PCT

[0017] The treatment system may display, via a GUI on a display, a representation of the second treatment coil. The treatment system may track movement of the second treatment coil based on additional images received via the one or more cameras.

[0018] The treatment system may display, via a GUI on a display, a representation of the first treatment coil. The treatment system may track movement of the first treatment coil based on additional images received via the one or more cameras. The treatment system may identify a housing of the coil based on an orientation of a subset of the identified plurality of vectors.

[0019] A method for training a neural network to detect a location and an orientation of a treatment coil relative to a head of a human subject. The method may comprise receiving a plurality of images from one or more cameras. The plurality of images may comprise images of the treatment coil. The method may comprise determining a surface of the treatment coil within the plurality of images. The method may comprise identifying a plurality of vectors associated with the surface of the treatment coil. The method may comprise training a machine learning model to identify the surface of the treatment coil using training data. The training data may comprise the plurality of images, an identification of the surface of the treatment coil within the plurality of images, and the plurality of vectors associated with the surface of the treatment coil.

[0020] Finally, although the Summary is drafted from the perspective of an apparatus or a method, the concepts described herein may be captured as any combination of an apparatus, method, and / or a computer-readable storage medium (e. , a non-transitoiy computer-readable storage medium) that is located on one or more apparatuses.BRIEF DESCRIPTION OF THE DRAWINGS

[0021] FIG. 1 A is a perspective diagram of an example of a treatment system.

[0022] FIG. IB is a block diagram of an example treatment system.

[0023] FIG. 2 is a block diagram of an example sensor.

[0024] FIG. 3 is a flowchart of an example procedure for tracking and placement of transcranial magnetic stimulation coils for therapeutic and diagnostic procedures.NNI 0507 DCVS PCT

[0025] FIGs. 4A-4C are example graphical user interfaces (GUIs) that include an example head model, an example indication of a treatment coil, and / or an example indication of a target location and orientation for the treatment coil.

[0026] FIG. 5 is a diagram illustrating an example interface of a treatment system that is using image recognition to determine the presence and / or location of a face of a human subject and a treatment coil.

[0027] FIG. 6 is a diagram of an example predetermined head model, a point cloud, and a fitted head model.

[0028] FIG. 7 is another diagram of an example predetermined head model superimposed with a plurality of points determined by a sensor.

[0029] FIG. 8 is a diagram of an example fitted head model that includes an EEG 10-20 coordinate grid.

[0030] FIG. 9 is a flow chart of a procedure for training a machine learning model to determine the location and the orientation of the treatment coil.

[0031] FIG. 10 is a flowchart of a procedure for training and / or using a trained machine learning model to determine the location and / or the orientation of a treatment coil.DETAILED DESCRIPTION

[0032] A detailed description of illustrative embodiments will now be described with reference to the various Figures. Although this description provides a detailed example of possible implementations, it should be noted that the details are intended to be examples and in no way limit the scope of the application.

[0033] In 1831, Michael Faraday discovered that the magnitude of an electric field induced on a conductive loop is proportional to the rate of change of magnetic flux that cuts across the area of the conductive loop. Faraday’s law may be represented as E ~ -(dB / dt), where E is the induced electric field in volts / meter, dB / dt is the time rate of change of magnetic flux density in Tesla / second. In other words, the amount of electric field induced in an object like a conductor may be determined byNNI 0507 DCVS PCT two factors: the geometry and the time rate of change of the flux. The greater the derivative of the magnetic flux, the greater the induced electric field and resulting current density. Because the magnetic flux density decreases quickly with distance from the source of the magnetic field, the flux density is greater the closer the conductor is to the source of the magnetic field. When the conductor is a coil, the current induced in the coil by the electric field may be increased in proportion to the number of turns of the coil.

[0034] When the electric field is induced in a conductor, the electric field creates a corresponding current flow in the conductor. The current flow is in the same direction of the electric field vector at a given point. The peak electric field occurs when dB / dt is the greatest and diminishes at other times. If the magnetic field changes, for example during a magnetic pulse, the current flows in a direction that tends to preserve the magnetic field (e.g., Lenz's Law).

[0035] In the context of electrical stimulation of the anatomy, certain parts of the anatomy (e.g., nerves, tissue, muscle, brain) act as a conductor and carry electric current when an electric field is presented. The electric field may be presented to these parts of the anatomy transcutaneously by applying a time varying (e.g., pulsed) magnetic field to the portion of the body. For example, in the context of TMS, a time-varying magnetic field may be applied across the skull to create an electric field in the brain tissue, which produces a current. If the induced current is of sufficient density, neuron membrane potential may be reduced to the extent that the membrane sodium channels open and an action potential response is created. An impulse of current is then propagated along the axon membrane which transmits information to other neurons via modulation of neurotransmitters. Such magnetic stimulation has been shown to acutely affect glucose metabolism and local blood flow in cortical tissue. In the case of major depressive disorder, neurotransmitter dysregulation and abnormal glucose metabolism in the prefrontal cortex and the connected limbic structures may be a likely pathophysiology. Repeated application of magnetic stimulation to the prefrontal cortex may produce chronic changes in neurotransmitter concentrations and metabolism so that depression is alleviated.

[0036] Before beginning a therapeutic and / or diagnostic procedure on a human subject, the size and the shape of the human subject’s head may be determined. This determination may be made in order to properly determine where and how the procedure is to be performed on the specific humanNNI 0507 DCVS PCT subject. Since each human subject’s head size and / or shape may be unique, and since the margin of error when determining these locations may be low, accurate means for determining the size and / or shape of the human subject’s head may be a time consuming and delicate procedure. Further, replicating theses determinations for each human subject and / or for each procedure for a human subject may be difficult. As such, the determination of the size and / or shape of the human subject’s head may be performed one time before the first procedure for the human subject and saved for use during subsequent procedures.

[0037] In some examples, the patient may wear a swim cap. Hand measurements may be used to place the cap on the patient’s head in a reproducible way. A mark on the cap edge may be placed above the nasion. The distance to the nasion may be measured to reproduce the position later. The cap may be centered above the patients’ ears. Once the cap is in a fixed location, measurement from anatomical landmarks may be performed to find the initial placement of the coil. One or more of the international EEG 10-20 system landmarks may be used. A grid and / or line of points may be placed relative to the anatomical landmarks to guide the search for the location where stimulation causes an observable response, such as a twitch of a body part like the thumb, fingers, arm and / or foot. Once the motor threshold location and level are found, the stimulation location and / or stimulation level can be saved (e.g., by making a partial outline of the coil on the cap and / or by storing the motor threshold location and / or stimulation level in memory of the treatment system). Then the treatment coil is placed at the treatment location, for instance, by moving the treatment coil a fixed distance relative to motor threshold location in a direction relative to anatomical landmarks or features.

[0038] In some examples, mechanical pointing mechanisms are used. For example, a pointer mechanism is placed in fixed relationship to patient’s head, and lasers and / or mechanical indicators can be used to align to anatomical landmarks. The initial stimulation may be performed by placing the coil relative coil landmarks using anatomical features and / or standard pointer location measurements. The coil rotation can be set and / or measured as angle relative to the pointer direction. Increments in the pointer location may aid the motor threshold search. Finally, increments in the pointer location may be used to find the treatment location. Approximately 10-20 locations may be found by determining fractional distances from anatomical landmarks.NNI 0507 DCVS PCT

[0039] Optical target neuronavigation systems can be used to assist transcranial magnetic stimulation, where markers that can be identified in images are used. The terms “tracker” and “marker” may be used interchangeably herein. The size and / or distortion of the marker in the image may determine the position and / or rotation of the marker relative to one or more cameras. In examples, three trackers are used - one marker mounted on the patient’s head, one mounted to the treatment coil, and one mounted to a pointer. The relationship between the location of the markers, the treatment coil, and the pointer may be measured or determined by construction. The pointer may indicate the relative position of patient landmarks to the patient marker. Distance for equivalent landmarks in medical images may guide the treatment at anatomically and / or physiologically defined targets. In some instances, using markers can be challenging due to their difficulty to attach to the patient, coil, and pointer. Moreover, markers may be bumped and / or moved throughout the procedure.

[0040] In some examples, electromagnetic neuronavigation is also used, where the location of the markers is determined by the size and possibly orientation of electromagnetic fields detected at the tracker from multiple transmission coils or antennae.

[0041] In some examples, surface guided tracking may be used, wherein such instances, depth cameras may measure the shape of the patient’s head. In such instances, multiple camera locations and / or scanning methods may improve the measurement of the patient’s head shape. Facial recognition techniques may identify anatomical landmarks. A tracked pointer may indicate the patient’s head shape with or without the use of a swim cap. The treatment coil may be used as this pointer. A generic model and / or atlas may be distorted to match the measured patient’s head shape. Locations relative to measured head features may then be determined and optical markers may be attached to the coil.

[0042] In some examples, optical Target Neuronavigation systems are used, where a marker is placed on the patient’s head. The accuracy of the system depends on how steadily the marker stays in position. Typically, the markers are placed on glasses or a head band or adhered to the patient’s forehead. The system learns where the marker is relative to the head, so a pointer is used to indicate the key landmarks. This alignment process needs to be repeated every day. The treatment locationNNI 0507 DCVS PCT error has the sum of the errors in landmark selection, pointer location measurement, head target stability, coil target measurement, coil pointer measurement, and / or coil geometry target selection.

[0043] Electromagnetic navigation may use markers and have the same placement and alignment problems found in the optical target neuronavigation systems. These methods use metallic equipment which may distort and / or reflect the electromagnetic signals causing errors.

[0044] Another problem with the measured deviations in the point cloud method is that not all items near the patients head move with the coil. Such items may include, but are not limited to, a head rest, side cushion, coil support arm, power cord, chair, etc. While facial recognition software may determine the head position, it cannot determine the position of these aforementioned items. Therefore, the indicated movement of the coil is not a simple calculation of the deviation of the point cloud shape from the reference point cloud.

[0045] Methods and / or systems configured to provide effective markerless tracking of the treatment coil, treatment location, and / or patient pose are described herein. These methods and / or systems enable the determination of the initial treatment location and the treatment without registering markers to the patient or the treatment coil. One or more cameras are placed in front of the patient with a view of the patient’s face and ears. Facial recognition software and / or a trained model (e.g., a neuro network) may recognize the face and / or ears of the patient. A pointer may demark one or more locations on the patient’s head (e.g., one location near the top of the head and one location on the back of the head). In examples, a generic head model may be used, then scaled, and then translated and / or rotated to match the patient based on the images captured by the cameras and / or the locations captured by the pointer.

[0046] In some examples, instead of using a geometric marker or memorized treatment configuration to track the coil, a trained model can be used to identify the location of the treatment coil within the images. Additionally or alternatively, one or more geometric features on the treatment coil may be identified to detect the location and / or orientation of the treatment coil. For instance, a flat side of the coil may be identified by finding surfaces with normal components in the same direction (e.g., using methods such a random sample consensus (RANSAC) that can be used to select points best representing the surface or geometric feature). The known shape of the treatmentNNI 0507 DCVS PCT coil may determine the location and / or orientation (e.g., roll, tilt, and / or yaw) of the surface and / or geometric feature.

[0047] The distance and direction to the points on the surfaces with normal components in the same direction can be used to determine the position of the treatment coil. The orientation of the normal components (e.g, the orientation of the normal components relative to the patient and other objects in the space) can be used to determine the orientation of the treatment coil. The systems and / or methods described herein may use principal component analysis to determine the location and / or orientation of the treatment coil. A known relationship between the geometric features of the treatment coil and an induced electric field of the treatment coil may display the relative position of the treatment coil to the target location. The treatment system may display the error between the actual location and / or orientation of the treatment coil and the target location and / or orientation of the treatment coil. In examples, the location of the treatment coil may be projected onto a head model in a GUI to assist a technician in positioning the treatment coil to the target location. Further, in some examples, the treatment system may use contact sensors to determine whether the treatment coil properly contacts the head.

[0048] FIG. 1 A is a perspective diagram of an example treatment system. FIG. IB is a block diagram illustrating an example of the treatment system 100. The treatment system 100 may comprise a controller (e.g, the controller 150 of FIG. IB) such as a processor, a power supply (e.g., the power supply 140 of FIG. IB), memory (e.g., memory 180 of FIG. IB), a transceiver, (e.g., the transceiver 190 of FIG. IB), a treatment coil 102, an articulating arm 104, a user interface (e.g., the user interface 130 of FIG. IB), one or more sensors, and / or a human subject positioning apparatus 122. The user interface may include a display device 106. The one or more sensors may include a first camera 110a, a second camera 110b, and / or a contact sensor 170,

[0049] The treatment system 100 may be stationary or movable. For example, the treatment system 100 may be integrated into a movable cart 124, for example, as shown in FIG. 1 A. In one or more examples, the treatment system 100 may be a TMS treatment system (e.g., NeuroStar®) and / or any other therapeutic and / or diagnostic procedure system. The treatment coil 102 may be used to administer a therapeutic and / or diagnostic procedure to a human subject 120, for example, TMS. Although illustrated in FIG. 1A to include a treatment coil 102, the treatment system 100 mayNNI 0507 DCVS PCT include any device for administration of therapeutic and / or diagnostic procedure of the human subject. The treatment system 100 may be used for a diagnostic procedure (e.g., solely for a diagnostic procedure).

[0050] The controller 150 of the treatment system 100 may be a general purpose processor, a special purpose processor, a conventional processor, a digital signal processor (DSP), a plurality of microprocessors, one or more microprocessors in association with a DSP core, a controller, a microcontroller, Application Specific Integrated Circuits (ASICs), Field Programmable Gate Array (FPGAs) circuits, any other type of integrated circuit (IC), a state machine, and the like. The processor may perform signal coding, data processing, power control, input / output processing, and / or any other functionality that enables the treatment system 100 to operate. The processor may be integrated together with one or more other components of the treatment system 100 in an electronic package or chip. Further, in some examples, the one or more processors may be distributed, such as, for instance, located in different physical locations, such as at the treatment system 100 and / or in one or more remote servers (e.g., cloud servers).

[0051] The controller 150 of the treatment system 100 may be coupled to and may receive user input data from and / or output user input data to the treatment coil 102, the articulating arm 104, the display device 106 (e.g., a liquid crystal display (LCD) display unit or organic light-emitting diode (OLED) display unit), the first and second cameras 110a, 110b, and / or the human subject positioning apparatus 122. The controller 150 may access information from, and store data in, any type of suitable memory, such as memory 180, which may be non-removable memory and / or removable memory. The non-removable memory may include random-access memory (RAM), read-only memory (ROM), a hard disk, or any other type of memory storage device. The removable memory may include a subscriber identity module (SIM) card, a memory stick, a secure digital (SD) memory card, and the like. The controller 150 may access information from, and store data in, memory that is not physically located within the treatment system 100, such as on a server (not shown).

[0052] The controller 150 may receive power from the power supply 140. The controller 150 may be configured to distribute and / or control the power to the other components in the treatment system 100. The power supply 140 may be any suitable device for powering the treatment system 100. The power supply 140 may be any type of power source that provides sufficient energy for the treatmentNNI 0507 DCVS PCT coil 102 to generate the pulsing magnetic field 160 for its intended purpose, for example, for TMS, repetitive transcranial magnetic stimulation (rTMS), magnetic seizure therapy (MST) or any other type of application. For example, the power supply 140 may be a conventional 120 or 240 VAC main power source. The controller 150 may be configured to send and / or receive wired or wireless signals using the transceiver 190, which may include a transmitter circuit and a receiver circuit.

[0053] The human subject 120 may be positioned within the human subject positioning apparatus 122. The human subject positioning apparatus 122 may be a chair, recliner, bed, stool, and / or the like. When performing treatment, the treatment coil 102 may be situated such that the human subject 120’ s head is positioned under the treatment coil 102. The treatment coil 102 may be adjusted by means of the articulating arm 104 and / or the like.

[0054] The user interface 130 may be any type of interface in which a user of the treatment system 100 may initiate, adjust, and / or end the magnetic stimulation procedure. For example, the user interface may include a personal computer (PC), a keyboard, a mouse, a touchscreen, a wireless device, and / or the like, that allows for an interface between the user and the treatment system 100.

[0055] The controller 150 may be configured to receive inputs from the user interface 130, the sensor 110, and / or the contact sensor 170 to conduct magnetic stimulation therapy accordingly. For example, the controller 150 may perform signal coding, data processing, power control, input / output processing, and / or any other functionality that enables the controller 150 to operate the magnetic stimulation component for magnetic stimulation. Although the controller 150 of the treatment system 100 may be configured to control the treatment coil 102, the one or more sensors 110, and the contact sensor 170, the treatment system 100 may include two or more controllers for individually controlling two or more of the components of the treatment system 100.

[0056] The treatment system 100 may comprise one or more computer software applications running on the controller 150. The computer software applications may provide a system graphical user interface (GUI) (e.g., a TMS system GUI) on the display device 106. The computer software applications may incorporate workflow management to guide a technician through the therapeutic and / or diagnostic procedure, and / or supervise and / or control one or more subsystems of the treatment system 100. For example, the computer software applications may control internal system functions, monitor the system status to ensure safe operation, and / or provide the user with aNNI 0507 DCVS PCT graphical means to manage the preparation for and / or the administration of the therapeutic and / or diagnostic procedures.

[0057] Interaction with the computer software applications may be provided via the user interface 130. In examples, the user interface 130 may be the display device 106, (e.g., a touch screen display). The display device 106 may include touch activated images of alphanumeric keys and / or buttons for user interaction with the treatment system 100. The display device 106 may provide graphic representations of the system activity, messages, and / or alarms. Interactive buttons, fields, and / or images may be displayed via the display device 106, and may enable the technician to direct and / or interact with system functions, for example, such as entering data, starting and / or stopping the procedure, running diagnostics, adjusting positioning and / or configuration of the treatment coil 102, adjusting the position of one or more sensors 110 (e.g., the first and second cameras 110a, 110b), and / or the like.

[0058] As described in more detail herein, the treatment system 100 may include more than one camera 110a, 110b. The first and second cameras 110a, 110b may be located and / or orientated at different locations within the room such that the cameras 110a, 110b are configured to capture images that provide varying perspectives of the human subject 120 and / or the treatment coil 102. The use of multiple cameras may allow the treatment system 100 to capture the entirety of the treatment coil 102 and / or the head of the human subject 120 in the images captured by the cameras 110a, 110b without having to move the cameras 110a, 100b. Further, the use of multiple cameras may help to ensure that the images captured by the cameras 110a, 110b include treatment coil 102 and / or the head of the human subject 120 irrespective of where the technician is located within the room.

[0059] Each of the first and second cameras 110a, 110b may include any combination of sensors, such as one or more depth cameras e.g., red-green-blue depth (RGB-d) camera or a greyscale depth camera), infrared (IR) sensors (e.g., IR camera), ultrasonic transducers, image sensors, color sensors, light sensors, radio-frequency (RF) sensors, accelerometers, orientation sensors, microphone arrays, laser scanners, and / or the like. When the depth camera is an RGB-d camera, the depth camera may capture both color images (e.g., RGB images) and depth information at the same time. The depth information may provide a per-pixel distance measurement between the camera and object(s) in theNNI 0507 DCVS PCT scene, which may allow for three-dimensional (3D) perception. When the depth camera is a greyscale depth camera, the depth camera may capture depth images (e.g., or a depth map) encodes a distance from the camera to object(s) in the scene for each pixel. The depth images from the greyscale depth camera may be visualized in grayscale (e.g., where closer objects appear lighter or darker, depending on the convention). The depth images may store or indicate each pixel as a single value representing depth (e.g., distance in meters, or a normalized grayscale value). As described in more detail below, the treatment system 100 may be configured to generate the point cloud based on the depth images captured by the first and second cameras 110a, 110b.

[0060] In some examples, the treatment system 100 may be configured to assign one of the first camera 110a or the second camera 110b as a master camera. For example, the treatment system 100 may establish one of the first camera 110a or the second camera 110b as the master camera to avoid flipping back and forth constantly between cameras. For example, in some scenarios, there can be inconsistency and / or jitter if the treatment system switches the camera that is being used to located and track the treatment coil 102. The treatment system may be configured to determine a confidence metric that indicates how confident the treatment system is that the treatment system has identified the treatment coil 102 in one or more images (e.g., depth images) generated by a camera (e.g., each of the first and second cameras 110a, 110b). Accordingly, the treatment system may be configured to determine a first confidence metric that indicates a confidence level that the treatment coil 102 was identified in one or more images (e.g., depth images) generated by the first camera 110a, and determine a second confidence metric that indicates a confidence level that the treatment coil 102 was identified in one or more images (e.g., depth images) generated by the second camera 110b. The treatment system may assign the first camera 110a or the second camera 110b the master based on which camera has the higher confidence rating. Then, moving forward throughout the procedure, the treatment system may continue to track and monitor the treatment coil 102 using the camera that is assigned as the master (e.g., and / or based on the respective confidence metrics for the first and second cameras 110a, 110b, which for instance, may change and update throughout the procedure in some examples). In situations where the treatment system does not detect the treatment coil 102 in the images generated by one camera, or when there is a string of low-confidence metrics determinations for a particular camera, the treatment system may switch to using the other camera for detection of the treatment coil 102 (e.g., assign the other camera as the master).NNI 0507 DCVS PCT

[0061] Although illustrated as fixed devices in FIG. 1 A, the first and second cameras 110a, 110b may be mobile. The first and second cameras 110a, 110b may communicate with the controller 150 of the treatment system 100 via a wired interface (e. , as shown in FIG. 1 A) and / or wireless interface (e.g., radio frequency (RF) communication, such as, but not limited to WiFi®, Bluetooth®, LTE®, and / or the like). In some examples, the first and second cameras 110a, 110b may be different from one another. Further, a depth camera may be configured to determine the distance between two objects (e.g., determine the distance between the camera itself and a target or between two objects within the space). For instance, types of depth cameras include stereo sensors (e.g., that includes two cameras and are configured to determine depth based on disparity information between the images captured by both cameras), time-of-flight sensors (e.g., that are configured to calculate depth based on photon transmission and reception times), light detection and ranging (LiDAR) sensors, structured light (SL) sensors, and / or the like. In other examples, the treatment system 100 may be configured to determine the depth of an object in an image based on the distortion of the projected image (e.g., where the distortion can be used to infer depth).

[0062] Although two cameras 110a, 110b are illustrated in FIG. 1 A, the treatment system 100 may comprise a single camera or a plurality of cameras (e.g., more than two cameras). Further, in some examples, the first and second cameras 110a, 110b may include an inertial measurement unit (IMU) that is configured to provide information about camera motion from frame to frame, which may be used to predict location of one or more objects in the fames, such as the patient and treatment coil 102. For examples, the treatment system 100 may be configured to further determine the position of the patient and / or the treatment coil 102 in a video frame based on feedback from an IMU of one or more of the first and second cameras 110a, 110b.

[0063] The treatment system 100 (e.g., the controller 150 of the treatment system 100) may receive a plurality of images from the first and second cameras 110a, 110b. The images may include images of the treatment coil 102 and of a head of a human subject 120. As described in more detail herein, the treatment system 100 may use the cameras 110a, 110b (e.g., the images captured by the cameras) to create a two-dimensional (2D) and / or a three-dimensional (3D) digital reconstruction model of the human subject 120’s head (e.g., a head model) and / or a 2D and / or 3D digital reconstruction of the treatment coil 102 (e.g., a coil model). When depth camera(s) are used, the output of the depth camera may be depth images. A depth image may be used to generate a point cloud possibly, forNNI 0507 DCVS PCT example, with addition information included with the point cloud (e.g., and additionally may include color and intensity information associated with each point cloud). For example, the treatment system 100 (e.g., or another system) may be configured to convert a depth image to a point cloud based on a depth value associated with one or more pixels of the depth image. For instance, the treatment system 100 may project the depth value into a 3D coordinate using one or more characteristics of the camera (e.g., focal length, principal point, and / or depth scale), and in some examples, based on a transformation library (e.g., Open3D). In some examples, the point cloud of the human subject 120 may be defined by a Standard Tessellation Language (STL) file that, for instance, may be a 3D model file format that represents the surface geometry of an object (e.g., the human subject 120) using a collection of connected triangles, called a mesh.

[0064] The treatment system 100 may determine the size and / or shape of a human subject’s head to generate a fitted head model, (e.g., a fitted head model of the human subject 120). The treatment system 100 may generate the fitted head model using a predetermined head model and one or more points that are determined using the first and second cameras 110a, 110b. The predetermined head model may be a generic head model that does not include any characteristics specific to one individual. The predetermined head model may be used, for example, to keep the anonymity of the human subject 120, to provide information relating to one or more predefined coordinates, to reduce the total number of artifacts in the fitted head model, to reduce or eliminate the asymmetric nature or abnormal features of the human subject 120 (e.g., loss of ear, asymmetric bump, etc.), and / or the like. Example procedures for generating a fitted head model using a point cloud and a predetermined head model are illustrated in FIGs. 6 and 7, respectively.

[0065] The fitted head model may be a 2D and / or 3D model. The 2D and / or 3D digital reconstruction of the human subject’s head may include a fitted head model. For example, the cameras 110a, 110b (e.g., the images captured by the cameras 110a, 110b) may be used to determine one or more points (e.g., the cloud of points) associated with the human subject’s head and / or the treatment coil 102. In examples where the cameras 110a, 110b are depth cameras, the plurality of images captured by the cameras 110a, 110b may be depth images. The treatment system 100 may identify the points associated with the face of the human subject 120 using one or more facial recognition techniques (e.g., a Viola-Jones object detection framework, a 3D facial recognition technique (e.g., Apple Face ID), and the like). The terms facial recognition techniques and imageNNI 0507 DCVS PCT processing techniques may be used interchangeable herein. In some examples, the treatment system 100 may receive images from a plurality of cameras, and as such, the treatment system 100 may detect the face of the human subject 120 in images from a plurality of cameras. To do so, the treatment system 100 may perform a post-processing technique to ensure that the resulting point cloud that represents the face of the human subject 120 is consistent between the images. Such postprocessing techniques may include averaging points from the images of a plurality of cameras and / or using the points from one camera (e.g., only one camera) if it is a location that is best (e.g., or only) identifiable from the images of a single camera.

[0066] The treatment system 100 may generate the 2D and / or 3D digital reconstruction of the human subject’s head using the points associated with the face of the human subject 120 and, in some examples, using a predetermined head model. The treatment system 100 may use a RANSAC technique to select the points that form the point cloud for the head model and / or the treatment coil 102. The treatment system 100 may use images captured by the one or more sensors 110, such as first and second cameras 110a, 110b, to generate the head model. Although primarily described in context of generating a fitted head model based on an existing, preconfigured head model, in some examples the treatment system 100 may generate the head model using (e.g., using only) the images captured from the first and second cameras 110a, 110b. The treatment system 100 may select from a library of different head models to select the one that best fits the human subject 120 (e.g., with minimal manipulation). The library of different head models may be pre-existing, pre-defined, and / or otherwise previously known and / or available to the treatment system 100.

[0067] In some examples, the treatment system 100 may use principal component analysis in selecting points that form a point cloud (e.g., alternatively or additionally, may use another measure of the orientation of a non-rotationally symmetric surface). The coil model may be matched to the point cloud using a point cloud registration method (e.g., iterative closest point). The head model may be used to determine a treatment location relative to a memorized treatment location. The head model may comprise medical images. The medical images may be overlayed and / or combined. The medical images may be derived from an MRI and / or a functional MRI with overlay with structural images. The medical images may be functional connectivity and / or resting state functional MRI images. The medical images may comprise positron emission tomography (PET) scan images, single photon emission computed tomography (SPECT) images, and / or ultrasound images.NNI 0507 DCVS PCT

[0068] As described in more detail herein, the fitted head model may be used to assist in a therapeutic and / or diagnostic procedure of the human subject 120. The treatment system 100 may store the fitted head model in memory. For example, the treatment system 100 may use polygonal mesh (e.g., as a mathematical structure) to store the fitted head model.

[0069] The first and second cameras 110a, 110b may be used to register one or more anatomical landmarks associated with the human subject’s head. For example, the treatment system 100 may identify one or more anatomical landmarks based on captured images (e.g., based on the cloud of points determined from the captured images). An anatomical landmark may comprise a nasion, an inion, a lateral canthus, an external auditory meatus (e.g., ear attachment point), such as the left and right tragi, and / or one or more preauricular points of the human subject 120. An anatomical landmark may be associated with an x coordinate, a y coordinate, and a z coordinate of the head model.

[0070] The first and second cameras 110a, 110b may determine (e.g., capture) a cloud of points and / or perform anatomical landmark registration of the human subject’s head with or without the use of an indicator tool. The indicator tool may be a second sensor, the finger of the technician, an additional tool, etc. The first and second cameras 110a, 110b may use the indicator tool to identify one or more points that are used to generate the 2D and / or 3D digital reconstruction of the human subject’s head. The indicator tool may include active components (e.g., a LED) and / or passive components (e.g., a reflector) to aid in detection by the first and second cameras 110a, 110b. For example, a reflector may be an example of a passive component / tool, while an LED may be an example of an active component / tool.

[0071] In some examples, the treatment system 100 may include an indicator tool that may be used to identify one or more specific locations on the human subject’s head that are difficult for the cameras to detect, such as the top of their head (e.g., beneath their hair) and / or the back of their head. The treatment system 100 may generate the 2D and / or 3D digital reconstruction of the human subject’s head based on the locations identified using the indicator tool and the images captured by the cameras 110a, 110b (e.g., and image processing techniques that identify one or more facial features of the human subject 120).NNI 0507 DCVS PCT

[0072] The treatment system 100 may determine the location and / or the position of the treatment coil 102 based on the images captured by the cameras 110a, 110b. The treatment system 100 may perform markerless treatment coil identification. As noted herein, markers on the treatment coil 102 may be marred, distorted, and / or moved during the treatment process, thereby creating issues when performing neuronavigation. The treatment system 100 may identify the location and / or orientation of the treatment coil 102 without the use of any markers attached to the treatment coil 102. For example, the treatment system 100 may determine a surface of the treatment coil 102 within an image of the plurality of images based on a direction of a plurality of vectors within the image.

[0073] In some examples, the treatment system 100 may determine the location and / or the position of the treatment coil 102 based on the images (e.g., depth images) captured by the cameras 110a, 110b using a plane fitting algorithm. The treatment system 100 may use the plane fitting algorithm to identify a surface or side of the treatment coil 102 in one or more images (e.g., depth images) to, for example, segment the treatment coil 100 from a cloud of points identified in the space. In some examples, the treatment system 100 may detect a subset of the points of the point cloud (e.g, generated from the depth images) that make up the treatment coil 102 based on the detection of a surface of the treatment coil 102.

[0074] The treatment system 100 may detect a plurality of vectors associated with the treatment coil 120 using one or more image processing techniques (e.g, a plane fitting algorithm, such as support vector machines (SVM), RANSAC, LAD, a Viola-Jones object detection framework, and / or other versions of these techniques, etc.). The treatment system 100 may detect vectors that are at a point on a curve, and the treatment system 100 may decompose the vector as a sum of two vectors: one tangent to the curve, called the tangential component of the vector, and another one perpendicular to the curve, called the normal component of the vector. The treatment system 100 may identify a surface of the treatment coil 102 (e.g, a flat surface of the treatment coil 102) based on the detection of a plurality of vectors that all point in the same direction. For instance, the treatment system 100 may detect a flat surface 103 (e.g, substantially flat surface) of the treatment coil 102 based on the identification of a surface that has normal components in the same direction. The treatment system 100 may determine the orientation (e.g, roll, tilt, and / or yaw) of the vectors to identify the treatment coil 102. For example, the treatment system 100 may perform a principal component analysis or similar methods to determine the orientation of the flat surface 103 of the treatment coil 102 basedNNI 0507 DCVS PCT on the vectors (e.g., if flat surface 103 is not circular). Using this information and / or the direction and distance to the flat surface(s), the treatment system 100 may determine the location of the treatment coil 102 (e.g., the x-y-z coordinates) and / or determine the position of the treatment coil 102 (e.g., the orientation, or the tilt, yaw, or roll, of the treatment coil 102, such as the relative position of the flat surface 103 of the treatment coil 102) based on the plurality of vectors that are determined based on the images captured by the cameras 110a, 110b. Finally, the treatment system 100 may display, via a GUI on the display device 106, the location and the position of the treatment coil 102 relative to a target location on the head of the human subject 120. The target location may be the treatment location or a location (e.g., the motor threshold location) used to determine the treatment location.

[0075] As used herein, location of the treatment coil may refer to the location of the treatment location in a three-dimensional space (e.g., the x-y-z coordinates of the treatment location). Further, as used herein, the position of the treatment coil may refer to the orientation (e.g., roll, tilt, and / or yaw) of the treatment coil. In some examples, the terms position and orientation are used interchangeably.

[0076] In some examples, the treatment system 100 may determine the location of the treatment location based on one or more of the anatomical landmarks of the human subject 120. For example, the treatment system 100 may determine a front-to-back size of the head of the human subject 120 based on a distance between the nasion and inion. The treatment system 100 may determine a left- to-right size of the head of the human subject 120 based on a distance between the tragi. The treatment system 100 may determine a vertical size of the head of the human subject 120 based on a distance between the top of the head and a plane fit to the other four anatomical landmarks (e.g., the nasion, the inion, and the left and right tragi). These measurements are compared to an Montreal Neurological Institute (MNI) head model to determine one or more scale factors. The treatment system 100 may use the scale factors to scale the MNI model to the specific size and dimensions of the head of the human subject 120. Once the MNI model is scaled, the treatment system 100 can take any point input (e.g., in either Talairach, 10-20, or MNI coordinates) and map that point input to the scaled model that is specific to the human subject 120, which can then be projected onto the closest point on the patient model.NNI 0507 DCVS PCT

[0077] The treatment system 100 may be configured to determine whether the treatment coil 102 is in contact with the head of the human subject 120. For example, the treatment system 100 may include the contact sensor 170 that, for example, may be located on a bottom surface of the treatment coil 102 (e.g, the surface of the treatment coil 102 that is intended to be in direct contact with the head of the human subject 120 during treatment). The contact sensor 170 may be located on (e.g., affixed to) the bottom surface (not shown), or treatment faces (not shown), of the treatment coil 102, so as to be adjacent the patient’s head when the treatment coil 102 is in the desired position. As such, the treatment system 100 may determine whether the treatment coil 102 is in contact with the head of the human subject 120 based on the signals received from the contact sensor 170. The location of the contact between the treatment coil 102 and the human subject 120 may be used by the system to determine information about the orientation of the treatment coil 102. Further, the treatment system 100 may display, via the GUI on the display device 106, an indication of whether the treatment coil 102 is in contact with the head of the human subject 120. The contact sensor 170 may be configured to detect a change in force between a surface of the treatment coil 102 and the patient. Examples of the contact sensor 170 that may be used in the treatment system 100 are described in United States Patent No. 8,177,702, which is incorporated herein by reference in its entirety.

[0078] In some examples, the treatment coil 102 may include a sensor that is configured to measure the orientation of the treatment coil 102 (e.g., an accelerometer, a magnetometer, and / or a gyroscopes). The sensor of the treatment coil 102 may be configured to measure the orientation of the treatment coil 102 with respect to gravity, earth’s magnetic field, or a human subject positioning apparatus 122. The treatment system 100 may be configured to determine the orientation of the treatment coil 102 based on feedback from the sensor of the treatment coil 102. Further, in some examples, the treatment system 100 may be configured to determine the position or orientation of the treatment coil 102 based on an inertial measurement unit of the treatment coil 102 (e.g, based on feedback from an accelerometer of the treatment coil 102, which may indicate how the treatment coil 102 and / or the patient’s head has moved between images (e.g., in addition to determining the position or orientation of the treatment coil 102 based on image analysis).

[0079] In some examples, the treatment system 100 may determine the location of the treatment coil 102 based on a relationship between one or more geometric features of the treatment coil 102 and anNNI 0507 DCVS PCT induced electric field of the treatment coil 102. For instance, the treatment coil 102 may induce an electric field that extends beyond the geometric extent of the treatment coil 102. The spatial distribution of the induced electric field and related currents may be distorted by the patient. Accordingly, the induced electric field can change depending on the location and orientation of the patient. The treatment system 100 may determine a geometric relationship between the treatment coil 102 and the induced electric field, and determine the location of the treatment coil 102 accordingly.

[0080] The treatment system 100 may be used for any therapeutic and / or diagnostic procedure. For example, the treatment system may be used for TMS, transcranial direct current stimulation (tDCS), electroencephalogram (EEG), Deep brain stimulation (DBS), a diagnostic procedure, and / or the like. For example, the treatment system 100 may be used for any therapeutic and / or diagnostic procedure that includes the placement of electrodes, sensors, probes, and / or the like on a human subject 120, such as on the surface of a human subject’s head. Although described with reference to a head model, the treatment system 100 may be configured to generate a model of any part of the human subject 120, such as, but not limited to, the arm, neck, chest, leg, and / or the like. The treatment system 100 may generate the fitted head model using the cameras 110a, 110b and / or detect the location of the treatment coil 102 relative to a target location on the head of the human subject 120 using the cameras 110a, 110b. In examples, the treatment system may generate a model of the human subject 120’ s body or specific body part other than the head. In examples, the treatment system 100 may detect the location of the treatment coil 102 relative to a target location on the human subject 102’s body or specific body part other than the head.

[0081] The treatment system 100 may measure and / or report (e.g., continuously measure and / or report) the relative position of one or more objects (e.g., the treatment coil 102, the head of the human subject 120, etc.) for display (e.g., real-time display) to the technician. For example, the treatment system 100 may measure the relative location of treatment coil 102 with respect to the human subject’s head. The treatment system 100 may determine the location of the treatment coil 102 that is in contact with the human subject’s head (e.g., which portion(s) of the treatment tool are in contact with the head) based on feedback from the contact sensor 170. The treatment system 100 may use treatment tool’s location information to determine if procedure is being performed properly. For example, for TMS, if the treatment coil 102 falls outside of a predefined range with respect toNNI 0507 DCVS PCT the human subject’s head (e.g., is too far or too close to the human subject’s head), then the treatment system 100 may alert the technician and / or alter or stop the therapeutic and / or diagnostic procedure.

[0082] TMS may refer to TMS, repetitive transcranial magnetic stimulation (rTMS), deep TMS (dTMS), controlled and / or pulse shape TMS (cTMS), or the like. The treatment coil 102 may include a single treatment coil, multiple treatment coils and / or an array of treatment coils. The treatment area may be the prefrontal cortex, for example. The treatment coil 102 may or may not include a core, such as a magnetic core (e.g., ferromagnetic core), for example. The pulsing magnetic field 160 may include one or more pulse bursts. A pulse burst e.g., each pulse burst) of the pulsing magnetic field 160 may include one or more pulses.

[0083] FIG. 2 is a diagram of an example of a sensor 200. The sensor 200 may be used with a treatment system (e.g., the treatment system 100). The sensor 200 may be an example of a camera 110a, 110b of the treatment system 100. The sensor 200 may comprise a processor 201, memory 202, a power supply 203, an IR emitter 204, an IR depth sensor 205, a color sensor 206, an orientation monitor 207, a microphone array 208 (e.g., which may comprises one or more microphones), and / or other peripherals 209 (e.g., a transceiver). In some examples, the sensor 200 may comprise a depth camera, such as a red-green-blue depth (RGB-d) camera. In an example, the sensor 200 may comprise the Kinect® system made by Microsoft® and / or an equivalent 3D IR depth sensing camera. The sensor 200 (e.g., and / or the controller 150 of the treatment system 100) may include software that can be used to perform facial recognition, motion capture, video recording, and / or the like. The sensor 200 may include a user interface (e.g., a touch screen) and / or the sensor 200 may use the user interface (e.g., the display device 106) of the treatment system. The sensor 200 may be a fixed or mobile device. The sensor 200 may communicate with the controller 150 of the treatment system 100 via a wired interface and / or a wireless interface.

[0084] The processor 201 of the sensor 200 may be a general purpose processor, a special purpose processor, a conventional processor, a digital signal processor (DSP), a plurality of microprocessors, one or more microprocessors in association with a DSP core, a controller, a microcontroller, ASICs, FPGA circuits, any other type of IC, a state machine, and the like. The processor 201 may perform signal coding, data processing, power control, input / output processing, and / or any other functionalityNNI 0507 DCVS PCT that enables the sensor 200 to operate. The processor 200 may be integrated together with one or more other components of the sensor 200 in an electronic package or chip. For example, the treatment system 100 and sensor 200 may share the same processor.

[0085] The processor 201 may be coupled to and may receive user input data from and / or output user input data to the memory 202, the power supply 203, the infrared (IR) emitter 204, the IR depth sensor 205, the color sensor 206, the orientation monitor 207, the microphone array 208, and / or other peripherals 209. The processor 201 may access information from, and store data in, memory 202, such as non-removable memory and / or removable memory. The non-removable memory may include random-access memory (RAM), read-only memory (ROM), a hard disk, or any other type of memory storage device. The removable memory may include a subscriber identity module (SIM) card, a memory stick, a secure digital (SD) memory card, and the like. The processor 201 may access information from, and store data in, memory that is not physically located within the sensor 200, such as on the treatment system 100 and / or on a server (not shown), etc.

[0086] The processor 201 may receive power from the power supply 203. The processor 201 may be configured to distribute and / or control the power to the other components in the sensor 200. The power supply 203 may be any suitable device for powering the sensor 200.

[0087] FIG. 3 is a flowchart of a procedure 300 for determining the location and the orientation of the treatment coil for TMS. The procedure 300 may be executed by a controller of a treatment system, such as the controller 150 of the treatment system 100. The treatment system may perform the procedure 300 in response to a user input, such as in response to a technician who is setting up a treatment procedure for a human subject.

[0088] At 302, the treatment system may receive a plurality of images from one or more sensors (e.g, the sensors 110) and / or cameras (e.g, the first and second cameras 110a, 110b). The plurality of images may comprise images of the treatment coil (e.g., the treatment coil 102) and / or of a head of a human subject (e.g., the human subject 120). In some examples, the treatment system may capture or detect a plurality of points (e.g., a point cloud), and detect the bounds of the face of the human subject in each image captured by the cameras. The treatment system may detect face point landmarks within the images using facial recognition techniques. In some examples, the treatment system may determine a feature of the human subject based on a facial recognition algorithm, andNNI 0507 DCVS PCT determining the location and / or orientation of the human subject based on the feature of the human subject. In such examples, the treatment system may use facial recognition to downsample the one or more images that comprise the human subject (e.g, for faster surface registration).

[0089] Further, the treatment system may identify one or more points associated with the head of the human subject using an indicator tool. For instance, the indicator tool may include a quick response (QR) code or other identifiable image on the tool so that it can be easily identified within the images by the treatment system. Further, in some examples, the camera may include an infrared camera (e.g., a time-of-flight (TOF) camera) and the images may be infrared images, and the system may determine one or more points associated with the head of the human subject based on the infrared images. In such examples, the system may determine the one or more points from infrared images captured using a time-of-flight depth camera.

[0090] Further, as described herein, the system may create a 2D and / or 3D digital reconstruction of the human subject’s head based on the images captured by the cameras (e.g, based on the plurality of points associated with the head of the human subject). The 2D and / or 3D digital reconstruction of the human subject’s head may include a fitted head model. For example, the sensors and / or cameras (e.g, the images captured by the cameras) may be used to determine one or more points (g.g, the cloud of points) associated with the human subject’s head and / or the treatment coil. Examples of methods that can be used to generate a fitted head model using a plurality of points that are determined based on images captured by cameras are described in more detail with respect to FIGs.6 and 7 below. Further, the treatment system may register one or more anatomical landmarks associated with the human subject’s head on the head model.

[0091] At 304, the treatment system may determine the treatment coil within an image of the plurality of images. The treatment system may determine the location and / or orientation of the treatment coil. In some examples, the treatment system may determine a subset of points that represent the treatment coil from the plurality of points that are associated with an image, such as a depth image (e.g., segment the treatment coil from the cloud of points identified in the space). For instance, the treatment system may detect a subset of the points of the point cloud (e.g, generated from the depth images) that make up the treatment coil based on the detection of a surface of the treatment coil. For example, the system may determine a surface of the treatment coil within anNNI 0507 DCVS PCT image of the plurality of images based on a direction of a plurality of vectors within the image. For example, the treatment system may determine a surface of the treatment coil within an image of the plurality of images based on a direction of a plurality of vectors within the image. For example, the treatment system may detect a plurality of vectors associated with the treatment coil using one or more image processing techniques (e.g., support vector machines (SVM), a Viola-Jones object detection framework, etc.). The treatment system may detect vectors that are at a point on a curve, and the treatment system 100 may decompose the vector as a sum of two vectors: one tangent to the curve, called the tangential component of the vector, and another one perpendicular to the curve, called the normal component of the vector. The treatment system may identify a surface of the treatment coil e.g., a flat surface of the treatment coil, such as the flat surface 103) based on the detection of a plurality of vectors that all point in the same direction. For instance, the treatment system may detect a flat surface of the treatment coil based on the identification of a surface that has normal components in the same direction. In some examples, the treatment system may determine the position of the treatment coil based on the distance and direction to the flat surface of the treatment coil.

[0092] In some examples, the treatment system may detect the treatment coil using a trained machine learning model (e.g., as described herein). In some examples, the machine learning model may include a deep neural network (DNN). In some instances, a training data set was developed using a prototype treatment system, where the prototype system captured images with different patients and / or backgrounds. The captured images could be manually annotated and / or annotated by a model to identify the location of the coil in each image (e.g., by identifying the specific points within each image that made up the treatment coil). The DNN was trained to reliably indicate the correct location of the coil within the images from the multiple cameras in the prototype. For instance, the treatment system may locate the coil in the one or more images using the machine leaning model. The treatment system may determine if the coil is identified in the images. If the coil is not identified in the images, the treatment system may fit the image and / or coil to a multi-task neighborhood interaction (MNN) model. If the coil is identified in the images, the treatment system may estimate the coil direction using the machine leaning model. Further, in some examples, the treatment system may determine the location of the treatment coil based on one or more previously determined locations of the treatment coil (e.g., location of the treatment coil as determined usingNNI 0507 DCVS PCT previous images). Further, in some examples, the treatment system may determine the location and / or position of the treatment coil in one or more images based on a surface registration technique, for instance, between the point cloud and a geometry of the treatment coil (e.g, using an iterative closest point method).

[0093] In some examples, a trained machine learning model may have been trained to identify the location of the treatment coil and the head of the human subject in one or more two-dimensional images. In such examples, the treatment system may determine the location of the treatment coil based on such information, which for example, may be used to narrow a search in the depth images for determining the location and orientation of the treatment coil and / or the head of the patient.

[0094] The treatment system may perform image segmentation to partition the images into multiple segments. The treatment system may find normal vectors based on local points near the treatment coil (e.g., after the treatment coil is identified). The treatment system may obtain a subset of segment candidates (e.g., approximately three to seven) based on the normal vector uniformity. The treatment system may run a RANSAC to fit a plane to each segment candidate. The treatment system may select the segment candidate(s) that has the size and / or eccentricity expected of the treatment coil based on the distance and / or orientation of the normal vectors associated with the segment candidate(s). The treatment system may determine whether a suitable segment candidate is found, and if so, the treatment system may align the coil model to the segment candidate’s principal components and fit the point cloud to the known geometry of the treatment coil. If a suitable surface (e.g., flat surface) candidate is not found, the treatment system may align the coil model with an Iterative Conformal Prediction (ICP) initialized at a previous transform, and / or fit the image and / or treatment coil to a model (e.g, the MNN model). Additionally or alternatively, a marker rigidly attached to the treatment coil can be used as a backup method of finding the coil in the image and, if necessary, to determine the location. This backup method may be necessary if an operator’s hand or the patient’s hair obscures the view of the treatment coil.

[0095] As noted herein, the treatment system may generate a 2D and / or a 3D model of the treatment coil based on one or more points associated with the treatment coil. For instance, the treatment system may execute RANSAC to fit a plane of points and normal vectors associated with the treatment coil. The treatment system may determine if the plane dimensions match the dimensionsNNI 0507 DCVS PCT of the treatment coil (e.g., a surface, such as a flat surface, of the treatment coil), and if so, treatment system may align the model of the treatment coil to the plane via principal components and / or other fitting methods.

[0096] At 306, the system may determine an orientation of the treatment coil. For example, the system may determine the orientation of the treatment coil based on the plurality of vectors (e.g., normal vectors) identified at 304. As such, the treatment system may determine the location of the treatment coil (e.g., the x-y-z coordinates) and / or determine the orientation of the treatment coil (e.g., the tilt, yaw, or roll of the treatment coil), such as the relative position of the flat surface of the treatment coil, based on the plurality of vectors that are determined based on the images captured by the cameras. In some examples, the treatment system can determine the location of the treatment coil based on the distance and / or direction between the points of the point cloud (e.g., points of the depth images) and the camera(s) that are identified as being part of the treatment coil. The visible size and shape of the surface (e.g., flat surface) may assist in determining the distance and orientation of the treatment coil.

[0097] At 308, the system may display, via a GUI on a display (e.g., the display device 106), the location and the orientation of the treatment coil relative to a target location on the head of the human subject (e.g., the human subject 120). For example, the system may generate a display that includes a digital representation of the treatment coil relative to a digital representation of the human subject’s head (e.g., the head model) via the GUI, examples of which are described herein. The target location could be the treatment location or a location (e.g., the motor threshold (MT) location) that is used to determine the treatment location.

[0098] As such, the system may generate a GUI on the display device that includes the head model (e.g., the 2D and / or 3D digital reconstruction of the human subject’s head) and / or a digital representation of the location and / or orientation of the treatment coil relative to the head of the human subject. The system may be configured to track and / or update the GUI (e.g., in real-time) based on the feedback from the sensors (e.g., the cameras 110a, 110b and / or the contact sensors 170). For example, the system may provide continuous tracking of the treatment coil (e.g., in realtime) by tracking the position and / or orientation of the treatment coil during a procedure based on the images captured by the cameras of the system (e.g., and / or without having markers attached toNNI 0507 DCVS PCT the treatment coil). The treatment system may generate a notification (e.g., audible alert, GUI on the display device, etc.) to indicate to the technician that the treatment coil is outside of the target location and / or orientation and / or no longer in contact with the head of the human subject.

[0099] Further, as described in more detail below, the treatment system may detect the face of the human subject based on the one or more images that were captured (e.g., the treatment system may determine a position of the face and / or one or more facial features of the human subject based on a point cloud). The treatment system may generate a fitted head model based on the images captured by the one or more cameras.

[0100] The treatment system may determine one or more target locations on the fitted head model using the anatomical landmarks. A target location may be a treatment location or a reference point that is used to determine a treatment location, for example, the MT location and / or treatment location of the human subject. The treatment system may determine a target location on the fitted head model based on one or more anatomical landmarks without the use of additional image information, for example, such as, but not limited to an MRI image, a CT image, an X-ray image, and / or the like. In some examples, the treatment system may use an indicator tool to indicate the treatment location or other target on the head model. The treatment system may use the model head, which may be adjusted for the patient based on the fitting, to determine the treatment location. A treatment location(s) can be found relative to one or more other location(s) based movements relative to identified locations. These identified locations may be automatically generated by facial recognition, ear recognition, the use of a pointer, and / or similar methods.

[0101] Anatomical landmarks may be used to assist in neuronavigation of the human subject’s head. For example, the treatment system may display the fitted head model, the treatment coil, the anatomical landmarks, and / or a target location on the display device, for example, to assist in a therapeutic and / or diagnostic procedure. For example, the treatment system may perform TMS using the fitted head model and / or the target location. The treatment system may save the fitted head model, along with the anatomical landmarks and / or target locations, for example, so that it may be recalled and / or used during a subsequent therapeutic or diagnostic procedure.

[0102] The treatment system may include software that may perform neuronavigation, for example, to assist in the therapeutic and / or diagnostic procedure of the human subject. For example, theNNI 0507 DCVS PCT treatment system may comprise software that may perform neuronavigation for TMS treatment. Neuronavigation may refer to the procedure by which the treatment system allows for a technician to navigate in and / or around a head, a vertebral column, and / or other body part of a human subject, for example, before and / or during a procedure. For example, spatial information that may be part of the fitted head model may be used by the software for neuronavigation. The neuronavigation may be performed in real-time. The neuronavigation may track one or more objects, for example, simultaneously. As such, the treatment system may be used to perform neuronavigation based on 3D IR object recognition and motion tracking.

[0103] FIGs. 4A-4C depict example graphical user interfaces (GUIs) that indicate an example fitted head model, an example indication of the current treatment coil pose (e.g., location and orientation), and an example indication of a target coil position and orientation. FIG. 4A depicts an example diagram 400 as displayed by a GUI. The diagram 400 includes the head model 409 of the human subject and one or more indicators, such as a treatment coil indicator 402 and a target location indicator 406. The treatment coil indicator 402 indicates the present location and orientation of a treatment coil, such as the treatment coil 102. The coil location may be projected to the graphical representation of the head model to make positioning easier, for example, particularly where a contact sensor (e.g., a contact sensor 170) is present to assure the coil is close to the head surface. For example, the treatment coil indicator 402 may include a location indicator 403 that indicates the location of the treatment coil relative to the head of the human subject. The treatment coil indicator 402 may also include an orientation indicator 404 that indicates the orientation (e.g., tilt, roll, or yaw) of the treatment coil relative to the head of the human subject. The target location indicator 406 includes a location indicator 407 that indicates the location of a target location on the head model of the human subject, and an orientation indicator 408 that indicates the tilt (e.g., tilt, roll, or yaw) that a treatment coil needs to be placed relative to the target location. The target location may be a treatment location or a location that is used to determine the treatment location, such as the motor threshold location of the patient. The target location indicator 406may be determined based on a known location in a head model. As shown in greater detail in FIG. 4B, when the treatment coil indicator 402 aligns with the target indicator 406, the treatment coil is placed at the proper location and orientation for treatment.NNI 0507 DCVS PCT

[0104] As noted herein, the treatment system may be used to help a technician position the treatment coil relative to the target location, for example, using neuronavigation, such as the GUIs shown in FIG. 4A-4C. The treatment system may be configured to adjust the color of the target location indicator 406 to indicate whether or not the treatment coil is located at or close to the target location and / or orientation relative to the target location. Alternatively or additionally, the treatment system may use the distance between the location of the treatment coil on the head surface and the target location to determine the proximity of the treatment coil for treatment (e.g., further based on confirmation from the contact sensor that confirms that the treatment coil is in contact with the human subject’s head). This accounts for errors in the fitting of the head model. For example, as shown in FIG. 4A, the color of the location indicator 407 may indicate whether the location of the treatment coil is aligned with the location of the target location. For example, as shown in FIG. 4A and 4B, the color of the location indicator 407 may be a particular color (e.g., green) to show that the location of the treatment coil is aligned with the location of the target location. Further, the color of the orientation indicator 408 may indicate whether the orientation of the treatment coil is aligned with the required orientation that the target location. In FIG. 4A, the color of the orientation indicator 408 may indicate that the treatment coil is at an incorrect orientation relative to the required orientation of the target location (e.g., the orientation indicator 408 may be red).

[0105] FIG. 4B illustrates a diagram 410 that includes a GUI that indicates that the position and orientation of the treatment coil are correct relative to the target location. For example, in FIG. 4B, the treatment system may generate a GUI where the color of the location indicator 403 and the orientation indicator 404 of the treatment coil indication 402 are of a specific color (e.g., green) to indicate that both the location and orientation of the treatment coil are correct relative to the target location. After the treatment coil is at the proper location and orientation, the technician may conduct a treatment procedure if the target location is the treatment location. Or, alternatively, if the target location is a location used to find the treatment location (e.g., is a motor threshold location), the technician may determine the human subject’s motor threshold level (e.g., a stimulation level for driving the treatment coil that elicits a neurological response in the human subject). Further, the treatment system may display, via the GUI on the display, an indication of a direction or a distance that the position of the treatment coil needs to be moved to reach the target location. The indication of the direction or the distance may comprise an indicator, such as an arrow or other shape, that isNNI 0507 DCVS PCT overlaid on top of a graphical representation of the head of the human subject. Further, in some examples, the diagram 410 may include contact indicators 430a, 430b that indicate that the treatment coil is in contact with the head of the human subject (e.g, based on feedback from the context sensor). Further, in some examples, the contact indicators 430a, 430b may indicate the location of the contact between the treatment coil and the head of the human subject.

[0106] After determining the human subject’s motor threshold level, the treatment system may generate a GUI that assists the technician in moving the treatment coil from the target location to the treatment location as shown in FIG. 4C. FIG. 4C depicts an example diagram 420 as displayed by a GUI that indicates the head model 409 of the human subject, the treatment coil indicator 402 that indicates the location of the treatment coil relative to the head of the human subject, and a treatment location indicator 414. The treatment location indicator 414 may include a location indicator 413 that indicates the location of a target location on the head model of the human subject, and an orientation indicator 414 that indicates the orientation (e.g., tilt, yaw, or roll) that a treatment coil needs to be placed relative to the treatment location. In FIG. 4C, the GUI may indicate that the treatment coil is in the incorrect location relative to the treatment location (e.g, the color of the location indicator 413 may be in a color, such as red, that indicates that the location is incorrect) and also indicate that the orientation of the treatment coil is correct relative to the required orientation at the treatment location (e.g, the color of the orientation indicator 414 may be in a color, such as green, that indicates that the orientation of the treatment coil is correct). Further, the GUI may include an indication 416 of a displacement error. The indication 416 of the displacement error may indicate the direction and / or orientation that the treatment coil needs to be moved relative to the head of the human subject to be moved from the target location (e.g., the MT location) to the treatment location. The indication 416 of the displacement error may assist the technician in moving the treatment location to the treatment location. Accordingly, in some examples, the treatment system may display, via the GUI on the display, the indication of the displacement error of the treatment coil relative to the target location during the procedure 300. In examples, the displacement error indicator 416 may indicate the direction and / or orientation needed to move the treatment coil from the MT location to the treatment location.

[0107] As noted above, the treatment system may include the contact sensor that, for example, may be located on a bottom surface of the treatment coil (e.g., the surface of the treatment coil that isNNI 0507 DCVS PCT intended to be in direct contact with the head of the human subject during treatment). The treatment system may determine whether the treatment coil is in contact with the head of the human subject based on the signals received from the contact sensor. For example, the treatment system may receive signals from a contact sensor located on the treatment coil, and determine whether the treatment coil is in contact with the head of the human subject based on the signals received from the contact sensor. Further, the treatment system may display, via the GUI on the display device, an indication of whether the treatment coil is in contact with the head of the human subject (e.g., in addition to the indicators 402, 406, and / or 412 shown in FIG. 4A-4C). As noted above, in some examples, the GUI may comprise one or more contact indicators 430 that indicate that the treatment coil is in contact with the head of the human subject.

[0108] In some examples, the treatment system may determine a distance between the treatment coil indicator 402 and the target location and / or treatment location indicator 406, 412 based on one or more medical images. The treatment system may indicate the direction that the treatment coil needs to move to arrive at the target location and / or treatment location indicator 406, 412 based on the medical images. The one or more structural and / or functional medical images may comprise MRI images, PET scan images, SPECT scan images, and / or ultrasound images.

[0109] In some examples, the treatment system may determine the location of the target location and / or treatment indicators based on generic image modeling information. For example, the treatment system may fit the images and / or treatment coil to a MNI model. An MNI model may include standard anatomical templates are used in human neuroimaging processing pipelines to facilitate group level analyses and comparisons across different subjects and populations. In some examples, the treatment system may fit a generic head model, such as the Montreal Neurological Institute (MNI) Template or the Talairach template, to the patient’s head. One or more generic head models with known coordinates may be used. The head model whose coordinates match the most closely with the dimensions of the patient’s head can be selected as a better alternative to using a single template. One example of an MNI model is the MNI-ICBM152 template, which represents an average of 152 healthy adult brains. The treatment system may determine the location of the target location and / or treatment indicators based on the shifting of locations in the model after performing the patient head fitting process. The treatment system may use detected landmarks, indicated land marks, a down sampled point cloud, and / or the full point cloud for the head fitting process. TheNNI 0507 DCVS PCT treatment system may translate and / or orient the MNI model to the patient’s face landmark points or use a fitting algorithm with the measured point cloud from the cameras (e.g., depth cameras). For instance, the treatment system may deform the MNI model to match the patient’s head dimensions. Accordingly, the treatment system may determine the location of the target location and / or treatment indicators based on the MNI model and the head model of the human subject.

[0110] FIG. 5 is a diagram illustrating an example interface 500 of a treatment system that is using image recognition to determine the presence and location of a face of a human subject and a treatment coil. The treatment system may be an example of the treatment system 100. The treatment system may receive a plurality of images from the one or more cameras of the treatment system, and use one or more image processing techniques to detect the treatment coil 540 of the treatment system and / or the face of the human subject 520. The treatment coil 540 may be an example of the treatment coil 102 of FIG. 1, and the human subject 520 may be an example of the human subject 120 of FIG. 1.

[0111] The treatment system may be configured to detect a plurality of points 511 that are associated with the face of the human subject 520 using one or more image processing techniques (e.g, facial recognition techniques, support vector machines (SVM), a Viola-Jones object detection framework, etc.). The treatment system may be configured to generate an object 510 (e.g., a bounding box) that includes the points 511 that are associated with the face of the human subject 520. In some examples, the treatment system may detect one or more points associated with the head of the human subject 520 using an indicator toll 550. For instance, the treatment system may use the indicator tool 550 to identify one or more specific locations on the human subject’s head that are difficult for the cameras to detect, such as the top of their head (e.g., beneath their hair) and / or the back of their head. The treatment system 100 may generate the 2D and / or 3D digital reconstruction of the human subject’s head based on the locations identified using the indicator tool and the points 511 captured by the cameras (e.g., and image processing techniques that identify one or more facial features of the human subject), for instance, as described herein.

[0112] The treatment system may be configured to identify the treatment coil 540. For example, as noted above, the treatment system may determine a surface 542 of the treatment coil 540 within an image of the plurality of images based on a direction of a plurality of vectors within the image and / orNNI 0507 DCVS PCT the shape of the surface. For example, the treatment system may detect the location of the treatment coil 540 using one or more image processing techniques (e.g., support vector machines (SVM), a Viola-Jones object detection framework, etc ). The treatment system may detect vectors that are at a point on a curve, and the treatment system may decompose the vector as a sum of two vectors: one tangent to the curve, called the tangential component of the vector, and another one perpendicular to the curve, called the normal component of the vector. The treatment system may identify the surface 542 of the treatment coil 540 (e.g., a flat surface of the treatment coil 540) based on the detection of a plurality of vectors that all point in the same direction. For instance, the treatment system may detect the surface 542 of the treatment coil 540 based on the identification of a surface that has multiple normal vector components in the same direction. Further, the treatment system may display, via the GUI on the display, an object 530 (e.g., a bounding box) that highlights the position of the treatment coil 540. As such, the treatment system may track movement of the treatment coil 540 e.g., and / or the head of the human subject 520) based on additional images received via the one or more cameras (e.g. , during a procedure and without the use of markers attached to the treatment coil).

[0113] FIG. 6 is a diagram of an example of a predetermined head model, a point cloud, and a fitted head model. For example, a treatment system (e.g., the treatment system 100) may generate a fitted head model 620 using a predetermined head model 601 and a plurality of points 610. The predetermined head model 601 may be a generic head model that does not include any characteristics that are specific to one individual. A predetermined head model 610 may be used, for example, to keep the anonymity of the human subject, to provide information relating to one or more predefined coordinates, to reduce the total number of artifacts in the fitted head model, to reduce or eliminate the asymmetric nature or abnormal features of the human subject (e.g., loss of ear, asymmetric bump, etc.), and / or the like.

[0114] The treatment system may determine the plurality of points 610 based on the images captured by a sensor, such as one or more cameras (e.g, the cameras 110a, 110b). The treatment system may use a facial recognition technique to determine the points 610. Further, as noted herein, the treatment system may include an indicator tool that may be used to identify one or more specific locations on the human subject’s head that are difficult for the cameras to detect, such as the top of their head (e.g., beneath their hair) and / or the back of their head. As such, the points 610 mayNNI 0507 DCVS PCT include points that were detected using the indicator tool. The plurality of points 610 may comprise facial feature information 61 la-c. The facial feature information 61 la-c may include the relative location of one or more of the human subject’s facial features. For example, the points 611a may comprise information relating to the relative location of the human subject’s right ear, points 611b may comprise information relating to the relative location of the human subject’s right eye, and points 611c may comprise information relating to the relative location of the human subject’s nose. Although facial feature information relating to the human subject right ear, right eye, and nose are identified in the example in FIG. 6, it should be understood that more or less facial feature information may be provided by a point cloud. A partial set of facial features may be used to correspond to rigid parts of the patient’s anatomy. Using a partial set of facial features for point cloud generation may prevent extraneous motions (e.g., the patient’s jaw movement) form causing tracking errors.

[0115] Although not illustrated in the example in FIG. 6, the predetermined head 601 model may comprise predefined coordinates, for example, such as an EEG 10-20 coordinate grid, one or more treatment locations, and / or the like. The predetermined head model 601 may include location information relating to the one or more anatomical landmarks. An anatomical landmark may comprise a nasion, an inion, a lateral canthus, an external auditory meatus (e.g., ear attachment point), and / or one or more preauricular points of the human subject. An anatomical landmark may be associated with an x coordinate, a y coordinate, and a z coordinate of the predetermined head model 601.

[0116] The treatment system may generate the fitted head model 620 using the predetermined head model 601 and the plurality of points 610. The fitted head model 620 may have a smooth surface. For example, the fitted head model 620 may have a smooth surface because it is created using a predetermined head model that also has a smooth surface. Further, in one or more embodiments, the treatment system may determine that one or more points of the point cloud 610 are associated with rippling of the human subject’s skin, loss of a facial feature, and / or an asymmetric lump, and generate the fitted head model 620 without the use of the these points. For example, the treatment system may recognize an unusual fluctuation between points, an asymmetry between points on opposite sides of the head, and / or a variance between the points and that of the predetermined head mode 601, and determine that one or more points of the point cloud 610 are associated with ripplingNNI 0507 DCVS PCT of the human subject’s skin, loss of a facial feature, and / or an asymmetric lump. The treatment system may not use these points when generating the fitted head model 620.

[0117] The fitted head model 620 may not include the human subject’s hair. The treatment system may identify, exclude, and / or remove the human subject’s hair (e.g, facial hair and / or head hair) from the fitted head model 620. The hair may act as interference when generating the fitted head model 620. The exclusion of the human subject’s hair may improve the accuracy when determining one or more target locations, for example, a MT location, a treatment location, and / or the like. The removal and / or exclusion of the human subject’s hair from the fitted head model 620 may be performed by the treatment system in one or more ways, which for example, may be performed in any combination.

[0118] Similarly, the treatment system may exclude from the fitted head model 620 any parts of the patient’s anatomy that create extraneous motions (e.g., the patient’s jaw movement). The extraneous motions may interfere and / or create errors when generating the fitted head model 620. The exclusion of the parts of the patient’s anatomy that create extraneous motions may improve the accuracy when determining one or more target locations, for example, a MT location, a treatment location, and / or the like.

[0119] The treatment system may determine that one or more points of the point cloud 610 that are used to create the fitted head model 620 are associated with the human subject’s hair. Thereafter, the treatment system may generate the fitted head model 620 without the use of the points associated with the human subject’s hair. For example, the treatment system may use the human subject’s facial features (e.g., those in close proximity to the hair in the front and / or back of the head) as a guide to determine the location of the hair, which for example, may be performed since the sensor’s scan of the human patient may include the face of the human subject along with the rest of the human subject’s head. The treatment system may determine one or more location of the surface of the head under the hair using a probe (e.g., finger, tool, etc.) and use these locations to determine the location of the hair (e.g., the points in the point cloud that rest outside of these locations). The treatment system may estimate the surface of the head under the hair using one or more facial features and / or anatomical landmarks. For example, removal of the hair may be performed by estimating intermediate points and using points relating to the human subject’s scalp to map the headNNI 0507 DCVS PCT that is under the hair. The treatment system may generate and use a mathematical head model using the known facial features and / or anatomical landmarks. A gel that is visible to a sensor (e.g., the sensor 110, such as the cameras 110a, 110b) may be spread on the hair of the patient, such that the treatment system may determine which of the plurality of points detected by the sensor correspond to the human subject’s hair and exclude those points when generating the fitted head model 620.

[0120] The plurality of points that are captured using the sensor and used to create the fitted head model 620 may be devoid of information relating to the human subject’s hair, such that the fitted head model 620 is devoid of information relating to the human subject’s hair when it is generated by the treatment system.

[0121] The fitted head model 620 may include one or more reference points, for example, such as predefined coordinate systems (e.g., an EEG 10-20 coordinate grid) that may be used for a therapeutic and / or diagnostic procedure, one or more target and / or treatment locations, and / or the like.

[0122] The fitted head model 620 may include information relating to one or more facial features. For example, the facial feature information may be provided via the cloud of points 610. The facial feature information may be registered with the fitted head model 620. In one or more embodiments, the technician may identify and / or confirm a facial feature using a gesture, for example, by placing a finger on the facial feature. The treatment system 100 may determine the gesture using the sensor, for example, to identify that the technician’s finger is on the facial feature. Once identified, the technician may confirm the facial feature via the user interface before the facial feature is registered with the fitted head model 620 by the treatment system.

[0123] FIG. 7 is a diagram of an example of a predetermined head model superimposed with a plurality of points determined by a sensor, such as the cameras 110a, 110b. The predetermined head model 700 may be an example of the predetermined head model 601. The plurality of points may be an example of the point cloud 610. The points 710 are a subset of the plurality of points that may be determined by a sensor (e.g, the sensor 110, sensor 200, and / or the like). It should be understood that not all points of the point cloud are labeled 710 for purposes of simplicity and clarity. As shown by the example of FIG. 7, some (e.g, all) of the points of the point cloud may not correspond directly with the predetermined head model 700. The treatment system may generate the fitted headNNI 0507 DCVS PCT model using the predetermined head model 700 and the point cloud 710, for example, by morphing the predetermined head model 700 to the point cloud 710. As such, the fitted head model may have a similar look and / or feel as the predetermined head model 700, but to the dimensions of the point cloud 710 determined by the sensor.

[0124] FIG. 8 is a diagram of an example fitted head model that includes an EEG 10-20 coordinate grid. The fitted head model 800 may be an example of the fitted head model 620. The fitted head model 800 may comprise EEG 10-20 coordinate grid locations 810. The EEG 10-20 coordinate grid locations 810 may be used to determine one or more target locations and / or treatment locations. A target location (e.g., the motor threshold (MT) location) may be a reference point used to determine a treatment location. A treatment location may be used as the position for the treatment coil 102 for the diagnostic and / or therapeutic treatment, and / or the like.

[0125] The treatment system (e.g., the treatment system 100) may determine one or more anatomical landmarks and associate the anatomical landmarks on the fitted head model. This may be referred to as anatomical landmark registration. The treatment system may determine one or more anatomical landmarks without the use of additional image information, such as MRI images, computer tomography (CT) images, an X-ray image, and / or the like. An anatomical landmark may comprise a nasion, an inion, a lateral canthus, an external auditory meatus (e.g., ear attachment point), and / or one or more preauricular points of the human subject. An anatomical landmark may be associated with an x coordinate, a y coordinate, and a z coordinate of the fitted head model. The treatment system may register the anatomical landmarks with the fitted head model.

[0126] The treatment system may determine the location of the anatomical landmarks using the facial feature information provided via the point cloud. For example, the treatment system may perform triangulation and / or trilateration using the facial feature information (e.g., one or more points associated with the facial feature information) to determine the location of one or more anatomical landmarks of the human subject on the fitted head model. In examples, the treatment system may determine the location of the anatomical landmarks using the facial feature information provided via the point cloud if the predetermined head model does not include information relating to the anatomical landmarks.NNI 0507 DCVS PCT

[0127] The fitted head model may comprise the one or more anatomical landmarks when it is generated by the treatment system. For example, the treatment system may integrate the anatomical landmarks from the predetermined head model into the fitted head model when created, for example, using the facial feature information as indicators of the specific location of the anatomical landmarks. The head model and / or its associated landmarks may be placed relative to the detected facial features reducing the computational complexity and time of using the full depth image data. Alternately, the head model may be fitted to a subset of the depth data segmented by using the region near the facial feature.

[0128] The treatment system may identify the anatomical landmarks and the technician may confirm the identification of the anatomical landmarks, for example, to associate the anatomical landmarks with the fitted head model. The technician may confirm the identification of the anatomical landmarks via the user interface of the treatment system.

[0129] Once associated with the fitted head model, the anatomical landmarks may be used to assist in the therapeutic and / or diagnostic procedure of the human subject. For example, the anatomical landmarks may be used to assist the technician in neuronavigation of the human subject’s head. Further, the anatomical landmarks may be used to identify and / or store locations that may be used for therapeutic and / or diagnostic procedure, for example, the MT location and / or treatment location of the human subject. In one or more embodiments, anatomical landmark registration may be performed prior to the procedure (e.g., each procedure) for the human subject, for example, to orient the fitted head model with the human subject’s actual head. It should be appreciated that, in some examples, the treatment system may generate the fitted head model using the cloud of points without the use of a predefined head model.

[0130] FIG. 9 is a flow chart of a procedure 900 for training a machine learning model to determine the location and / or the orientation of the treatment coil. The procedure 900 may be performed by a controller of a treatment system (e.g, the controller 150 of the treatment system 100 of FIG. 1 A). Alternatively or additionally, the procedure 900 may be performed by an external processing system (e.g, one or more external servers), and the machine learning mode could be transferred to the treatment system for later use (e.g., to perform the procedure 300 of FIG. 3). The treatment systemNNI 0507 DCVS PCT may perform the procedure 900 to train a machine learning model to detect a location and / or an orientation of a treatment coil relative to a head of a human subject.

[0131] At 902, the treatment system may receive a plurality of images from one or more cameras (e.g, the cameras 110a, 110b). The plurality of images may comprise images of the treatment coil (e.g., the treatment coil 102). In some examples, the images may be depth images, such as RGB-d images or greyscale depth images.

[0132] At 904, the treatment system may determine a surface of the treatment coil within the plurality of images. For instance, the treatment system may identify one or more surfaces that corresponds to the size and shape of the treatment coil. In some examples, the treatment coil (c.g, the points associated with the treatment coil) may be identified (e.g, manually) in the plurality of images that are received by the treatment system at 902. For instance, a user may identify the treatment coil in the plurality of images prior to providing the images in the treatment system for training. In some examples, the treatment system may use the plane fitting algorithm to identify a surface or side of the treatment coil in one or more images (c.g., depth images) to, for example, determine a subset of points that represent the treatment coil from the plurality of points that are associated with an image (e.g, segment the treatment coil from the cloud of points identified in the space). In some examples, the treatment system may detect a subset of the points of the point cloud (e.g, generated from the depth images) that make up the treatment coil based on the detection of a surface of the treatment coil. For example, the treatment system may use a neural network to detect the approximate location of the treatment coil in one or more images (e.g., depth images, or a point cloud based on those depth images) to detect the treatment coil. The treatment system may detect one or more surfaces of the treatment coil using a plane-fitting algorithm to, for example, extract one or more surfaces (e.g., flat surface) of the treatment coil from the images (e.g, the depth images and / or point cloud).

[0133] At 906, the neural network may identify a plurality of vectors associated with the surface of the treatment coil. For example, the treatment system may determine a surface of the treatment coil 102 within an image of the plurality of images based on a direction of a plurality of vectors within the image. For example, the treatment system may detect a plurality of vectors associated with the treatment coil using one or more image processing techniques (c.g, support vector machines (SVM),NNI 0507 DCVS PCT a Viola-Jones object detection framework, RANSAC, principal component analysis, iterative closest point methods, machine learning, pattern recognition, Adaptive thresholding (e.g., detect markers), digital filtering and averaging, perspective transformation, warping to correct for lens distortion, etc.). As described above, the treatment system may detect vectors that are at a point on a curve, and the treatment system may decompose the vector as a sum of two vectors: one tangent to the curve, called the tangential component of the vector, and another perpendicular to the curve, called the normal component of the vector. The treatment system may identify a surface of the treatment coil based on a unique alignment of vectors that are specific to one or more surfaces of the treatment coil. For instance, the treatment system may identify a surface of the treatment coil (e.g, a flat surface of the treatment coil) based on the detection of a plurality of vectors that are all one point in the same direction. For instance, the treatment system may detect a flat surface of the treatment coil based on the identification of a surface that has normal components in the same direction.

[0134] At 908, the treatment system may train a machine learning model to identify the surface of the treatment coil using training data. The training data may comprise the plurality of images, an identification of the surface of the treatment coil taken from the plurality of images, and / or the plurality of vectors associated with the surface of the treatment coil e.g., as described at 902-906). The images may be captured in different rooms that have different lighting. As such, the machine learning model may be trained using images (e. ., depth images) and / or point clouds that represent a unique surface of the treatment coil in a variety of different settings. Thereafter, the treatment system may use the trained machine learning model to identify the treatment coil during a diagnostic or treatment procedure, and in some examples, to conduct neuronavigation (e.g., as described herein, for example, with reference to the procedure 300).

[0135] The treatment system may receive a second plurality of images associated with a second treatment coil and determine a surface of the second treatment coil using the trained machine learning model based on the second plurality of images. The second treatment coil may have substantially similar dimensions, shape, and / or size to the first treatment coil (e.g., the second treatment coil may be a second instance or duplicate of the same treatment coil). For example, the first treatment coil may be used to train the machine learning model, while the second treatment coil may be used during a therapeutic and / or diagnostic procedure of the human subject. As noted herein, the treatment system may detect the location and / or orientation of the second treatment coilNNI 0507 DCVS PCT using the machine learning mode, and display, via a GUI on a display, a representation of the second treatment coil. The treatment system may track movement of the second treatment coil based on additional images received via the one or more cameras. The treatment system may identify a housing of the coil based on a orientation of a subset of the identified plurality of vectors.

[0136] Although described with reference to the head of a human subject, one or more of the methods and / or procedures described herein may applied to anatomies other than the head. For example, the treatment system may be used to create a fitted model of an arm, leg, hand, foot, chest, and / or other anatomy of a human subject. Further, the treatment system may be used to create a fitted model of an anatomy of a subject other than humans, such as birds, amphibians, reptiles, fish, insects, and / or other mammals.

[0137] FIG. 10 is a flowchart of a procedure 1000 for training and / or using a trained machine learning model to determine the location and / or the orientation of a treatment coil. A controller of a treatment system (e.g., the controller 150 of the treatment system 100 of FIG. 1A) may perform the procedure 1000. Additionally or alternatively, the procedure 1000 may be performed by an external processing system (e.g., one or more external servers). A trained machine learning model may be transferred to the treatment system for later use (e.g., to perform the procedure 300 as shown in FIG. 3). The treatment system may perform the procedure 1000 to train a machine learning model to detect a location and / or an orientation of a treatment coil relative to a head of a human subject.

[0138] At 1002, the treatment system may receive a plurality of images (e.g., such as depth images, or points clouds) from one or more cameras, such as depth cameras (e.g., the first and second cameras 110a, 110b). The plurality of images may comprise images of the patient and / or treatment coil (e.g., the treatment coil 102). At 1004, the treatment may use a trained machine learning model and / or neural network to identify the treatment coil in the captured images. In some examples, the treatment system may use one or more markers to identify the treatment coil. In some examples, the treatment coil (e.g., the points associated with the treatment coil) may be identified (e.g., manually) in the plurality of images that are received by the treatment system at 1002. For instance, a user may identify the treatment coil in the plurality of images prior to providing the images in the treatment system for training. In some examples, the treatment system may determine a subset of points that represent the treatment coil from the plurality of points that are associated with an image (e.g. ,NNI 0507 DCVS PCT segment the treatment coil from the cloud of points identified in the space). In some examples, the treatment system may detect a subset of the points of the point cloud (e.g., generated from the depth images) that make up the treatment coil based on the detection of a surface of the treatment coil.

[0139] At 1006, the treatment system may identify regions in point clouds where a surface of the treatment coil is flat (e.g., substantially flat). For instance, the treatment system may detect a plurality of vectors associated with the treatment coil using one or more image processing techniques (e.g., support vector machines (SVM), RANSAC, LAD, a Viola-Jones object detection framework, or improved versions of these, < / < .). The treatment system may detect vectors that are at a point on a curve. The treatment system may decompose the vector as a sum of two vectors: one tangent to the curve, called the tangential component of the vector, and another one perpendicular to the curve, called the normal component of the vector. The treatment system may identify a collection of surface normal vectors all pointing the same direction to identify a flat surface of the treatment coil. This collection of surface normal vectors all pointing in the same direction makes it known to the treatment system that the surface from which the surface normal vectors originate is flat (e.g., substantially flat).

[0140] At 1008, the treatment system may then determine the orientation e.g., roll, tilt, and / or yaw) of the treatment coil from the direction of the collected surface normal vectors. The treatment system may determine the orientation of the surface using a principal component analysis or similar method. In some examples, the treatment system may determine the pitch of the coil from the orientation of the shape of the surface of the treatment coil. At 1010, the treatment system may determine the location of the treatment coil based on normal vectors in the one or more images (e.g., based on the distance to and position of the normal vectors in the depth image). In some examples, the treatment system may determine the location of the treatment coil based on the orientation of the treatment coil and / or by fitting the shape of the treatment coil to the detected surface(s). Further, in some examples, the treatment system train a machine learning model based on the detected location of the treatment coil. Further, in some examples, the treatment system may use the procedure 1000 to detect a location and / or an orientation of a treatment coil relative to a head of a human subject. For instance, the treatment system may detect a location and / or an orientation of a treatment coil relative to a head of a human subject based on the detected location and / or orientation of the treatment coil and a detected located on the head of the human subject. As previously noted, theNNI 0507 DCVS PCT head of the human subject may be detected based on a point cloud generated from one or more depth images. Further, the point cloud of the human subject may be defined by an STL file that, for instance, may be a 3D model file format that represents the surface geometry of an object (e.g., the human subject) using a collection of connected triangles, called a mesh.

[0141] Although features and elements are described above in particular combinations, one of ordinary skill in the art will appreciate that each feature or element can be used alone or in any combination with the other features and elements. In addition, the methods described herein may be implemented in a computer program, software, or firmware incorporated in a computer-readable medium for execution by a computer or processor (e.g., that resides at the treatment system, partially or fully, and / or at one or more external servers). Examples of computer-readable medium include electronic signals (transmitted over wired or wireless connections) and computer-readable storage medium. Examples of computer-readable storage mediums (e.g., non-transitory computer-readable storage mediums) include, but are not limited to, a read only memory (ROM), a random access memory (RAM), a register, cache memory, semiconductor memory devices, magnetic media such as internal hard disks and removable disks, magneto-optical media, and optical media such as CD- ROM disks, and digital versatile disks (DVDs).

Claims

NNI 0507 DCVS PCTCLAIMSWhat is claimed is:

1. A method for identifying a location and an orientation of a treatment coil used for transcranial magnetic stimulation (TMS), the method comprising: receiving a plurality of images from one or more cameras, wherein the plurality of images comprises images of the treatment coil and of a head of a human subject; determining a surface of the treatment coil within an image of the plurality of images based on a direction of a plurality of vectors within the image; determining the location of the treatment coil based on the plurality of vectors; determining the orientation of the treatment coil based on an orientation of the plurality of vectors; and displaying, via a graphical user interface (GUI) on a display, the location and the orientation of the treatment coil relative to a target location on the head of the human subject.

2. The method of claim 1, wherein the orientation of the treatment coil comprises one or more of a roll, a tilt, or a yaw of the treatment coil, and wherein the orientation of the plurality of vectors comprises one or more of a roll, a tilt, or a yaw of the plurality of vectors.

3. The method of claim 1, wherein the orientation of the treatment coil is determined based on an orientation of the plurality of vectors using a principal component analysis.

4. The method of claim 1, wherein the surface of the treatment coil is determined within an image of the plurality of images based on a normal component of each of the plurality of vectors associated with the image being aligned in a particular direction.

5. The method of claim 4, further comprising: determining the normal component of each of the plurality of vectors by identifying one or more segment candidate, wherein each of the one or more segment candidates is associated with the one or more vectors of the plurality of vectors; andNNI 0507 DCVS PCT applying a random sample consensus (RANSAC) to each of the one or more segment candidates to determine the normal component of each of the plurality of vectors.

6. The method of claim 1, wherein the surface is a flat surface.

7. The method of claim 1, wherein the surface of the treatment coil is determined within the image based on a respective direction of each of a first subset of the plurality of vectors.

8. The method of claim 1, further comprising: displaying, via the GUI on the display, an indication of a displacement error of the treatment coil relative to the target location.

9. The method of claim 8, further comprising: determining that the treatment coil has been moved to the target location; and changing a color of the indication of the displacement error to indicate that the position of the treatment coil has been moved to the target location.

10. The method of claim 1, further comprising: receiving signals from a contact sensor located on the treatment coil; determining whether the treatment coil is in contact with the head of the human subject based on the signals received from the contact sensor; and displaying, via the GUI on the display, an indication of whether the treatment coil is in contact with the head of the human subject.

11. The method of claim 1, further comprising: displaying, via the GUI on the display, an indication of a direction or a distance that the treatment coil needs to be moved to reach the target location.

12. The method of claim 11, wherein the indication of the direction or the distance comprises an arrow that is overlaid on top of a graphical representation of the head of the human subject.NNI 0507 DCVS PCT13. The method of claim 11, further comprising: displaying, via the GUI on the display, an indication that the treatment coil has been moved to the target location.

14. The method of claim 13, further comprising: determining that the treatment coil has been moved to the target location; and changing a color of a location indicator to indicate that the treatment coil has been moved to the target location.

15. The method of claim 1, further comprising: displaying, via the GUI on the display, an object that highlights the location of the treatment coil; and tracking movement of the treatment coil based on additional images received via the one or more cameras.

16. The method of claim 1 , wherein the one or more cameras comprise a first red-green-blue depth (RGB-d) camera and a second RGB-d camera.

17. The method of claim 16, further comprising: determining the surface of the treatment coil based on images captured by the first RGB-d camera; and determining the head of the human subject based on images captured by the second RGB-d camera.

18. The method of claim 1, wherein the surface of the treatment coil is determined using a trained machine learning model.NNI 0507 DCVS PCT19. The method of claim 18, wherein the trained machine learning model has been trained to identify the location of the treatment coil and the head of the human subject in one or more two- dimensional images.

20. The method of claim 1, further comprising: determining a feature of the human subject based on a facial recognition algorithm; and determining a location and orientation of the human subject based on the feature of the human subject.

21. The method of claim 1, wherein location of the treatment coil is determined based on a surface registration technique.

22. The method of claim 1, wherein the location of the treatment coil is determined based on one or more previously determined locations of the treatment coil.

23. The method of claim 1, wherein the location and orientation of the treatment coil within the images is determined using a neural network.

24. The method of claim 1, wherein a marker or indicator tool is used to determine the location of the treatment coil in the image.

25. The method of claim 1, further comprising: detecting the surface of a circular surface of the treatment coil using a marker or indicator tool.

26. The method of claim 1, wherein the treatment coil comprises a sensor that is configured to measure the orientation of the treatment coil.NNI 0507 DCVS PCT27. The method of claim 26, wherein the sensor is configured to measure the orientation of the treatment coil with respect to gravity, earth’s magnetic field, or a human subject positioning apparatus.

28. The method of claim 1, wherein the position or orientation of the treatment coil is determined based on an inertial measurement unit of the treatment coil.

29. The method of claim 1, wherein the target location may be a treatment location, a motor threshold location, or other reference point.

30. The method of claim 1, wherein the location of the treatment coil is determined based on a relationship between one or more geometric features of the treatment coil and an induced electric field of the treatment coil.

31. The method of claim 1, wherein each image of the plurality of images is a point cloud.

32. A method for training a machine learning model to detect a location and an orientation of a treatment coil relative to a head of a human subject, the method comprising: receiving a plurality of images from one or more cameras, wherein the plurality of images comprises images of the treatment coil; determining a surface of the treatment coil within the plurality of images; identifying a plurality of vectors associated with the surface of the treatment coil; and training a machine learning model to identify the surface of the treatment coil using training data, wherein the training data comprises the plurality of images, an identification of the surface of the treatment coil within the plurality of images, and the plurality of vectors associated with the surface of the treatment coil.

33. The method of claim 32, further comprising: receiving a second plurality of images associated with a second treatment coil; andNNI 0507 DCVS PCT determining a surface of the second treatment coil using the trained machine learning model based on the second plurality of images.

34. The method of claim 33, further comprising: displaying, via a GUI on a display, a representation of the second treatment coil; and tracking movement of the second treatment coil based on additional images received via the one or more cameras.

35. The method of claim 32, further comprising: displaying, via a GUI on a display, a representation of the first treatment coil; and tracking movement of the first treatment coil based on additional images received via the one or more cameras.

36. The method of claim 32, further comprising identifying a housing of the treatment coil based on an orientation of a subset of the identified plurality of vectors.

37. A method for identifying a location of a treatment coil on a patient’s head used for transcranial magnetic stimulation (TMS), the method comprising: receiving a plurality of depth images from one or more depth cameras, wherein the plurality of depth images comprises depth images of the treatment coil and of a head of a live subject, and wherein each depth image of the plurality of depth images comprises a plurality of points, and wherein each depth camera are configured to determine the distance between two objects; determining the location of the treatment coil in each image of the plurality of images using a trained machine learning model; determining the orientation and location of the treatment coil within each image of the plurality of images based on a known geometry of the treatment coil correlating to a subset of the plurality of points within each image of the plurality of images; and displaying, via a graphical user interface (GUI) on a display, the orientation and the location of the treatment coil relative to a target location on the head of the live subject.NNI 0507 DCVS PCT38. The method of claim 37, further comprising: applying a random sample consensus (RANSAC) to the plurality of points associated with each image to identify the subset of the plurality of points within each image of the plurality of images that correlate to the known geometry of the treatment coil.

39. A method for identifying a location and an orientation of a treatment coil used for transcranial magnetic stimulation (TMS), the method comprising: receiving a plurality of depth images from one or more depth cameras, wherein the plurality of depth images comprises depth images of the treatment coil and of a head of a human subject; determining the location of the treatment coil and the orientation of the treatment coil based on the one or more depth images; and displaying, via a graphical user interface (GUI) on a display, the location and the orientation of the treatment coil relative to a target location on the head of the human subject, wherein the GUI further comprises an indication of a direction or a distance that the treatment coil needs to be moved to reach the target location.

40. The method of claim 39, wherein the indication of the direction or the distance comprises an arrow that is overlaid on top of a graphical representation of the head of the human subject.

41. The method of claim 39, wherein the GUI further comprises an indication that the treatment coil has been moved to the target location.

42. The method of claim 41, further comprising: determining that the treatment coil has been moved to the target location; and changing a color of a location indicator to indicate that the treatment coil has been moved to the target location.

43. A transcranial magnetic stimulation (TMS) system comprising: a treatment coil;NNI 0507 DCVS PCT one or more cameras, wherein each of the one or more cameras are configured to generate a plurality of images of the treatment coil and of a head of a human subject; and one or more processors configured to: receive the plurality of images from the one or more cameras; determine a surface of the treatment coil within an image of the plurality of images based on a direction of a plurality of vectors within the image; determine the location of the treatment coil based on the plurality of vectors; determine the orientation of the treatment coil based on an orientation of the plurality of vectors; and display, via a graphical user interface (GUI) on a display device, the location and the orientation of the treatment coil relative to a target location on the head of the human subject.

44. The TMS system of claim 43, wherein the orientation of the treatment coil comprises one or more of a roll, a tilt, or a yaw of the treatment coil, and wherein the orientation of the plurality of vectors comprises one or more of a roll, a tilt, or a yaw of the plurality of vectors.

45. The TMS system of claim 43, wherein the orientation of the treatment coil is determined based on an orientation of the plurality of vectors using a principal component analysis.

46. The TMS system of claim 43, wherein the surface of the treatment coil is determined within an image of the plurality of images based on a normal component of each of the plurality of vectors associated with the image being aligned in a particular direction.

47. The TMS system of claim 46, wherein the one or more processors are configured to: determine the normal component of each of the plurality of vectors by identifying one or more segment candidate, wherein each of the one or more segment candidates is associated with the one or more vectors of the plurality of vectors; and apply a random sample consensus (RANSAC) to each of the one or more segment candidates to determine the normal component of each of the plurality of vectors.NNI 0507 DCVS PCT48. The TMS system of claim 43, wherein the surface is a flat surface.

49. The TMS system of claim 43, wherein the surface of the treatment coil is determined within the image based on a respective direction of each of a first subset of the plurality of vectors.

50. The TMS system of claim 43, wherein the one or more processors are configured to: display, via the GUI on the display, an indication of a displacement error of the treatment coil relative to the target location.

51. The TMS system of claim 50, wherein the one or more processors are configured to: determine that the treatment coil has been moved to the target location; and change a color of the indication of the displacement error to indicate that the position of the treatment coil has been moved to the target location.

52. The TMS system of claim 43, wherein the one or more processors are configured to: receive signals from a contact sensor located on the treatment coil; determine whether the treatment coil is in contact with the head of the human subject based on the signals received from the contact sensor; and display, via the GUI on the display, an indication of whether the treatment coil is in contact with the head of the human subject.

53. The TMS system of claim 43, wherein the one or more processors are configured to: display, via the GUI on the display, an indication of a direction or a distance that the treatment coil needs to be moved to reach the target location.

54. The TMS system of claim 53, wherein the indication of the direction or the distance comprises an arrow that is overlaid on top of a graphical representation of the head of the human subject.

55. The TMS system of claim 53, wherein the one or more processors are configured to:NNI 0507 DCVS PCT display, via the GUI on the display, an indication that the treatment coil has been moved to the target location.

56. The TMS system of claim 55, wherein the one or more processors are configured to: determine that the treatment coil has been moved to the target location; and change a color of a location indicator to indicate that the treatment coil has been moved to the target location.

57. The TMS system of claim 43, wherein the one or more processors are configured to: display, via the GUI on the display, an object that highlights the location of the treatment coil; and track movement of the treatment coil based on additional images received via the one or more cameras.

58. The TMS system of claim 43, wherein the one or more cameras comprise a first red-green- blue depth (RGB-d) camera and a second RGB-d camera.

59. The TMS system of claim 58, wherein the one or more processors are configured to: determine the surface of the treatment coil based on images captured by the first RGB-d camera; and determine the head of the human subject based on images captured by the second RGB-d camera.

60. The TMS system of claim 43, wherein the surface of the treatment coil is determined using a trained machine learning model.

61. The TMS system of claim 60, wherein the trained machine learning model has been trained to identify the location of the treatment coil and the head of the human subject in one or more two- dimensional images.NNI 0507 DCVS PCT62. The TMS system of claim 43, wherein the one or more processors are configured to: determine a feature of the human subject based on a facial recognition algorithm; and determine a location and orientation of the human subject based on the feature of the human subject.

63. The TMS system of claim 43, wherein location of the treatment coil is determined based on a surface registration technique.

64. The TMS system of claim 43, wherein the location of the treatment coil is determined based on one or more previously determined locations of the treatment coil.

65. The TMS system of claim 43, wherein the location and orientation of the treatment coil within the images is determined using a neural network.

66. The TMS system of claim 43, wherein a marker or indicator tool is used to determine the location of the treatment coil in the image.

67. The TMS system of claim 43, wherein the one or more processors are configured to: detect the surface of a circular surface of the treatment coil using a marker or indicator tool.

68. The TMS system of claim 43, wherein the treatment coil comprises a sensor that is configured to measure the orientation of the treatment coil.

69. The TMS system of claim 26, wherein the sensor is configured to measure the orientation of the treatment coil with respect to gravity, earth’s magnetic field, or a human subject positioning apparatus.

70. The TMS system of claim 43, wherein the position or orientation of the treatment coil is determined based on an inertial measurement unit of the treatment coil.NNI 0507 DCVS PCT71. The TMS system of claim 43, wherein the target location may be a treatment location, a motor threshold location, or other reference point.

72. The TMS system of claim 43, wherein the location of the treatment coil is determined based on a relationship between one or more geometric features of the treatment coil and an induced electric field of the treatment coil.

73. The TMS system of claim 43, wherein each image of the plurality of images is a point cloud.

Images