Device parameter determination using prediction uncertainty

Abstract

Presented herein are methods and systems for configuring one or more operational parameters of a recipient device. A plurality of past operational parameters associated with configuring a recipient device (e.g., implantable medical device) for a recipient in one or more previous configuration sessions is obtained. A plurality of operational parameters for the recipient device is generated, via a trained probabilistic regression network, based on the plurality of past operational parameters. The recipient device is configured with the plurality of operational parameters.

Description

Atty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1DEVICE PARAMETER DETERMINATION USING PREDICTION UNCERTAINTY BACKGROUNDField of the Invention[oooi] The present invention relates generally to configuring one or more operational parameters of a recipient device based on uncertainty associated with predicted operational parameters.Related Art

[0002] Medical devices are devices that are intended to be used for medical purposes. They can vary in both their intended use and indications for use. Examples range from simple, low-risk medical supplies to complex, potentially high-risk devices that are implanted and / or sustain life, such as deep brain stimulators and brain-computer interfaces. Other categories of medical device include diagnostic equipment.

[0003] Hearing devices act on an actual or potential auditory perception of an individual, including to improve perception of sound signals, to reduce perception of sound signals, etc. In particular, a hearing device can deliver sound signals to a user in any form, including in the form of acoustical stimulation, mechanical stimulation, electrical stimulation, etc., and / or can operate to suppress all or some sound signals. As such, a hearing device can be a device for use by a hearing-impaired person (e.g., hearing aids, middle ear auditory prostheses, bone conduction devices, direct acoustic stimulators, electro-acoustic hearing prostheses, auditory brainstem stimulators, bimodal hearing prostheses, bilateral hearing prostheses, dedicated tinnitus therapy devices, tinnitus therapy devices, etc.) or a device for use by a person with normal hearing (e.g., a consumer device that provides audio streaming, a consumer headphone, an earphone, etc.), a hearing protection device (e.g., a noise cancellation headset, a loudness reduction apparatus, etc.), etc.SUMMARY

[0004] In one aspect, a method is provided. The method comprises: obtaining a plurality of past operational parameters associated with configuring a recipient device for a recipient in one or more previous configuration sessions; generating, via a trained probabilistic regression network, a plurality of operational parameters for the recipient device based on the plurality of past operational parameters; and configuring the recipient device with the plurality of operational parameters.Atty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1

[0005] In one aspect, a second method is proved. The method comprises: obtaining a plurality of historical stimulation parameters associated with configuring one or more medical devices; generating, via a probabilistic regression network, a plurality of stimulation parameters for a recipient medical device based on the plurality of historical stimulation parameters; and configuring the recipient medical device with the plurality of stimulation parameters.

[0006] In another aspect, one or more non-transitory computer readable storage media are provided. The one or more non-transitory computer readable storage media comprising instructions that, when executed by a processor, cause the processor to predict, via a first machine learning model, a plurality of predicted operational parameters for a recipient device; generate, via a second machine learning model, a plurality of operational parameters based on the plurality of predicted operational parameters; and configure the recipient device based on the plurality of operational parameters.

[0007] In another aspect, a system is provided. The system comprises: a memory; and at least one processor operable coupled to the memory, wherein the at least one processor is configured to: generate, via a machine learning model, a plurality of operational parameters for a device; determine a plurality of uncertainty values associated with the plurality of operational parameters; update a graphical representation of the plurality of operational parameters based on the plurality of uncertainty values; select one or more electrodes from a plurality of electrodes associated with the device based on the graphical representation; and configure the one or more electrodes of the device based on the plurality of operational parameters.BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Embodiments of the present invention are described herein in conjunction with the accompanying drawings, in which:

[0009] FIG. 1A is a schematic diagram illustrating a cochlear implant system with which aspects of the techniques presented herein can be implemented;[ooio] FIG. IB is a side view of a recipient wearing a sound processing unit of the cochlear implant system of FIG. 1A;[ooii] FIG. 1C is a schematic view of components of the cochlear implant system of FIG. 1 A;

[0012] FIG. ID is a block diagram of the cochlear implant system of FIG. 1 A;Atty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1

[0013] FIG. 2 is a functional block diagram illustrating an example system for generating a plurality of predicted parameters for configuring a recipient device, in accordance with certain embodiments presented herein;

[0014] FIG. 3 is a functional block diagram illustrating another example system for generating a plurality of predicted parameters for configuring a recipient device, in accordance with certain embodiments presented herein;

[0015] FIG. 4 is a functional block diagram illustrating another example system for generating a plurality of predicted parameters for configuring a recipient device, in accordance with certain embodiments presented herein;

[0016] FIG. 5 is a functional block diagram illustrating another example system for generating a plurality of predicted parameters for configuring a recipient device, in accordance with certain embodiments presented herein;

[0017] FIG. 6 is a flowchart illustrating a method for obtaining one or more historical operational parameters, in accordance with certain embodiments presented herein;

[0018] FIG. 7A is a graphical depiction illustrating a graph for visualizing uncertainty associated with a plurality of predicted parameters, in accordance with certain embodiments presented herein;

[0019] FIG. 7B is another graphical depiction illustrating a graph for visualizing uncertainty associated with a plurality of predicted parameters, in accordance with certain embodiments presented herein;

[0020] FIG. 7C is another graphical depiction illustrating a graph for visualizing uncertainty associated with a plurality of predicted parameters, in accordance with certain embodiments presented herein;

[0021] FIGs. 7D-7K are graphical depictions each including a graph for visualizing uncertainty values associated with a plurality of predicted parameters, in accordance with certain embodiments presented herein;

[0022] FIG. 8 is a flowchart illustrating a method for configuring a recipient device with a plurality of operational parameters, in accordance with certain embodiments presented herein;

[0023] FIG. 9 is a flowchart illustrating a method for configuring a recipient medical device with a plurality of stimulation parameters, in accordance with certain embodiments presented herein;Atty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1

[0024] FIG. 10 is a flowchart illustrating operations for configuring a recipient device, wherein the operations are performed by a processor executing instructions stored in one or more non-transitory computer readable storage media, in accordance with certain embodiments presented herein;

[0025] FIG. 11 is a flowchart illustrating operations performed by a system comprising a memory and at least one processor operable coupled to the memory, wherein the at least one processor is configured to perform the operations, in accordance with certain embodiments presented herein;

[0026] FIG. 12 is a schematic diagram illustrating a vestibular stimulator system with which aspects of the techniques presented herein can be implemented;

[0027] FIG. 13 is a schematic diagram illustrating a retinal prosthesis system with which aspects of the techniques presented herein can be implemented; and

[0028] FIG. 14 illustrates a fitting system with which aspects of the techniques presented herein can be implemented.DETAILED DESCRIPTION

[0029] Presented herein are techniques for configuring one or more operational settings or parameters, such as stimulation settings / parameters (e.g., threshold levels (“T-levels”), comfort levels (“C-levels”), a dynamic range, etc.), of a recipient device (e.g., implantable medical device) using predicted operational settings or parameters generated by a machine learning model.

[0030] More specifically, certain implantable medical devices, such as cochlear implants, electro-acoustic hearing prosthesis, auditory brainstem implants, etc., operate by converting at least a portion of received sound signals into electrical stimulation signals (current signals) for delivery to a recipient’s auditory system. The window / range of electrical amplitudes (current levels) at which electrical stimulation signals are delivered to the recipient’s auditory system is limited. In particular, if the amplitude of the electrical stimulation signals is too low, then the associated sounds used to generate the electrical stimulation signals will not be perceived by the recipient (i.e., the stimulation signals will either not evoke a neural response in the cochlea or evoke a neural response that cannot be perceived by the recipient). Conversely, if the amplitude of the electrical stimulation signals is too high, then the associated sounds used to generate the electrical stimulation signals will be perceived as too loud or uncomfortable by the recipient.Atty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1

[0031] As such, electrical stimulation signals are generally delivered between a lower limit, referred to herein as a “threshold level,” at which the associated sound signals are barely audible to the recipient, and an upper limit, referred to herein as a “comfort level,” above which the associated sound signals are uncomfortably loud to the recipient. The difference in electrical amplitudes between the threshold level and the comfort level is referred to herein as the “dynamic range.” In general, the term “stimulation parameters” herein can include the threshold level, the comfort level, the dynamic range, or other attributes (e.g., rate, frequency, modulation, etc.) of the electrical stimulation signals to be delivered to a recipient, regardless of whether or not the electrical stimulation signals are generated based on sound signals.

[0032] In the specific example of cochlear implants, due to a recipient’s specific anatomical features, the insertion depth of a given electrode, or other variables, the dynamic range can be different for different electrodes implanted in a recipient. That is, different electrodes implanted in a recipient can have different associated threshold and comfort levels. The range in acoustic amplitudes of sound signals received by a cochlear implant (or other auditory prosthesis) is considerably larger than the dynamic range associated with an electrode. As such, the conversion of the received sound signals into electrical stimulation signals for delivery to the recipient includes, among other operations, mapping (compression) of the acoustic amplitudes into electrical amplitudes within the dynamic range of the corresponding electrode(s) (i.e., the stimulating contact(s) at which the electrical stimulation is delivered to the recipient).

[0033] Medical practitioners (e.g., clinicians) often participate in clinical sessions to configure a recipient device (e.g., medical device) that is worn by, implanted in, etc., a recipient. These clinical sessions could be performed for a variety of reasons, such as to initially “fit” or configure (e.g., program) the recipient device for the recipient, to adjust or refine param eters / settings of the recipient device, to troubleshoot problems with the recipient device, etc. In conventional clinical sessions, medical practitioners configure or adjust operational parameters of a recipient device based on feedback from the recipient. In the specific example of cochlear implants, medical practitioners (e.g., audiologists) often need to perform new measurements to determine operational parameters (e.g., C-levels or T-levels) during configuration, or “fitting,” sessions. In many cases, due to a lack of information on the operational parameters, a medical practitioner must perform measurements on all electrodes of an electrode array even if some of the electrodes do not require configuration. This process can be time-consuming and inconvenient for the recipient. Thus, it is desirable to leverage dataAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1modeling techniques to predict / estimate operational parameters that have not yet been measured.

[0034] For example, a computer-implemented deterministic or stochastic decision-making artificial intelligent agent (e.g., implemented via a computer program / software application) can leverage a machine learning model to predict / estimate operational parameters. The machine learning model can be trained to predict operational parameters for configuring a recipient device (e.g., hearing aid or cochlear implant). For example, a priori information such as historical operational parameters and / or historical recipient-specific information associated with past device configuration sessions for a recipient population (e.g., cochlear implant recipients) can be obtained from a parameter database. Based on the obtained historical operational parameters and / or historical recipient-specific information, the machine learning model can generate a plurality of predicted operational parameters. The predicted operational parameters can be fine-tuned with observations and / or measurements made by the medical practitioner performing the configuration.

[0035] In one example, based on the predicted C-levels and / or T-levels, an audiologist can perform additional audiogram, psychoacoustical, and / or neurophysiological measurements and / or objective measurements (e.g., phoneme discrimination tests, speech tests, electrically evoked compound action potential (eCAP), etc.) on the recipient. The results of these additional measurements / tests can be used to adjust (e.g., fine-tune) the predicted operational parameters. The process of performing additional measurements and / or tests and fine-tuning the predicted operational parameters can proceed iteratively until a convergence condition is reached. Although fine-tuning the predicted parameters can improve predictions generated by the machine learning model, parameter refinement is often cumbersome and lengthy, particularly for inexperienced medical practitioners, because it requires performing additional tests on the recipient. Thus, it is desirable to implement a parameter refinement process for machine learning-based parameter prediction without reliance on psychoacoustic and / or neurophysiologic tests.

[0036] In order to address the above and other challenges, the techniques presented herein leverage a machine learning model trained on a priori information from past device configuration sessions associated with a recipient population and fine-tuned via expert evaluation to generate predicted operational parameters. Moreover, the techniques presented herein include determining and providing prediction uncertainty associated with operational parameters predicted by the machine learning model to assist medical practitioners in decidingAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1which prediction can be reliable, and which parameter requires further configuration. Experts (e.g., medical practitioners and / or recipients) can provide interactive and iterative evaluation of predicted operational parameters generated by the machine learning model, thus enabling an efficient and effective device configuration process. Further, the techniques presented herein include iteratively refining predicted parameters to minimize the difference between predicted and observed parameters, thus providing continuous improvement of predictions generated by the machine learning model.

[0037] There are a number of different types of device in / with which the techniques presented herein can be implemented. Merely for ease of description, the techniques presented herein are primarily described with reference to a specific device. However, it is to be appreciated that the techniques presented herein can also be partially or fully implemented by any of a number of different types of devices or systems, including consumer electronic devices (e.g., consumer hearing devices, consumer computing devices such as mobile phones and tablets, audio equipment such as home theatre and car audio systems, etc.), computing systems (e.g., servers in data centers, Intemet-of-Things (loT) devices), various types of software systems, such as databases, machine learning and artificial intelligence systems, other medical devices, such as diagnostic equipment or life sustaining equipment, etc. For example, the techniques presented herein could be used in or with sensory protheses, including hearing aids and cochlear implants, and various medical devices, such as pacemakers, drug delivery systems, implantable defibrillators, functional electrical stimulation devices, sleep disorder devices (e.g., sleep apnea devices), seizure devices (e.g., devices for monitoring and / or treating epileptic events), balance or movement disorder devices (e.g., vestibular stimulation devices), tinnitus management devices, visual implants (e.g., bionic eyes), and other neuromodulation devices (e.g., braincomputer interfaces).

[0038] FIGs. 1 A-1D illustrates an example cochlear implant system 102 with which aspects of the techniques presented herein can be implemented. The cochlear implant system 102 comprises an external component 104 and an implantable component 112. In the examples of FIGs. 1A-1D, the implantable component is sometimes referred to as a “cochlear implant.” FIG. 1A illustrates the cochlear implant 112 implanted in the head 154 of a user, while FIG. IB is a schematic drawing of the external component 104 worn on the head 154 of the user. FIG. 1C is another schematic view of the cochlear implant system 102, while FIG. ID illustrates further details of the cochlear implant system 102. For ease of description, FIGs.1 A-1D will generally be described together.Atty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1

[0039] Cochlear implant system 102 includes an external component 104 that is configured to be directly or indirectly attached to the body of the user and an cochlear implant 112 configured to be implanted in the user. In the examples of FIGs. 1A-1D, the external component 104 comprises a sound processing unit 106, while the cochlear implant 112 includes an implantable coil 114, an implant body 134, and an elongate stimulating assembly 116 configured to be implanted in the user’s cochlea.

[0040] In the example of FIGs. 1A-1D, the sound processing unit 106 is an off-the-ear (OTE) sound processing unit, sometimes referred to herein as an OTE component, that is configured to send data and power to the cochlear implant 112. In general, an OTE sound processing unit is a component having a generally cylindrically shaped housing 111 and which is configured to be magnetically coupled to the user’s head (e.g., includes an integrated external magnet 150 configured to be magnetically coupled to an implantable magnet 152 in the cochlear implant 112). The OTE sound processing unit 106 also includes an integrated external (headpiece) coil 108 that is configured to be inductively coupled to the implantable coil 114.

[0041] It is to be appreciated that the OTE sound processing unit 106 is merely illustrative of the external devices that could operate with cochlear implant 112. For example, in alternative examples, the external component can comprise a behind-the-ear (BTE) sound processing unit or a micro-BTE sound processing unit and a separate external. In general, a BTE sound processing unit comprises a housing that is shaped to be worn on the outer ear of the user and is connected to the separate external coil assembly via a cable, where the external coil assembly is configured to be magnetically and inductively coupled to the implantable coil 114. It is also to be appreciated that alternative external components could be located in the user’s ear canal, worn on the body, etc.

[0042] As noted above, the cochlear implant system 102 includes the sound processing unit 106 and the cochlear implant 112. However, as described further below, the cochlear implant 112 can operate independently from the sound processing unit 106, for at least a period, to stimulate the user. For example, the cochlear implant 112 can operate in a first general mode, sometimes referred to as an “external hearing mode,” in which the sound processing unit 106 captures sound signals which are then used as the basis for delivering stimulation signals to the user. The cochlear implant 112 can also operate in a second general mode, sometimes referred as an “invisible hearing” mode, in which the sound processing unit 106 is unable to provide sound signals to the cochlear implant 112 (e.g., the sound processing unit 106 is not present, the sound processing unit 106 is powered-off, the sound processing unit 106 is malfunctioning,Atty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1efc.). As such, in the invisible hearing mode, the cochlear implant 112 captures sound signals itself via implantable sound sensors and then uses those sound signals as the basis for delivering stimulation signals to the user. Further details regarding operation of the cochlear implant 112 in the external hearing mode are provided below, followed by details regarding operation of the cochlear implant 112 in the invisible hearing mode. It is to be appreciated that reference to the external hearing mode and the invisible hearing mode is merely illustrative and that the cochlear implant 112 could also operate in alternative modes. In FIGs. 1 A and 1C, the cochlear implant system 102 is shown with an external device 110, configured to implement aspects of the techniques presented. The external device 110 is a computing device, such as a computer (e.g., laptop, desktop, tablet), a mobile phone, remote control unit, etc.

[0043] In FIGs. 1A-1D, the OTE sound processing unit 106 comprises one or more input devices that are configured to receive input signals (e.g., sound or data signals). The one or more input devices include one or more sound input devices 118 (e.g., one or more external microphones, audio input ports, telecoils, etc.), one or more auxiliary input devices 128 (e.g., audio ports, such as a Direct Audio Input (DAI), data ports, such as a Universal Serial Bus (USB) port, cable port, efc.), and a wireless transmitter / receiver (transceiver) 120 (e.g., for communication with the external device 110). However, it is to be appreciated that one or more input devices can include additional types of input devices and / or less input devices (e.g., the wireless short range radio transceiver 120 and / or one or more auxiliary input devices 128 could be omitted).

[0044] The OTE sound processing unit 106 also comprises the external coil 108, a charging coil 121, a closely-coupled transmitter / receiver (RF transceiver) 122, sometimes referred to as or radio-frequency (RF) transceiver 122, at least one rechargeable battery 132, and an external sound processing module 124. The external sound processing module 124 can comprise, for example, one or more processors and a memory device (memory) that includes sound processing logic. The memory device can comprise any one or more of: Non-Volatile Memory (NVM), Ferroelectric Random Access Memory (FRAM), read only memory (ROM), random access memory (RAM), magnetic disk storage media devices, optical storage media devices, flash memory devices, electrical, optical, or other physical / tangible memory storage devices. The one or more processors are, for example, microprocessors or microcontrollers that execute instructions for the sound processing logic stored in memory device.

[0045] The cochlear implant 112 comprises an implant body (main module) 134, a lead region 136, and the intra-cochlear stimulating assembly 116, all configured to be implanted under theAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1skin / tissue (tissue) 115 of the user. The implant body 134 generally comprises a hermetically-sealed housing 138 in which RF interface circuitry 140 and a stimulator unit 142 are disposed. The implant body 134 also includes the internal / implantable coil 114 that is generally external to the housing 138, but which is connected to the RF interface circuitry 140 via a hermetic feedthrough (not shown in FIG. ID).

[0046] As noted, stimulating assembly 116 is configured to be at least partially implanted in the user’s cochlea. Stimulating assembly 116 includes a plurality of longitudinally spaced intra-cochlear electrical stimulating contacts (electrodes) 144 that collectively form a contact or electrode array 146 for delivery of electrical stimulation (current) to the user’s cochlea.

[0047] Stimulating assembly 116 extends through an opening in the user’s cochlea (e.g., cochleostomy, the round window, efc.) and has a proximal end connected to stimulator unit 142 via lead region 136 and a hermetic feedthrough (not shown in FIG. ID). Lead region 136 includes a plurality of conductors (wires) that electrically couple the electrodes 144 to the stimulator unit 142. The cochlear implant 112 also includes an electrode outside of the cochlea, sometimes referred to as the extra-cochlear electrode (ECE) 139.

[0048] As noted, the cochlear implant system 102 includes the external coil 108 and the implantable coil 114. The external magnet 152 is fixed relative to the external coil 108 and the implantable magnet 152 is fixed relative to the implantable coil 114. The magnets fixed relative to the external coil 108 and the implantable coil 114 facilitate the operational alignment of the external coil 108 with the implantable coil 114. This operational alignment of the coils enables the external component 104 to transmit data and power to the cochlear implant 112 via a closely-coupled wireless link 148 formed between the external coil 108 with the implantable coil 114. In certain examples, the closely-coupled wireless link 148 is a radio frequency (RF) link. However, various other types of energy transfer, such as infrared (IR), electromagnetic, capacitive and inductive transfer, can be used to transfer the power and / or data from an external component to an implantable component and, as such, FIG. ID illustrates only one example arrangement.

[0049] As noted above, sound processing unit 106 includes the external sound processing module 124. The external sound processing module 124 is configured to convert received input signals (received at one or more of the input devices) into output signals for use in stimulating a first ear of a user (i.e., the external sound processing module 124 is configured to perform sound processing on input signals received at the sound processing unit 106). StatedAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1differently, the one or more processors in the external sound processing module 124 are configured to execute sound processing logic in memory to convert the received input signals into output signals that represent electrical stimulation for delivery to the user.

[0050] As noted, FIG. ID illustrates an embodiment in which the external sound processing module 124 in the sound processing unit 106 generates the output signals. In an alternative embodiment, the sound processing unit 106 can send less processed information (e.g., audio data) to the cochlear implant 112 and the sound processing operations (e.g., conversion of sounds to output signals) can be performed by a processor within the cochlear implant 112.

[0051] Returning to the specific example of FIG. ID, the output signals are provided to the RF transceiver 122, which transcutaneously transfers the output signals (e.g., in an encoded manner) to the cochlear implant 112 via external coil 108 and implantable coil 114. That is, the output signals are received at the RF interface circuitry 140 via implantable coil 114 and provided to the stimulator unit 142. The stimulator unit 142 is configured to utilize the output signals to generate electrical stimulation signals (e.g., current signals) for delivery to the user’s cochlea. In this way, cochlear implant system 102 electrically stimulates the user’s auditory nerve cells, bypassing absent or defective hair cells that normally transduce acoustic vibrations into neural activity, in a manner that causes the user to perceive one or more components of the received sound signals.

[0052] As detailed above, in the external hearing mode the cochlear implant 112 receives processed sound signals from the sound processing unit 106. However, in the invisible hearing mode, the cochlear implant 112 is configured to capture and process sound signals for use in electrically stimulating the user’s auditory nerve cells. In particular, as shown in FIG. ID, the cochlear implant 112 includes a plurality of implantable sound sensors 160 and an implantable sound processing module 158. Similar to the external sound processing module 124, the implantable sound processing module 158 can comprise, for example, one or more processors and a memory device (memory) that includes sound processing logic. The memory device can comprise any one or more of Non-Volatile Memory (NVM), Ferroelectric Random Access Memory (FRAM), read only memory (ROM), random access memory (RAM), magnetic disk storage media devices, optical storage media devices, flash memory devices, electrical, optical, or other physical / tangible memory storage devices. The one or more processors are, for example, microprocessors or microcontrollers that execute instructions for the sound processing logic stored in memory device.Atty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1

[0053] In the invisible hearing mode, the implantable sound sensors 160 are configured to detect / capture signals (e.g., acoustic sound signals, vibrations, efc.), which are provided to the implantable sound processing module 158. The implantable sound processing module 158 is configured to convert received input signals (received at one or more of the implantable sound sensors 160) into output signals for use in stimulating the first ear of a user (i.e., the processing module 158 is configured to perform sound processing operations). Stated differently, the one or more processors in implantable sound processing module 158 are configured to execute sound processing logic in memory to convert the received input signals into output signals 156 that are provided to the stimulator unit 142. The stimulator unit 142 is configured to utilize the output signals 156 to generate electrical stimulation signals (e.g., current signals) for delivery to the user’s cochlea, thereby bypassing the absent or defective hair cells that normally transduce acoustic vibrations into neural activity.

[0054] It is to be appreciated that the above description of the so-called external hearing mode and the so-called invisible hearing mode are merely illustrative and that the cochlear implant system 102 could operate differently in different embodiments. For example, in one alternative implementation of the external hearing mode, the cochlear implant 112 could use signals captured by the sound input devices 118 and the implantable sound sensors 160 in generating stimulation signals for delivery to the user.

[0055] As noted above, presented herein are techniques for configuring / determining / setting one or more operational settings or parameters of a recipient device, such as one or more operational parameters of cochlear implant system 102, based on / using predicted operational settings or parameters generated by a machine learning model. For example, the techniques presented herein leverage a machine learning model trained on a priori information from past device configuration sessions associated with a recipient population and fine-tuned via expert evaluation to generate predicted operational parameters. Moreover, the techniques presented herein include determining and providing prediction uncertainty associated with operational parameters predicted by the machine learning model to assist medical practitioners in deciding which prediction can be reliable, and which parameter requires further configuration. Experts (e.g., medical practitioners and / or recipients) can provide interactive and iterative evaluation of predicted operational parameters generated by the machine learning model, thus enabling an efficient and effective device configuration process. Further, the techniques presented herein include iteratively refining predicted parameters to minimize the difference between predictedAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1and observed parameters, thus providing continuous improvement of predictions generated by the machine learning model.

[0056] Shown in FIG. 2 is system 200 that is configured to perform aspects of the techniques presented herein. In particular, system 200 is configured to perform a plurality of operations during a training phase 262 and an inference phase 264. The training phase 262 is configured to train a machine learning model to predict a plurality of parameters for configuring a recipient device (e.g., a medical device). In operation, during the training phase 262, a parameter database 265 is accessed to obtain one or more input parameters. The one or more input parameters can include one or more operational parameters 266 and / or one or more recipientspecific parameters 267. For example, the parameter database 265 can store a dataset D = {0D, QD} having M records of the one or more operational parameters 266 QD= {q°, ql, qM} and / or one or more recipient-specific parameters 267 0D= {0°, 0l, 0M} associated with a plurality of configuration sessions at time T = {T T2, —,Tn}, wherein i represents an index and n represents a total number of timestamps. In certain embodiments, each of the dataset D, the one or more operational parameters 266 QD, and the one or more recipient-specific parameters 2670Dcan be a matrix that includes one or more vectors.

[0057] In certain embodiments, the parameter database 265 can represent the one or more operational parameters 266 and the one or more recipient-specific parameters 267 using a vector representation. For example, each vector q stored in the parameter database 265 can include the one or more operational parameters 266, and each vector 0 can include the one or more recipient-specific parameters 267. In certain embodiments, the parameter database 265 can store the one or more operational parameters 266 and / or the one or more recipient-specific parameters 267 associated with a specific population of recipients, such as recipients of a specific type of device (e.g., cochlear implants). The parameter database 265 can be continuously updated as new data becomes available from additional configuration sessions.

[0058] For example, the parameter database 265 is configured to store a plurality of operational settings or parameters for configuring one or more recipient devices. A plurality of operational settings or parameters, such as stimulation settings / parameters of an implantable medical device (e.g., auditory prosthesis), can be obtained from the parameter database 265. For example, the one or more operational parameters 266 can include T-level, C-level, dynamic range, frequency map, or any parameter associated with the configuration and / or stimulationAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1of a recipient device. The one or more operational parameters 266 stored in the parameter database 265 can be collected from a plurality of configuration sessions for configuring one or more devices for one or more recipients (e.g., a recipient of a cochlear implant).

[0059] In addition to operational or stimulation parameters, the parameter database 265 is configured to store the one or more recipient-specific parameters 267. For example, the one or more recipient-specific parameters 267 can include sex, age, device implantation time, device implantation time interval, electrode impedance, electrically evoked compound action potential (eCAP), phoneme and speech test outcomes, past operational parameters associated with a recipient, anatomical features, or any other parameter associated with the recipient. The one or more recipient-specific parameters 267 can be obtained during a configuration session of a device, or obtained through another source (e.g., medical records). In certain embodiments, the one or more recipient-specific parameters 267 are not available (e.g., during an initial configuration session of a new recipient) or only partially available. In such embodiments, the one or more recipient-specific parameters 267 can be optional for the training of a machine learning model.

[0060] As noted above, the one or more operational parameters 266 and / or one or more recipient-specific parameters 267 can be obtained from the parameter database 265. Then, the one or more operational parameters 266 and / or one or more recipient-specific parameters 267 are provided as input to a machine learning model 268. The machine learning model 268 can include a neural network, support vector machines, K-nearest neighbors, transformer-based models, etc. Neural networks can include multi-layer perceptron or deep neural network, such as convolutional neural network, recurrent neural network, probabilistic autoencoder neural network, etc. In certain embodiments, the machine learning model 268 can be a probabilistic regression network such as a conditional multivariate Gaussian process model. The machine learning model 268 can be an unsupervised, semi-supervised, or supervised model.

[0061] Based on the one or more operational parameters 266 and / or one or more recipientspecific parameters 267, the machine learning model 268 can be trained to generate one or more predicted parameters. In one example, the machine learning model 268 is trained to predict (e.g., reconstruct) one or more operational parameters and / or one or more recipientspecific parameters received in its inputs (e.g., observed parameters). In another example, the machine learning model 268 is trained to predict one or more operational parameters and / or one or more recipient-specific parameters that are unobserved. That is, for example, theAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1machine learning model 268 can be trained to predict a T-level or C-level associated with an electrode in an array of electrodes for which no T-level or C-level data has been obtained.

[0062] In certain embodiments, the machine learning model 268 can be trained over the entire dataset D or a subset of the dataset D (denoted as D'). The machine learning model 268 can be trained to estimate a probability of a vector q (e.g., an operational parameter vector) from one or more qprev(e.g., one or more previously observed operational parameter vectors in a same configuration session) and / or from one or more qpast(e.g., one or more observed operational parameter vectors obtained via past / historical configuration sessions). The vector q can further include an observed vector q0having observed operational parameters and / or recipient-specific parameters and an unobserved vector quhaving unobserved operational parameters and / or recipient-specific parameters.

[0063] In certain embodiments, the machine learning model 268 can be trained to minimize one or more differences between one or more predicted parameters (e.g., represented by one or more predicted vectors of operational parameters) and one or more actual parameters (e.g., represented by one or more ground truth vectors of operational parameters). For example, the one or more differences can be determined using a loss function. The loss function can be an error-based loss function (e.g., mean-square error loss function or mean absolute error loss function), a binary cross entropy loss function, a categorical cross entropy loss function, or any suitable loss function. The machine learning model 268 can be iteratively trained by minimizing the loss generated by the loss function, thus resulting in continuous improvement in its predictions. In the training phase 262, the machine learning model 268 can be trained iteratively until a stopping criterion is met. The stopping criterion can include an accuracy threshold (e.g., prediction accuracy at 90% or higher), a total number of iterations, execution time, etc. The stopping criterion can be configurable based on user requirement and / or preference.

[0064] After a stopping criterion has been met, the training of the machine learning model 268 can be halted. At this point, the training phase 262 can conclude, and the inference phase 264 commences. For example, the inference phase 264 can take place during a configuration session of a recipient device for a specific recipient. In certain embodiments, the inference phase 264 can optionally include one or more operations performed by a dataset and / or machine learning model rectifier 269. In certain embodiments, one or more recipient-specific parameters 270 can be optionally obtained, and based on the one or more recipient-specificAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1parameters 270, the dataset and / or machine learning model rectifier 269 is configured to select / rectify a subset of the dataset D (denoted as D' = {0'D, Q'D}) and / or a corresponding machine learning model associated with D', thus resulting in selected parameters and / or machine learning model 271. In certain embodiments, the selected parameters can be provided as input to a trained machine learning model 274, which can be the machine learning model selected by the dataset and / or machine learning model rectifier 269. In certain embodiments, the trained machine learning model 274 can be the machine learning model 268 after it has been trained to meet a stopping criterion as described above. Further, in certain embodiments, the trained machine learning model 274 can be obtained from another source (e.g., external source).

[0065] Moreover, data associated with one or more historical configuration sessions 272 can be obtained. For example, the data obtained includes one or more historical operational parameters 273 (represented by one or more vectors qpastor< associated with configuring one or more recipient devices during the one or more historical configuration sessions 272 at time T < Tn, wherein Tnrepresents a current time. The one or more historical configuration sessions 272 can be past configuration sessions associated with configuring a recipient device for a specific recipient (e.g., the recipient currently undergoing a configuration session) or past configuration sessions associated with a specific recipient population (e.g., recipients of the same type of recipient device being configured in a current configuration session). For example, during an initial configuration session for a recipient, past operational parameters associated with the recipient device are not available because the recipient device has not been configured. In this scenario, the one or more historical operational parameters 273 can include operational parameters from past configuration sessions associated with the corresponding recipient population. In subsequent configuration sessions, the one or more historical operational parameters 273 can include operational parameters from past configuration sessions associated with the recipient.

[0066] The trained machine learning model 274 can be executed iteratively. In each iteration, based on the one or more historical operational parameters 273 (represented by vector qpastor qT< rnthe selected parameters and / or machine learning model 271, and optionally the one or more recipient-specific parameters 270 (represented by vector 0), the trained machine learning model 274 is configured to determine a probability function 275 p for estimating a probability of a predicted operational parameter vector q that includes predicted observed operational parameters and / or predicted unobserved operational parameters. That is, for example, theAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1probability function 275 can be expressed as p (q = {q0, qu] | qPrev> T <Tn> wherein q0represents predicted observed operational parameters, qurepresents predicted unobserved operational parameters, and qprevrepresents previous operational parameters (which can be the same as qpastor qT < Tnduring an initial configuration session).

[0067] The predicted unobserved operational parameters 276 (represented by vector qu) can be obtained by maximizing the probability determined using the probability function 275. Further, the probability function 275 can be evaluated by one or more experts, and the evaluation provides expert adjustments 277 to tune the predicted unobserved operational parameters 276. That is, for example, an expert (e.g., medical practitioner or recipient) can evaluate the probability function 275 and adjusts / tunes one or more of the predicted unobserved operational parameters 276 based on loudness feedback from the recipient. In certain embodiments, the trained machine learning model 274 can suggest which parameters should be adjusted based on a probability distribution of the operational parameters. Based on the expert adjustments 277, one or more updated observed operational parameters 278 are generated.

[0068] Then, the predicted unobserved operational parameters 276 and the one or more updated observed operational parameters 278 are provided as inputs to a union operation 279, which generates a unified vector of updated operational parameters 280. The updated operational parameters 280 (expressed as a unified vector) are provided as feedback to the trained machine learning model 274 for the next iteration as a vector of previous operational parameters qprev. Then, the trained machine learning model 274 can calculate an updated probability function p based on the new values of qprev. The trained machine learning model 274 can also be iteratively refined based on the updated operational parameters 280 until a stopping criterion is met. The stopping criterion can include an accuracy threshold (e.g., prediction accuracy at 90% or higher), a total number of iterations, execution time, etc. The stopping criterion can be configurable based on user requirement and / or preference.

[0069] In certain embodiments, the trained machine learning model 274 can determine q0(e.g., predicted observed operational parameters) via a deterministic approach. In the deterministic approach, the trained machine learning model 274 does not calculate a probability function as in a probabilistic approach. Instead, the probability of q0calculated by the trained machine learning model 274 takes a single value (e.g., the value equals to 1.0 at qoprev (previous observed operational parameters)). In certain embodiments, the trained machine learningAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1model 274 can determine q0via a probabilistic approach where the trained machine learning model 274 generates a probability function (e.g., probability function 275). The probabilistic approach takes into consideration the possibility of imperfect expert observation (e.g., expert adjustments 277) by using a probability value.

[0070] As illustrated in FIG. 3, the system 300 is configured to perform a plurality of operations during a training phase 362 and an inference phase 364. The training phase 362 is configured to train a probabilistic regression network to predict a plurality of parameters for configuring a recipient device (e.g., a medical device). In operation, during the training phase 362, a parameter database 365 is accessed to obtain one or more input parameters. The one or more input parameters can include one or more operational parameters 366 and / or one or more recipient-specific parameters 367. For example, the parameter database 365 can store a dataset D = {&D> QD} having M records of the one or more operational parameters 266 QD= {q°, ql, qM} and / or one or more recipient-specific parameters 367 0D= {0°, 0l, 0M} associated with a plurality of configuration sessions at time T = {T T2, —,Tn}, wherein i represents an index and n represents a total number of timestamps. In certain embodiments, each of the dataset D, the one or more operational parameters 366 QD, and the one or more recipient-specific parameters 3670Dcan be a matrix that includes one or more vectors.

[0071] In certain embodiments, the parameter database 365 can represent the one or more operational parameters 366 and the one or more recipient-specific parameters 367 using a vector representation. For example, each vector q stored in the parameter database 365 can include the one or more operational parameters 366, and each vector 0 can include the one or more recipient-specific parameters 367. In certain embodiments, the parameter database 365 can store the one or more operational parameters 366 and / or the one or more recipient-specific parameters 367 associated with a specific population of recipients, such as recipients of a specific type of device (e.g., cochlear implants). The parameter database 365 can be continuously updated as new data becomes available from additional configuration sessions.

[0072] For example, the parameter database 365 is configured to store a plurality of operational settings or parameters for configuring one or more recipient devices. A plurality of operational settings or parameters, such as stimulation settings / parameters of an implantable medical device (e.g., auditory prosthesis), can be obtained from the parameter database 365. For example, the one or more operational parameters 366 can include T-level, C-level, dynamicAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1range, frequency map, or any parameter associated with the configuration and / or stimulation of a recipient device. The one or more operational parameters 366 stored in the parameter database 365 can be collected from a plurality of configuration sessions for configuring one or more devices for one or more recipients (e.g., a recipient of a cochlear implant).

[0073] In addition to operational or stimulation parameters, the parameter database 365 is configured to store the one or more recipient-specific parameters 367. For example, the one or more recipient-specific parameters 367 can include sex, age, device implantation time, device implantation time interval, electrode impedance, electrically evoked compound action potential (eCAP), phoneme and speech test outcomes, past operational parameters associated with a recipient, anatomical features, or any other parameter associated with the recipient. The one or more recipient-specific parameters 367 can be obtained during a configuration session of a device, or obtained through another source (e.g., medical records). In certain embodiments, the one or more recipient-specific parameters 367 not be available (e.g., during an initial configuration session of a new recipient) or only partially available. In such embodiments, the one or more recipient-specific parameters 367 can be optional for the training of a probabilistic regression network.

[0074] As noted above, the one or more operational parameters 366 and / or one or more recipient-specific parameters 367 can be obtained from the parameter database 365. Then, the one or more operational parameters 366 and / or one or more recipient-specific parameters 367 are provided as input to a probabilistic regression network 368. The probabilistic regression network 368 can include a multi-layer perceptron or deep neural network, such as convolutional neural network, recurrent neural network, probabilistic autoencoder neural network, etc.

[0075] Based on the one or more operational parameters 366 and / or one or more recipientspecific parameters 367, the probabilistic regression network 368 can be trained to generate one or more predicted parameters. In one example, the probabilistic regression network 368 is trained to predict (e.g., reconstruct) one or more operational parameters and / or one or more recipient-specific parameters received in its inputs (e.g., observed parameters). In another example, the probabilistic regression network 368 is trained to predict one or more operational parameters and / or one or more recipient-specific parameters that are unobserved. That is, for example, the probabilistic regression network 368 can be trained to predict a T-level or C-level associated with an electrode in an array of electrodes for which no T-level or C-level data has been obtained.Atty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1

[0076] In certain embodiments, the probabilistic regression network 368 can be trained over the entire dataset D or a subset of the dataset D (denoted as D'). The probabilistic regression network 368 can be trained to estimate a probability of a vector q (e.g., an operational parameter vector) from one or more qprev(e.g., one or more previously observed operational parameter vectors in a same configuration session) and / or from one or more qpast(e.g., one or more observed operational parameter vectors obtained via past / historical configuration sessions). The vector q can further include an observed vector q0having observed operational parameters and / or recipient-specific parameters and an unobserved vector quhaving unobserved operational parameters and / or recipient-specific parameters.

[0077] In certain embodiments, the probabilistic regression network 368 can be trained to minimize one or more differences between one or more predicted parameters (e.g., represented by one or more predicted vectors of operational parameters) and one or more actual parameters (e.g., represented by one or more ground truth vectors of operational parameters). For example, the one or more differences can be determined using a loss function. The loss function can be an error-based loss function (e.g., mean-square error loss function or mean absolute error loss function), a binary cross entropy loss function, a categorical cross entropy loss function, or any suitable loss function. The probabilistic regression network 368 can be iteratively trained by minimizing the loss generated by the loss function, thus resulting in continuous improvement in its predictions. In the training phase 362, the probabilistic regression network 368 can be trained iteratively until a stopping criterion is met. The stopping criterion can include an accuracy threshold (e.g., prediction accuracy at 90% or higher), a total number of iterations, execution time, etc. The stopping criterion can be configurable based on user requirement and / or preference.

[0078] After a stopping criterion has been met, the training of the probabilistic regression network 368 can be halted. At this point, the training phase 362 can conclude, and the inference phase 364 commences. For example, the inference phase 364 can take place during a configuration session of a recipient device for a specific recipient. In certain embodiments, the inference phase 364 can optionally include one or more operations performed by a dataset and / or probabilistic regression network rectifier 369. In certain embodiments, one or more recipient-specific parameters 370 can be optionally obtained, and based on the one or more recipient-specific parameters 370, the dataset and / or probabilistic regression network rectifier 369 is configured to select / rectify a subset of the dataset D (denoted as D' = {0'D, Q'D}) and / or a corresponding probabilistic regression network associated with D', thus resulting in selectedAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1parameters and / or probabilistic regression network 371. In certain embodiments, the selected parameters can be provided as input to a trained probabilistic regression network 374, which can be the probabilistic regression network selected by the dataset and / or probabilistic regression network rectifier 369. In certain embodiments, the trained probabilistic regression network 374 can be the probabilistic regression network 368 after it has been trained to meet a stopping criterion as described above. Further, in certain embodiments, the trained probabilistic regression network 374 can be obtained from another source (e.g., external source).

[0079] Moreover, data associated with one or more historical configuration sessions 372 can be obtained. For example, the data obtained includes one or more historical operational parameters 373 (represented by one or more vectors qpastor< associated with configuring one or more recipient devices during the one or more historical configuration sessions 372 at time T < Tn, wherein Tnrepresents a current time. The one or more historical configuration sessions 372 can be past configuration sessions associated with configuring a recipient device for a specific recipient (e.g., the recipient currently undergoing a configuration session) or past configuration sessions associated with a specific recipient population (e.g., recipients of the same type of recipient device being configured in a current configuration session). For example, during an initial configuration session for a recipient, past operational parameters associated with the recipient device are not available because the recipient device has not been configured. In this scenario, the one or more historical operational parameters 373 can include operational parameters from past configuration sessions associated with the corresponding recipient population. In subsequent configuration sessions, the one or more historical operational parameters 373 can include operational parameters from past configuration sessions associated with the recipient.

[0080] The trained probabilistic regression network 374 can be executed iteratively. In each iteration, based on the one or more historical operational parameters 373 (represented by vector pastorQr<Tn)^ the selected parameters and / or probabilistic regression network 371, and optionally the one or more recipient-specific parameters 370 (represented by vector 0), the trained probabilistic regression network 374 is configured to determine a probability function 375 p for estimating a probability of a predicted operational parameter vector q that includes predicted observed operational parameters and / or predicted unobserved operational parameters. That is, for example, the probability function 375 can be expressed as p (q = {q0, qu} | qprev >qr <Tn>^> D'), wherein q0represents predicted observed operationalAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1parameters, qurepresents predicted unobserved operational parameters, and qprevrepresents previous operational parameters (which can be the same as qpastor qT < Tnduring an initial configuration session).

[0081] The predicted unobserved operational parameters 376 (represented by vector qu) can be obtained by maximizing the probability determined using the probability function 375. Further, the probability function 375 can be evaluated by one or more experts, and the evaluation provides expert adjustments 377 to tune the predicted unobserved operational parameters 376. That is, for example, an expert (e.g., medical practitioner or recipient) can evaluate the probability function 375 and adjusts / tunes one or more of the predicted unobserved operational parameters 376 based on loudness feedback from the recipient. In certain embodiments, the trained probabilistic regression network 374 can suggest which parameters should be adjusted based on a probability distribution of the operational parameters. Based on the expert adjustments 377, one or more updated observed operational parameters 378 are generated.

[0082] Then, the predicted unobserved operational parameters 376 and the one or more updated observed operational parameters 378 are provided as inputs to a union operation 379, which generates a unified vector of updated operational parameters 380. The updated operational parameters 380 (expressed as a unified vector) are provided as feedback to the trained probabilistic regression network 374 for the next iteration as a vector of previous operational parameters qprev. Then, the trained probabilistic regression network 374 can calculate an updated probability function p based on the new values of qprev. The trained probabilistic regression network 374 can also be iteratively refined based on the updated operational parameters 380 until a stopping criterion is met. The stopping criterion can include an accuracy threshold (e.g., prediction accuracy at 90% or higher), a total number of iterations, execution time, etc. The stopping criterion can be configurable based on user requirement and / or preference.

[0083] In certain embodiments, the trained probabilistic regression network 374 can determine q0(e.g., predicted observed operational parameters) via a deterministic approach. In the deterministic approach, the trained probabilistic regression network 374 does not calculate a probability function as in a probabilistic approach. Instead, the probability of q0calculated by the trained probabilistic regression network 374 takes a single value (e.g., the value equals to 1.0 at qoprev (previous observed operational parameters)). In certain embodiments, the trainedAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1probabilistic regression network 374 can determine q0via a probabilistic approach where the trained probabilistic regression network 374 generates a probability function (e.g., probability function 375). The probabilistic approach takes into consideration the possibility of imperfect expert observation (e.g., expert adjustments 377) by using a probability value.

[0084] As illustrated in FIG. 4, the system 400 is configured to perform a plurality of operations during a training phase 401 and an inference phase 403. The training phase 401 is configured to train a machine learning model to predict a plurality of parameters for configuring a recipient device (e.g., a medical device). In operation, during the training phase 401, a parameter database 405 is accessed to obtain one or more input parameters. The one or more input parameters can include one or more operational parameters 407 and / or one or more recipient-specific parameters 409. For example, the parameter database 405 can store a dataset D = {&D> QD} having M records of the one or more operational parameters 407 QD= {q°, ql, qM} and / or one or more recipient-specific parameters 409 0D= {0°, 0l, 0M} associated with a plurality of configuration sessions at time T = {T T2, ... , Tn}, wherein i represents an index and n represents a total number of timestamps.

[0085] In certain embodiments, the parameter database 405 can represent the one or more operational parameters 407 and the one or more recipient-specific parameters 409 using a vector representation. For example, each vector q stored in the parameter database 405 can include the one or more operational parameters 407, and each vector 0 can include the one or more recipient-specific parameters 409. In certain embodiments, the parameter database 405 can store the one or more operational parameters 407 and / or the one or more recipient-specific parameters 409 associated with a specific population of recipients, such as recipients of a specific type of device (e.g., cochlear implants). The parameter database 405 can be continuously updated as new data becomes available from additional configuration sessions.

[0086] For example, the parameter database 405 is configured to store a plurality of operational settings or parameters for configuring one or more recipient devices. A plurality of operational settings or parameters, such as stimulation settings / parameters of an implantable medical device (e.g., auditory prosthesis), can be obtained from the parameter database 405. For example, the one or more operational parameters 407 can include T-level, C-level, dynamic range, frequency map, or any parameter associated with the configuration and / or stimulation of a recipient device. The one or more operational parameters 407 stored in the parameterAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1database 405 can be collected from a plurality of configuration sessions for configuring one or more devices for one or more recipients (e.g., a recipient of a cochlear implant).

[0087] In addition to operational or stimulation parameters, the parameter database 405 is configured to store the one or more recipient-specific parameters 409. For example, the one or more recipient-specific parameters 409 can include sex, age, device implantation time, device implantation time interval, electrode impedance, electrically evoked compound action potential (eCAP), phoneme and speech test outcomes, past operational parameters associated with a recipient, anatomical features, or any other parameter associated with the recipient. The one or more recipient-specific parameters 409 can be obtained during a configuration session of a device, or obtained through another source (e.g., medical records). In certain embodiments, the one or more recipient-specific parameters 409 are not available (e.g., during an initial configuration session of a new recipient) or only partially available. In such embodiments, the one or more recipient-specific parameters 409 can be optional for the training of a machine learning model.

[0088] As noted above, the one or more operational parameters 407 and / or one or more recipient-specific parameters 409 can be obtained from the parameter database 405. Then, the one or more operational parameters 407 and / or one or more recipient-specific parameters 409 are provided as input to a machine learning model 411. The machine learning model 411 can include a neural network, support vector machines, K-nearest neighbors, transformer-based models, etc. Neural networks can include multi-layer perceptron or deep neural network, such as convolutional neural network, recurrent neural network, probabilistic autoencoder neural network, etc. In certain embodiments, the machine learning model 411 can be a probabilistic regression network or any regression model implemented via machine learning. The machine learning model 411 can be an unsupervised, semi-supervised, or supervised model.

[0089] Based on the one or more operational parameters 407 and / or one or more recipientspecific parameters 409, the machine learning model 411 can be trained to generate one or more predicted parameters. That is, for example, the machine learning model 411 is trained to predict one or more operational parameters associated with a configuration session at time Tnbased on the one or more operational parameters 407 associated with time T = {T T2, —’Tn-i} and the one or more recipient-specific parameters 409. Then, one or more predicted operational parameters 413 (represented by Yn= func(Yx) where x = {x1,x2' ■■■'■^n-i}) generated by the machine learning model 411 can be compared to one or more actual operational parameters 415 Ynobtained from the parameter database 405. The oneAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1or more actual operational parameters 415 are considered ground truths for evaluating accuracy of the machine learning model 411.

[0090] The one or more predicted operational parameters 413 (represented by Fn) and the one or more actual operational parameters 415 are provided as input to a subtract operation 417 to determine one or more differences, or error values 419. That is, for example, the one or more actual operational parameters 415 are subtracted from the one or more predicted operational parameters 413 to determine the error values 419 (e.g., error values 419 E = Yn— Fn). In certain embodiments, the error values 419 are determined for each recipient in the recipient population. Based on the error values 419, one or more covariance values can be calculated at operation 421 to generate a covariance matrix 422 (represented by £n), which expresses the uncertainty associated with the one or more predicted operational parameters 413 Yn. The covariance matrix 422 determines one or more tuning parameters of a second machine learning model 445 (such as a neural network, a conditional multivariate Gaussian process model, etc.). It should be understood that while a covariance matrix is used as an exemplary error metrics matrix in FIG. 4, other types of suitable error metrics matrix can be used in some embodiments. In certain embodiments, the error values 419 can be determined using a loss function. The loss function can be an error-based loss function (e.g., mean-square error loss function or mean absolute error loss function), a binary cross entropy loss function, a categorical cross entropy loss function, or any suitable loss function. In certain embodiments, the machine learning model 411 can be iteratively trained by minimizing the loss generated by the loss function, thus resulting in continuous improvement in its predictions. In the training phase 401, the machine learning model 411 can be trained iteratively until a stopping criterion is met. The stopping criterion can include an accuracy threshold (e.g., prediction accuracy at 90% or higher), a total number of iterations, execution time, etc. The stopping criterion can be configurable based on user requirement and / or preference.

[0091] After a stopping criterion has been met, the training of the machine learning model 411 can be halted. At this point, the training phase 401 can conclude, and the inference phase 403 commences. For example, the inference phase 403 can take place during a configuration session of a recipient device for a specific recipient. Data associated with one or more historical configuration sessions 423 can be obtained. For example, the data obtained includes one or more historical operational parameters 427 (represented by one or more vectors qpastorqT < Tn) associated with configuring one or more recipient devices during the one or moreAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1historical configuration sessions 423 at time T < Tn, T = {T T2, ■■■ wherein Tnrepresents a current time.

[0092] The one or more historical configuration sessions 423 can be past configuration sessions associated with configuring a recipient device for a specific recipient (e.g., the recipient currently undergoing a configuration session) or past configuration sessions associated with a specific recipient population (e.g., recipients of the same type of recipient device being configured in a current configuration session). For example, during an initial configuration session for a recipient, past operational parameters associated with the recipient device are not available because the recipient device has not been configured. In this scenario, the one or more historical operational parameters 427 can include operational parameters from past configuration sessions associated with the corresponding recipient population. In subsequent configuration sessions, the one or more historical operational parameters 427 can include operational parameters from past configuration sessions associated with the recipient. For example, in a subsequent configuration session at time T2, one or more operational parameters associated with the recipient can have been modified. Thus, one or more operational parameters at time T2can be predicted and / or estimated based on operational parameters from the past configuration session at T and / or impedance values. In certain embodiments, auxiliary parameters (e.g., device implantation time) can be used in the prediction to improve the prediction accuracy.

[0093] The one or more historical operational parameters 427 can be provided as input to a first machine learning model 429 (e.g., a trained machine learning model), which can be the machine learning model 411 after it has been trained to meet a stopping criterion as described above. Further, in certain embodiments, the first machine learning model 429 can be a trained machine learning model obtained from another source (e.g., external source). Based on the one or more historical operational parameters 427, the first machine learning model 429 is configured to generate one or more predicted current operational parameters associated with a recipient R at time Tn, and qprev, which represents a vector of one or more predicted previous operational parameters 441. Further, a mean vector 443 p = pRis taken from the one or more predicted current operational parameters associated with the recipient R at time Tn(e.g., pR= YRn)-

[0094] The one or more predicted previous operational parameters 441 are provided as input to initialize a second machine learning model 445. Then, after the second machine learningAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1model 445 is initialized, the mean vector 443 (of one or more predicted current operational parameters associated with a recipient R) and the covariance matrix 422 (determined based on parameters associated with a recipient population) are provided as input to the second machine learning model 445. That is, for example, a probability distribution associated with the second machine learning model 445 is initialized with the mean vector 443 and the covariance matrix 422. The second machine learning model 445 can be any suitable machine learning model that is configured to estimate a probability of a predicted operational parameter vector. In certain embodiments, the second machine learning model 445 can be a probabilistic regression network configured to output a probability function that follows a Multivariate Gaussian Distribution (expressed by a mean vector p and a covariance matrix 2 = £n).

[0095] It is noted that the predicted previous operational parameters 441 and the mean vector 443 are not provided at the same time to the second machine learning model 445. Instead, the predicted previous operational parameters 441 are used as an initial value to start the loop (when the mean vector 443 is not available yet). When the mean vector 443 is available, then the predicted previous operational parameters 441 are not used anymore.

[0096] In certain embodiments, the probability distribution can be determined via a recipientspecific approach such that it can be applicable for predicting operational parameters for any configuration session (initial and / or subsequent sessions for one or more recipients). That is, for example, the mean vector 443 p for the probability distribution for a given recipient can be taken from predicted operational parameters associated with the recipient. Although the covariance matrix 422 can be directly taken from the recipient population (e.g., determined based on measured operational parameters associated with the recipient population), the covariance matrix 422 would not necessarily express the uncertainty of the predicted previous operational parameters 441 of the recipient at time Tn. Instead, the covariance matrix 422 is a covariance matrix of the error values 419 and thus represents the uncertainty of the first machine learning model 429 in its predicted operational parameters (including the predicted operational parameters of the recipient in a current session). Thus, this approach leverages a population-based covariance matrix to express recipient-specific data uncertainty. Further, the covariance matrix of the recipient can be estimated based on covariance matrix of the recipient population (e.g., through the covariance matrix 422 of the error values 419) while removing the impact of the recipient population from the predicted operational parameters. Estimating the covariance matrix of the recipient based on covariance matrix of the recipient populationAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1ensures that accurate predictions can still be generated in scenarios where the patient-specific covariance matrix alone may not provide meaningful information.

[0097] Based on the probability distribution characterized by the mean vector 443 p and the covariance matrix 422 £n, the second machine learning model 445, via the Gaussian Process, is configured to determine a probability function 447 p(q) for estimating a probability of a predicted operational parameter vector q that includes one or more predicted operational parameters 449. The one or more predicted operational parameters 449 (e.g., new T-levels and / or C-levels) can be represented by p and the associated uncertainty can be represented by covariance matrix 2. For example, the one or more predicted operational parameters 449 can include predicted observed operational parameters and / or predicted unobserved operational parameters. Further, the probability function 447 can be evaluated by one or more experts, and the evaluation provides expert adjustments 451 to tune the one or more predicted operational parameters 449. In one example, an expert (e.g., medical practitioner or recipient) can evaluate the probability function 447 and adjusts / tunes the one or more predicted operational parameters 449 based on loudness feedback from the recipient. In another example, an expert (e.g., medical practitioner or recipient) can provide new inputs (e.g., new observations and / or measurements) based on which the one or more predicted operational parameters 449 can be adjusted. Based on the expert adjustments 451, one or more updated operational parameters 453 (expressed as vector q(updated ) are generated.

[0098] The one or more updated operational parameters 453 are provided as feedback to refine the second machine learning model 445 in the next iteration. Then, the second machine learning model 445 can provide one or more predicted operational parameters 449 as output based on the new values of the one or more updated operational parameters 453. The second machine learning model 445 can also be iteratively refined based on the one or more updated operational parameters 453 until a stopping criterion is met. The stopping criterion can include an accuracy threshold (e.g., prediction accuracy at 90% or higher), a total number of iterations, execution time, etc. The stopping criterion can be configurable based on user requirement and / or preference. Upon meeting a stopping criterion, the second machine learning model 445 can provide the one or more predicted operational parameters 449 as output. Thus, the one or more predicted operational parameters 449 can be provided as output as one or more output operational parameters 455. In certain embodiments, the one or more output operational parameters 455 can be utilized by a medical practitioner (e.g., audiologist) to configure a recipient device during a configuration session.Atty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1

[0099] As illustrated in FIG. 5, the system 500 is configured to perform a plurality of operations during a training phase 501 and an inference phase 503. The training phase 501 is configured to train a machine learning model to predict a plurality of parameters for configuring a recipient device (e.g., a medical device). In operation, during the training phase 501, a parameter database 505 is accessed to obtain one or more input parameters. The one or more input parameters can include one or more operational parameters 507 and / or one or more recipient-specific parameters 509. For example, the parameter database 505 can store a dataset D = {&D> QD} having M records of the one or more operational parameters 507 QD= {q°, ql, qM} and / or one or more recipient-specific parameters 509 0D= {0°, 0l, 0M} associated with a plurality of configuration sessions at time T = {T T2, ... , Tn}, wherein i represents an index and n represents a total number of timestamps.[ooioo] In certain embodiments, the parameter database 505 can represent the one or more operational parameters 507 and the one or more recipient-specific parameters 509 using a vector representation. For example, each vector q stored in the parameter database 505 can include the one or more operational parameters 507, and each vector 0 can include the one or more recipient-specific parameters 509. In certain embodiments, the parameter database 505 can store the one or more operational parameters 507 and / or the one or more recipient-specific parameters 509 associated with a specific population of recipients, such as recipients of a specific type of device (e.g., cochlear implants). The parameter database 505 can be continuously updated as new data becomes available from additional configuration sessions.[ooioi] For example, the parameter database 505 is configured to store a plurality of operational settings or parameters for configuring one or more recipient devices. A plurality of operational settings or parameters, such as stimulation settings / parameters of an implantable medical device (e.g., auditory prosthesis), can be obtained from the parameter database 505. For example, the one or more operational parameters 507 can include T-level, C-level, dynamic range, frequency map, or any parameter associated with the configuration and / or stimulation of a recipient device. The one or more operational parameters 507 stored in the parameter database 505 can be collected from a plurality of configuration sessions for configuring one or more devices for one or more recipients (e.g., a recipient of a cochlear implant).

[0102] In addition to operational or stimulation parameters, the parameter database 505 is configured to store the one or more recipient-specific parameters 509. For example, the one or more recipient-specific parameters 509 can include sex, age, device implantation time, device implantation time interval, electrode impedance, electrically evoked compound action potentialAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1(eCAP), phoneme and speech test outcomes, past operational parameters associated with a recipient, anatomical features, or any other parameter associated with the recipient. The one or more recipient-specific parameters 509 can be obtained during a configuration session of a device, or obtained through another source (e.g., medical records). In certain embodiments, the one or more recipient-specific parameters 509 are not available (e.g., during an initial configuration session of a new recipient) or only partially available. In such embodiments, the one or more recipient-specific parameters 509 can be optional for the training of a machine learning model.

[0103] As noted above, the one or more operational parameters 507 and / or one or more recipient-specific parameters 509 can be obtained from the parameter database 505. Then, the one or more operational parameters 507 and / or one or more recipient-specific parameters 509 are provided as input to a machine learning model 511. The machine learning model 511 can include a neural network, support vector machines, K-nearest neighbors, transformer-based models, etc. Neural networks can include multi-layer perceptron or deep neural network, such as convolutional neural network, recurrent neural network, probabilistic autoencoder neural network, etc. In certain embodiments, the machine learning model 511 can be a probabilistic regression network or any regression model implemented via machine learning. The machine learning model 511 can be an unsupervised, semi-supervised, or supervised model.

[0104] Based on the one or more operational parameters 507 and / or one or more recipientspecific parameters 509, the machine learning model 511 can be trained to generate one or more predicted parameters. That is, for example, the machine learning model 511 is trained to predict one or more operational parameters associated with a configuration session at time Tnbased on the one or more operational parameters 507 associated with time T = {T T2, —’Tn-i} and the one or more recipient-specific parameters 509. Then, one or more predicted operational parameters 513 (represented by Yn= func(Yx) where x = [x1,x2, ...,xn-1}) generated by the machine learning model 511 can be compared to one or more actual operational parameters 515 Ynobtained from the parameter database 505. The one or more actual operational parameters 515 are considered ground truths for evaluating accuracy of the machine learning model 511.

[0105] The one or more predicted operational parameters 513 (represented by Pn) and the one or more actual operational parameters 515 are provided as input to a subtract operation 517 to determine one or more differences, or error values 519. That is, for example, The one or moreAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1actual operational parameters 515 are subtracted from the one or more predicted operational parameters 513 to determine the error values 519 (e.g., error values 519 E = Yn— Fn). In certain embodiments, the error values 519 is determined for each recipient in the recipient population. Based on the error values 519, one or more covariance values can be calculated at operation 521 to generate a covariance matrix 522 (represented by fn), which expresses the uncertainty associated with the one or more predicted operational parameters 513 Yn. The covariance matrix 522 determines one or more tuning parameters of a probabilistic regression network 545 (such as a probabilistic neural network, a conditional multivariate Gaussian process model, etc.). It should be understood that while a covariance matrix is used as an exemplary error metrics matrix in FIG. 5, other types of suitable error metrics matrix can be used in some embodiments. In certain embodiments, the error values 519 can be determined using a loss function. The loss function can be an error-based loss function (e.g., mean-square error loss function or mean absolute error loss function), a binary cross entropy loss function, a categorical cross entropy loss function, or any suitable loss function. In certain embodiments, the machine learning model 511 can be iteratively trained by minimizing the loss generated by the loss function, thus resulting in continuous improvement in its predictions. In the training phase 501, the machine learning model 511 can be trained iteratively until a stopping criterion is met. The stopping criterion can include an accuracy threshold (e.g., prediction accuracy at 90% or higher), a total number of iterations, execution time, etc. The stopping criterion can be configurable based on user requirement and / or preference.

[0106] After a stopping criterion has been met, the training of the machine learning model 511 can be halted. At this point, the training phase 501 can conclude, and the inference phase 503 commences. For example, the inference phase 503 can take place during a configuration session of a recipient device for a specific recipient. Data associated with one or more historical configuration sessions 523 can be obtained. For example, the data obtained includes one or more historical operational parameters 527 (represented by one or more vectors qpastorqT<Tn) associated with configuring one or more recipient devices during the one or more historical configuration sessions 523 at time T < Tn, T = {Tp T2’ ■■■ >7n-i}> wherein Tnrepresents a current time.

[0107] The one or more historical configuration sessions 523 can be past configuration sessions associated with configuring a recipient device for a specific recipient (e.g., the recipient currently undergoing a configuration session) or past configuration sessions associated with a specific recipient population (e.g., recipients of the same type of recipientAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1device being configured in a current configuration session). For example, during an initial configuration session for a recipient, past operational parameters associated with the recipient device are not available because the recipient device has not been configured. In this scenario, the one or more historical operational parameters 527 can include operational parameters from past configuration sessions associated with the corresponding recipient population. In subsequent configuration sessions, the one or more historical operational parameters 527 can include operational parameters from past configuration sessions associated with the recipient. For example, in a subsequent configuration session at time T2, one or more operational parameters associated with the recipient can have been modified. Thus, one or more operational parameters at time T2can be predicted and / or estimated based on operational parameters from the past configuration session at T and / or impedance values. In certain embodiments, auxiliary parameters (e.g., device implantation time) can be used in the prediction to improve the prediction accuracy.

[0108] The one or more historical operational parameters 527 can be provided as input to a trained machine learning model 529, which can be the machine learning model 511 after it has been trained to meet a stopping criterion as described above. Further, in certain embodiments, the trained machine learning model 529 can be a trained machine learning model obtained from another source (e.g., external source). Based on the one or more historical operational parameters 527, the trained machine learning model 529 is configured to generate one or more predicted current operational parameters associated with a recipient R at time Tn, and qprev, which represents a vector of one or more predicted previous operational parameters 541. Further, a mean vector 543 p = pRis taken from the one or more predicted current operational parameters associated with the recipient R at time Tn(e.g., pR= YRn).

[0109] The one or more predicted previous operational parameters 541 are provided as input to initialize a probabilistic regression network 545. Then, after the probabilistic regression network 545 is initialized, the mean vector 543 (of one or more predicted current operational parameters associated with a recipient / ?) and the covariance matrix 522 (determined based on parameters associated with a recipient population) are provided as input to the probabilistic regression network 545. That is, for example, a probability distribution associated with the probabilistic regression network 545 is initialized with the mean vector 543 and the covariance matrix 522. The probabilistic regression network 545 can include a multi-layer perceptron or deep neural network, such as convolutional neural network, recurrent neural network, probabilistic autoencoder neural network, etc. In certain embodiments, the probabilisticAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1regression network 545 is configured to output a probability function that follows a Conditional Multivariate Gaussian Distribution (expressed by a mean vector p and a covariance matrix 2 = S ).[oono] It is noted that the predicted previous operational parameters 541 and the mean vector 543 are not provided at the same time to the probabilistic regression network 545. Instead, the predicted previous operational parameters 541 are used as an initial value to start the loop (when the mean vector 543 is not available yet). When the mean vector 543 is available, then the predicted previous operational parameters 541 are not used anymore.[oom] In certain embodiments, the probability distribution can be determined via a recipientspecific approach such that it can be applicable for predicting operational parameters for any configuration session (initial and / or subsequent sessions for one or more recipients). That is, for example, the mean vector 543 p for the probability distribution for a given recipient can be taken from predicted operational parameters associated with the recipient. Although the covariance matrix 522 can be taken from the recipient population (e.g., determined based on measured operational parameters associated with the recipient population), the covariance matrix 522 would not necessarily express the uncertainty of the predicted previous operational parameters 541 of the recipient at time Tn. Instead, the covariance matrix 522 is a covariance matrix of the error values 519 and thus represents the uncertainty of the trained machine learning model 529 in its predicted operational parameters (including the predicted operational parameters of the recipient in a current session). Thus, this approach leverages a populationbased covariance matrix to express recipient-specific data uncertainty. Further, the covariance matrix of the recipient can be estimated based on covariance matrix of the recipient population (e.g., through the covariance matrix 522 of the error values 519) while removing the impact of the recipient population from the predicted operational parameters. Estimating the covariance matrix of the recipient based on covariance matrix of the recipient population ensures that accurate predictions can still be generated in scenarios where the patient-specific covariance matrix alone may not provide meaningful information.

[0112] Based on the probability distribution characterized by the mean vector 543 p and the covariance matrix 522 £n, the probabilistic regression network 545, via a Gaussian Process (e.g., Conditional Multivariate Gaussian Process), is configured to determine a probability function 547 p(q) for estimating a probability of a predicted operational parameter vector q that includes one or more predicted operational parameters 549. The one or more predictedAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1operational parameters 549 (e.g., new T-levels and / or C-levels) can be represented by mean vector p and the associated uncertainty can be represented by covariance matrix 2. For example, the one or more predicted operational parameters 549 can include predicted observed operational parameters and / or predicted unobserved operational parameters.

[0113] In embodiments where the probability function 547 follows a Conditional Multivariate Gaussian Distribution expressed by the mean vector p and covariance matrix 2, the mean vector p and covariance matrix 2 associated with the one or more predicted operational parameters can be calculated by partitioning a TV-dimensional matrix x of operational parameter values, wherein N represents the number of dimensions. That is, for example, the TV-dimensionalmatrix x is partitioned as follows: x = with sizes Accordingly, p and 2r x 1 2n ^12are partitioned as follows: p = [ ] with sizes and 2 = with sizes (A - r) x 1 .^21 ^22.r x rwhere N is the total number of parameters and r is the (N - r) x rnumber of unobserved parameters. In certain embodiments, when a medical practitioner (e.g., audiologist) has not made any observation in a configuration session, r = N as all parameters are unobserved. Then, as the medical practitioner makes one or more observations, r (e.g., the number of unobserved parameters) decreases. When all N parameters have been observed by the medical practitioner, r = 0 as no unobserved parameter remains. The distribution of x conditional on x2= a is a multivariate normal distribution expressed by (x | x2= a) ~ A(p, 2) with mean p = px+ 2122221(a — p2) and covariance matrix 2 = 21;L— Si2S22S2i, wherein pxand 21;Lrespectively represent updated mean and updated covariance corresponding to unobserved operational parameters (e.g., unobserved C-levels and T-levels), and p2and 222respectively represent updated mean and updated covariance corresponding to measured operational parameters (e.g., measured C-levels and T-levels), and a represents a value. In certain embodiments, p2corresponds to q0(e.g., an observed operational parameter vector that can be updated by experts), and 222corresponds to p(q0), which is a deterministic function. Moreover, pxcorresponds to qu(e.g., an unobserved operational parameter vector) and 2n corresponds to p(quFurther, 212and 221correspond to the remaining terms in the full joint distribution p(q = {q0, qu}

[0114] In certain embodiments, the probability function 547 can be expressed as> <> wherein D' represents a subset of the dataset D and qprevrepresents previous operational parameters. As described above, the probability function 547 can be aAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1Multivariate Gaussian Distribution expressed by the mean vector p and a covariance matrix 2. The probabilistic regression network 545 can generate the probability function 547 characterized by a Conditional Multivariate Gaussian Distribution, thus the probabilistic regression network 545 can have a structure with two parameters (e.g., p and 2) calculated (or trained) over a subset of recipient population parameters from the parameter database 505 that have low distances (e.g., a distance is considered “low” if it is below a predetermined threshold) with the one or more recipient-specific parameters 509 0. In an initial configuration session at time 7 , past operational parameters (qpast ) associated with the recipient device are not available because the recipient device has not been configured. In this scenario, the one or more historical operational parameters 527 can include operational parameters from past configuration sessions associated with the corresponding recipient population.

[0115] In certain embodiments, 2 (e.g., covariance matrix of the population) can be iteratively updated upon new inputs (e.g., operational parameters of recipient device) provided by a medical practitioner (e.g., audiologist) in a configuration session to generate 2 (e.g., a patientspecific covariance matrix). When all inputs have been manually provided by the medical practitioner, 2 would be equal to 0, which means there is no more uncertainty on p (e.g., patientspecific operational parameters). As described above, in a subsequent configuration session at time T2, one or more operational parameters associated with the recipient can have been modified. Thus, one or more operational parameters at time T2can be predicted and / or estimated based on operational parameters from the past configuration session at T and / or impedance values. In certain embodiments, auxiliary parameters (e.g., device implantation time) can be used in the prediction to improve the prediction accuracy.

[0116] Further, the probability function 547 can be evaluated by one or more experts, and the evaluation provides expert adjustments 551 to tune the one or more predicted operational parameters 549. In one example, an expert (e.g., medical practitioner or recipient) can evaluate the probability function 547 and adjusts / tunes the one or more predicted operational parameters 549 based on loudness feedback from the recipient. In another example, an expert (e.g., medical practitioner or recipient) can provide new inputs (e.g., new observations and / or measurements) based on which the one or more predicted operational parameters 549 can be adjusted. Based on the expert adjustments 551, one or more updated operational parameters 553 (expressed as vector q(updated ) are generated.Atty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1

[0117] The one or more updated operational parameters 553 are provided as feedback to refine the probabilistic regression network 545 in the next iteration. Then, the probabilistic regression network 545 can provide one or more predicted operational parameters 549 as output based on the new values of the one or more updated operational parameters 553. The probabilistic regression network 545 can also be iteratively refined based on the one or more updated operational parameters 553 until a stopping criterion is met. The stopping criterion can include an accuracy threshold (e.g., prediction accuracy at 90% or higher), a total number of iterations, execution time, etc. The stopping criterion can be configurable based on user requirement and / or preference. Upon meeting a stopping criterion, the probabilistic regression network 545 can provide the one or more predicted operational parameters 549 as output. Thus, the one or more predicted operational parameters 549 can be provided as output as one or more output operational parameters 555. In certain embodiments, the one or more output operational parameters 555 can be utilized by a medical practitioner (e.g., audiologist) to configure a recipient device during a configuration session.

[0118] FIG. 6 is a flowchart illustrating a method 600, in accordance with certain embodiments presented herein. Method 600 describes details of obtaining one or more historical operational parameters that can be applied in certain embodiments presented herein. At 613, configuration session information associated with configuring a recipient device for a recipient is obtained. That is, for example, the configuration sessions information can include one or more timestamps associated with a configuration session (e.g., when the configuration commences and / or when the configuration session concludes), duration of the configuration session, type of configuration performed, and / or any other information associated with the configuration session. Then, at 615, a determination is made regarding whether a current configuration session is a first configuration session of the recipient device for the recipient.

[0119] Then, in response to determining that the current configuration session is a first configuration session of the recipient device for the recipient, the method 600 proceeds to obtain one or more past configuration sessions associated with a recipient population from a database (e.g., parameter database) at 617. For example, the recipient population can include recipients of the same type of recipient device that is being configured in the current configuration session. Then, one or more operational parameters of the one or more past configuration sessions associated with the recipient population are obtained at 619. In certain embodiments, the obtained one or more operational parameters can be provided as input to a machine learning model during a training phase and / or an inference phase.Atty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1

[0120] Alternatively, in response to determining that the current configuration session is not a first configuration session of the recipient device for the recipient, the method 600 proceeds to obtain one or more past configuration sessions associated with the recipient from the database (e.g., parameter database) at 621. Then, one or more operational parameters of the one or more past configuration sessions associated with the recipient are obtained at 623. In certain embodiments, the obtained one or more operational parameters can be provided as input to machine learning model during a training phase and / or an inference phase.

[0121] FIG. 7A is a graphical depiction illustrating a graph 700A for visualizing uncertainty associated with a plurality of predicted parameters, in accordance with certain embodiments presented herein. The graph 700A includes an x-axis representing an electrode number 781 identifying an electrode of a recipient device (e.g., electrode number 1 through 20 of a cochlear implant). The graph 700A further includes a y-axis representing operational parameters 783 corresponding to each of the electrodes. For example, the operational parameters can include C-level, T-level, or any operational parameter associated with configuring a recipient device.

[0122] The graph 700A further includes a line 785 representing predicted C-levels and a line 787 representing predicted T-levels. For example, the predicted C-levels and / or predicted T-levels can be generated by a machine learning model (e.g., probabilistic regression network) in accordance with certain embodiments presented herein. Moreover, the graph 700A further illustrates an observed measurement 788 (e.g., observed C-level associated with electrode number 11) along with a plurality of unobserved measurements 789. The graph 700A can further include a confidence area 791A to depict a level of confidence associated with one or more predicted operational parameters. In one example, since the observed measurement 788 has been made by a medical practitioner (e.g., audiologist), the confidence area 791A of predicted operational parameters surrounding the observed measurement 788 (e.g. measurement of electrode number 11) has a darker shade compared to the confidence areas (e.g., confidence area 791B and confidence area 791C) associated with other predicted operational parameters. That is, for example, darker shading around a certain predicted operational parameter indicates higher confidence and lower uncertainty associated with that specific prediction. In certain embodiments, each confidence area (e.g., confidence area 791 A) can be associated with an upper boundary and a lower boundary for the probability that a predicted operational parameter is a correct value. In these embodiments, darker shading of the confidence area (e.g., confidence area 791A) indicates the upper boundary and the lowerAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1boundary are coming closer together, and there is an increase in probability of finding the C-level in the vicinity of the expert’s adjustments.

[0123] The graph 700A enables a user to identify one or more electrodes that have predicted operational parameters with high uncertainty (or low confidence). This way, the confidence levels and / or uncertainty values allow users (e.g., a medical practitioner) to quickly determine which of the plurality of predicted operational parameters is reliable and which parameters require further clinical observation, thus providing increased efficiency and accuracy in the configuration of recipient devices. In certain embodiments, graph 700A can be displayed, via a graphical user interface (GUI), to a user (e.g., a medical practitioner) before, during, or after a recipient device configuration session. Based on the shaded confidence areas (e.g., confidence areas 791A-C) visualized in the graph 700A, the user can quickly and accurately decide which electrode needs further clinical observation (measurement) at a current or future session. For example, the user can speed up the recipient device configuration session by only obtaining new observations (measurements) of operational parameters for electrodes with prediction uncertainty higher than a predetermined threshold while relying on predicted parameters for the other electrodes. This provides a more targeted configuration experience, thus saving time and resources for the recipient and the medical practitioner. Further, augmenting model-generated parameter predictions with human observation and expertise ensures the recipient device is properly configured. In certain embodiments, the graphical user interface can be configured to receive input from the user (e.g., medical practitioner), such as feedback regarding the predicted operational parameters and / or the confidence areas 791 A-C. In certain embodiments, in addition to or instead of C-levels and / or T-levels, any predicted operational parameters can be visualized in the graph 700A in accordance with the techniques presented herein.

[0124] FIG. 7B is another graphical depiction illustrating a graph 700B for visualizing uncertainty associated with a plurality of predicted parameters, in accordance with certain embodiments presented herein. The graph 700B includes a line 793 representing predicted C-levels and a line 795 representing predicted T-levels. For example, the predicted C-levels and / or predicted T-levels can be generated by a machine learning model (e.g., probabilistic regression network) in accordance with certain embodiments presented herein. Moreover, the graph 700B further illustrates a plurality of observed measurements, including observed measurement 797A and observed measurement 797B (e.g., observed C-level values) as well as observed measurement 797C and observed measurement 797D (e.g., observed T-level values).Atty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1The graph 700 A can further depict a level of confidence associated with one or more predicted operational parameters. Predicted operational parameters in the area surrounding each of the observed measurements 797A-D are represented via a darker shade compared to predicted operational parameters corresponding to unobserved measurements. That is, for example, darker shading around a certain predicted operational parameter indicates higher confidence and lower uncertainty associated with that specific prediction. In certain embodiments, in addition to or instead of C-levels and / or T-levels, any predicted operational parameters can be visualized in the graph 700B in accordance with the techniques presented herein.

[0125] The graph 700B enables a user to identify one or more electrodes that have predicted operational parameters with high uncertainty (or low confidence). This way, the confidence levels and / or uncertainty values allow users (e.g., a medical practitioner) to quickly determine which of the plurality of predicted operational parameters is reliable and which parameters require further clinical observation, thus providing increased efficiency and accuracy in the configuration of recipient devices. In certain embodiments, graph 700B can be displayed, via a graphical user interface (GUI), to a user (e.g., a medical practitioner) before, during, or after a recipient device configuration session. Based on the shaded areas visualized in the graph 700B, the user can quickly and accurately decide which electrode needs further clinical observation (measurement) at a current or future session. For example, the user can speed up the recipient device configuration session by only obtaining new observations (measurements) of operational parameters for electrodes with prediction uncertainty higher than a predetermined threshold while relying on predicted parameters for the other electrodes. This provides a more targeted configuration experience, thus saving time and resources for the recipient and the medical practitioner. Further, augmenting model-generated parameter predictions with human observation and expertise ensures the recipient device is properly configured. In certain embodiments, the graphical user interface can be configured to receive input from the user (e.g., medical practitioner), such as feedback regarding the predicted operational parameters and / or the confidence associated with each of the predicted operational parameters.

[0126] FIG. 7C is another graphical depiction illustrating a graph 700C for visualizing uncertainty associated with a plurality of predicted operational parameters, in accordance with certain embodiments presented herein. The graph 700C includes an x-axis representing an electrode number 725 identifying an electrode of a recipient device (e.g., electrode number 1 through 20 of a cochlear implant). The graph 700C further includes a y-axis representingAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1operational parameters 727 corresponding to each of the electrodes. For example, the operational parameters can include C-level, T-level, or any operational parameter associated with configuring a recipient device.

[0127] The graph 700C further includes a line 729 representing predicted C-levels and a line 731 representing predicted T-levels. For example, the predicted C-levels and / or predicted T-levels can be generated by a machine learning model (e.g., probabilistic regression network) in accordance with certain embodiments presented herein. Moreover, the graph 700C further illustrates an observed measurement 733 (e.g., observed C-level associated with electrode number 10) along with a plurality of unobserved measurements. The graph 700C further includes one or more C-level 2 std (standard deviation) error zones 735 to depict uncertainty level associated with one or more predicted C-levels and one or more T-level 2 std (standard deviation) error zones 737 to depict uncertainty level associated with one or more predicted T-levels. In one example, since the observed measurement 733 has been made by a medical practitioner (e.g., audiologist), the one or more C-level 2 std (standard deviation) error zones 735 surrounding the observed measurement 733 have narrower spread than those that are further away from the observed measurement 733. As wider spread indicates higher uncertainty and narrower spread indicates lower uncertainty regarding the corresponding predicted operational parameters, predicted operational parameters in the areas closer to the observed measurement 733 have lower uncertainty. In certain embodiments, the confidence level, reflected via the corresponding error zone, is shown by an interval coming from a metric value applied to the probability function corresponding to the confidence interval. In certain embodiments, in addition to or instead of C-levels and / or T-levels, any predicted operational parameters can be visualized in the graph 700C in accordance with the techniques presented herein.

[0128] The graph 700C enables a user to identify one or more electrodes that have predicted operational parameters with high uncertainty (or low confidence). This way, the confidence levels and / or uncertainty values allow users (e.g., a medical practitioner) to quickly determine which of the plurality of predicted operational parameters is reliable and which parameters require further clinical observation, thus providing increased efficiency and accuracy in the configuration of recipient devices. In certain embodiments, graph 700C can be displayed, via a graphical user interface (GUI), to a user (e.g., a medical practitioner) before, during, or after a recipient device configuration session. Based on the one or more C-level 2 std (standard deviation) error zones 735 and one or more T-level 2 std (standard deviation) error zones 737Atty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1visualized in the graph 700C, the user can quickly and accurately decide which electrode needs further clinical observation (measurement) at a current or future session. For example, the one or more C-level 2 std error zones 735 and one or more T-level 2 std error zones 737 corresponding to the predicted operational parameters provide visual emphasis on one or more electrodes based on uncertainty associated with the corresponding predicted operational parameters.

[0129] For example, the user can speed up the recipient device configuration session by only obtaining new observations (measurements) of operational parameters for electrodes with prediction uncertainty higher than a predetermined threshold while relying on predicted parameters for the other electrodes. This provides a more targeted configuration experience, thus saving time and resources for the recipient and the medical practitioner. Further, augmenting model-generated parameter predictions with human observation and expertise ensures the recipient device is properly configured. In certain embodiments, the graphical user interface can be configured to receive input from the user (e.g., medical practitioner), such as feedback regarding the predicted operational parameters and / or the error zones.

[0130] With reference to FIGs. 7D-7K shown are respective graphical depictions 700D-700K illustrating a plurality of graphs for visualizing uncertainty values associated with a plurality of predicted parameters, in accordance with certain embodiments presented herein. The graphical depictions 700D-700K includes a graph 739A (FIG. 7D), a graph 739B (FIG. 7E), a graph 739C (FIG. 7F), a graph 739D (FIG. 7G), a graph 739E (FIG. 7H), a graph 739F (FIG.71), a graph 739G (FIG. 7J), and a graph 739H (FIG. 7K) each visualizing uncertainty values associated with a plurality of predicted parameters after a number of observed measurements. For example, the graph 739 A illustrates one or more C-level 2 std error zones and one or more T-level 2 std error zones after 0 measurements. The graph 739B illustrates one or more C-level 2 std error zones and one or more T-level 2 std error zones after 1 measurement. The graph 739C illustrates one or more C-level 2 std error zones and one or more T-level 2 std error zones after 2 measurements. The graph 739D illustrates one or more C-level 2 std error zones and one or more T-level 2 std error zones after 3 measurements. The graph 739E illustrates one or more C-level 2 std error zones and one or more T-level 2 std error zones after 4 measurements. The graph 739F illustrates one or more C-level 2 std error zones and one or more T-level 2 std error zones after 5 measurements. The graph 739G illustrates one or more C-level 2 std error zones and one or more T-level 2 std error zones after 6 measurements. The graph 739H illustrates one or more C-level 2 std error zones and one or more T-level 2 std error zones afterAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC17 measurements. As illustrated in the graphical depiction 700D, the spread of the error zones (e.g., C-level 2 std error zones and / or T-level 2 std error zones) continues to narrow as additional measurements are performed by a user (e.g., medical practitioner), thus indicating less uncertainty regarding the predicted operational parameters as more measurements are obtained. In certain embodiments, in addition to or instead of C-levels and / or T-levels, any predicted operational parameters can be visualized in each of the graphs 739A-H in accordance with the techniques presented herein.

[0131] Each of graphical depictions 700D-700K enables a user to identify one or more electrodes that have predicted operational parameters with high uncertainty (or low confidence). This way, the confidence levels and / or uncertainty values allow users (e.g., a medical practitioner) to quickly determine which of the plurality of predicted operational parameters is reliable and which parameters require further clinical observation, thus providing increased efficiency and accuracy in the configuration of recipient devices. In certain embodiments, the graphs 739A-H can be displayed, via a graphical user interface (GUI), to a user (e.g., a medical practitioner) before, during, or after a recipient device configuration session. In certain embodiments, the user can select a subset of the graphs 739A-H for display. The one or more C-level 2 std error zones and one or more T-level 2 std error zones corresponding to the predicted operational parameters in each of the graphs 739A-H provide visual emphasis on one or more electrodes based on uncertainty associated with the corresponding predicted operational parameters. Thus, the graphs 739A-H enable the user to quickly and accurately decide which electrode needs further clinical observation (measurement) at a current or future session. For example, the user can speed up the recipient device configuration session by only obtaining new observations (measurements) of operational parameters for electrodes with prediction uncertainty higher than a predetermined threshold while relying on predicted parameters for the other electrodes. This provides a more targeted configuration experience, thus saving time and resources for the recipient and the medical practitioner. Further, augmenting model-generated parameter predictions with human observation and expertise ensures the recipient device is properly configured. In certain embodiments, the graphical user interface can be configured to receive input from the user (e.g., medical practitioner), such as feedback regarding the predicted operational parameters and / or the error zones.

[0132] FIG. 8 is a flowchart illustrating a method 800, in accordance with certain embodiments presented herein. Method 800 begins at 801 where a plurality of past operational parametersAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1associated with configuring a recipient device for a recipient in one or more previous configuration sessions is obtained. At 803, based on the plurality of past operational parameters, a plurality of operational parameters for the recipient device is generated via a trained probabilistic regression network. Then, at 805, the recipient device is configured with the plurality of operational parameters.

[0133] FIG. 9 is a flowchart illustrating a method 900, in accordance with certain embodiments presented herein. Method 900 beings at 901 where a plurality of historical stimulation parameters associated with configuring one or more medical devices is obtained. At 903, based on the plurality of historical stimulation parameters, a plurality of stimulation parameters for a recipient medical device is generated via the probabilistic regression network. Then, at 905, the recipient medical device is configured with the plurality of stimulation parameters.

[0134] FIG. 10 is a flowchart illustrating a method 1000, in accordance with certain embodiments presented herein, where the operations are performed by a processor executing instructions stored in one or more non-transitory computer readable storage media. At 1001, a plurality of predicted operational parameters for a recipient device is predicted via a first machine learning model, at 1003, a plurality of operational parameters based on the plurality of predicted operational parameters is generated via a second machine learning model. Then, at 1003, the recipient device is configured based on the plurality of operational parameters.

[0135] FIG. 11 is a flowchart illustrating a method 1100, in accordance with certain embodiments presented herein, where the operations are performed by a system comprising a memory and at least one processor operable coupled to the memory. At 1101, a plurality of operational parameters for a device is generated via a machine learning model. At 1103, a plurality of uncertainty values associated with the plurality of operational parameters is determined. Then, at 1105, a graphical representation of the plurality of operational parameters is updated based on the plurality of uncertainty values. At 1107, one or more electrodes are selected from a plurality of electrodes associated with the device based on the graphical representation. Then, at 1109, the one or more electrodes of the device are configured based on the plurality of operational parameters.

[0136] As previously described, the technology disclosed herein can be applied in any of a variety of circumstances and with a variety of different devices. Example devices that can benefit from technology disclosed herein are described in more detail in FIGS. 12 and 13. The techniques of the present disclosure can be applied to other devices, such as neurostimulators,Atty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1cardiac pacemakers, cardiac defibrillators, sleep apnea management stimulators, seizure therapy stimulators, tinnitus management stimulators, and vestibular stimulation devices, as well as other medical devices that deliver stimulation to tissue. Further, technology described herein can also be applied to consumer devices. These different systems and devices can benefit from the technology described herein.

[0137] FIG. 12 illustrates an example vestibular stimulator system 1202, with which embodiments presented herein can be implemented. As shown, the vestibular stimulator system 1202 comprises an implantable component (vestibular stimulator) 1212 and an external device / component 1204 (e.g., external processing device, battery charger, remote control, etc.). The external device 1204 comprises a transceiver unit 1260. As such, the external device 1204 is configured to transfer data (and potentially power) to the vestibular stimulator 1212.

[0138] The vestibular stimulator 1212 comprises an implant body (main module) 1234, a lead region 1236, and a stimulating assembly 1216, all configured to be implanted under the skin / tissue (tissue) 1215 of the recipient. The implant body 1234 generally comprises a hermetically-sealed housing 1238 in which RF interface circuitry, one or more rechargeable batteries, one or more processors, and a stimulator unit are disposed. The implant body 1234 also includes an intemal / implantable coil 1214 that is generally external to the housing 1238, but which is connected to the transceiver via a hermetic feedthrough (not shown).

[0139] The stimulating assembly 1216 comprises a plurality of electrodes 1244(l)-(3) disposed in a carrier member (e.g., a flexible silicone body). In this specific example, the stimulating assembly 1216 comprises three (3) stimulation electrodes, referred to as stimulation electrodes 1244(1), 1244(2), and 1244(3). The stimulation electrodes 1244(1), 1244(2), and 1244(3) function as an electrical interface for delivery of electrical stimulation signals to the recipient’s vestibular system.

[0140] The stimulating assembly 1216 is configured such that a surgeon can implant the stimulating assembly adjacent the recipient’s otolith organs via, for example, the recipient’s oval window. It is to be appreciated that this specific embodiment with three stimulation electrodes is merely illustrative and that the techniques presented herein can be used with stimulating assemblies having different numbers of stimulation electrodes, stimulating assemblies having different lengths, etc.

[0141] One or more operational parameters of the vestibular stimulator 1212, the external device 1204, and / or another external device can be set / determined using the techniquesAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1presented herein. Thereafter, the vestibular stimulator 1212, the external device 1204, and / or another external device could be programmed / configured with the determined operational parameters (e.g., the determined operational parameters could be instantiated in the vestibular stimulator 1212, the external device 1204, and / or another external device).

[0142] FIG. 13 illustrates a retinal prosthesis system 1301 that comprises an external device 1310 (which can correspond to the wearable device 100) configured to communicate with an implantable retinal prosthesis 1300 via signals 1351. The retinal prosthesis 1300 comprises an implanted processing module 1325, and a retinal prosthesis sensor-stimulator 1390 is positioned proximate the retina of a recipient. The external device 1310 and the processing module 1325 can communicate via coils 1308, 1314.

[0143] In an example, sensory inputs (e.g., photons entering the eye) are absorbed by a microelectronic array of the sensor-stimulator 1390 that is hybridized to a glass piece 1392 including, for example, an embedded array of microwires. The glass can have a curved surface that conforms to the inner radius of the retina. The sensor-stimulator 1390 can include a microelectronic imaging device that can be made of thin silicon containing integrated circuitry that convert the incident photons to an electronic charge.

[0144] The processing module 1325 includes an image processor 1323 that is in signal communication with the sensor-stimulator 1390 via, for example, a lead 1388 that extends through surgical incision 1389 formed in the eye wall. In other examples, processing module 1325 is in wireless communication with the sensor-stimulator 1390. The image processor 1323 processes the input into the sensor-stimulator 1390 and provides control signals back to the sensor-stimulator 1390 so the device can provide an output to the optic nerve. That said, in an alternate example, the processing is executed by a component proximate to, or integrated with, the sensor-stimulator 1390. The electric charge resulting from the conversion of the incident photons is converted to a proportional amount of electronic current which is input to a nearby retinal cell layer. The cells fire and a signal is sent to the optic nerve, thus inducing a sight perception.

[0145] The processing module 1325 can be implanted in the recipient and function by communicating with the external device 1310, such as a BTE unit, a pair of eyeglasses, etc. The external device 1310 can include an external light / image capture device (e.g., located in / on a behind-the-ear device or a pair of glasses, etc.), while, as noted above, in some examples, theAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1sensor-stimulator 1390 captures light / images, in which sensor-stimulator 1390 is implanted in the recipient.

[0146] One or more operational parameters of the retinal prosthesis system 1301 can be set / determined using the techniques presented herein. Thereafter, the retinal prosthesis system 1301 could be programmed / configured with the determined operational parameters (e.g., the determined operational parameters could be instantiated in the retinal prosthesis system 1301).

[0147] It is to be appreciated that FIGs. 12 and 13 are merely illustrative of devices / systems that with which aspects of the techniques presented herein can be implemented. In other embodiments, the techniques presented herein can be implemented in / by / with sleep disorder devices (e.g., sleep apnea devices), seizure devices (e.g., devices for monitoring and / or treating epileptic events), other balance or movement disorder devices (e.g., vestibular stimulation devices), tinnitus management devices, other visual implants (e.g., bionic eyes) and other neuromodulation devices (e.g., brain-computer interfaces). For example, the techniques presented herein could be used to set one or more operational parameters of any of these or other devices. Therefore, specific reference to cochlear implants or other devices herein is merely illustrative

[0148] Turning to FIG. 14, depicted therein is a block diagram illustrating an example fitting system 1470 configured to perform aspects of the techniques presented herein (e.g., perform one or more of the operations of FIGs. 2, 3, 4, 5, 6, 7A-7K 8, 9, 10, and / or 11). Fitting system 1470 is, in general, a computing device that comprises a plurality of interfaces / ports 1478(1)-1478(N), a memory 1480, a processor 1484, and a user interface 1486. The interfaces 1478(1)-1478(N) can comprise, for example, any combination of network ports (e.g., Ethernet ports), wireless network interfaces, Universal Serial Bus (USB) ports, Institute of Electrical and Electronics Engineers (IEEE) 1394 interfaces, PS / 2 ports, etc. In the example of FIG. 14, interface 1478(1) is connected to cochlear implant system 102 having components implanted in a user 1471. Interface 1478(1) can be directly connected to the cochlear implant system 102 or connected to an external device that is communication with the cochlear implant systems (e.g., external device 110 of FIGs. 1A-1D). Interface 1478(1) can be configured to communicate with cochlear implant system 102 via a wired or wireless connection (e.g., telemetry, Bluetooth, efc.).

[0149] The user interface 1486 includes one or more output devices, such as a display screen (e.g., a liquid crystal display (LCD)) and a speaker, for presentation of visual or audibleAtty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1information to a clinician, audiologist, or other user. The user interface 1486 can also comprise one or more input devices that include, for example, a keypad, keyboard, mouse, touchscreen, etc.

[0150] The memory 1480 comprises latent variable fitting logic 1481 that can be executed to perform aspects of the techniques presented herein (e.g., executed to perform one or more of the operations of FIGs. 2, 3, 4, 5, 6, 7A-7K, 8, 9, 10, and / or 11). Memory 1480 can comprise read only memory (ROM), random access memory (RAM), magnetic disk storage media devices, optical storage media devices, flash memory devices, electrical, optical, or other physical / tangible memory storage devices. The processor 1484 is, for example, a microprocessor or microcontroller that executes instructions for the latent variable fitting logic 1481. Thus, in general, the memory 1480 can comprise one or more tangible (non-transitory) computer readable storage media (e.g., a memory device) encoded with software comprising computer executable instructions and when the software is executed (by the processor 1484) it is operable to perform the techniques described herein.

[0151] As should be appreciated, while particular uses of the technology have been illustrated and discussed above, the disclosed technology can be used with a variety of devices in accordance with many examples of the technology. The above discussion is not meant to suggest that the disclosed technology is only suitable for implementation within systems akin to that illustrated in the figures. In general, additional configurations can be used to practice the processes and systems herein and / or some aspects described can be excluded without departing from the processes and systems disclosed herein.

[0152] This disclosure described some aspects of the present technology with reference to the accompanying drawings, in which only some of the possible aspects were shown. Other aspects can, however, be embodied in many different forms and should not be construed as limited to the aspects set forth herein. Rather, these aspects were provided so that this disclosure was thorough and complete and fully conveyed the scope of the possible aspects to those skilled in the art.

[0153] As should be appreciated, the various aspects (e.g., portions, components, etc.) described with respect to the figures herein are not intended to limit the systems and processes to the particular aspects described. Accordingly, additional configurations can be used to practice the methods and systems herein and / or some aspects described can be excluded without departing from the methods and systems disclosed herein.Atty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1

[0154] According to certain aspects, systems and non-transitory computer readable storage media are provided. The systems are configured with hardware configured to execute operations analogous to the methods of the present disclosure. The one or more non-transitory computer readable storage media comprise instructions that, when executed by one or more processors, cause the one or more processors to execute operations analogous to the methods of the present disclosure.

[0155] Similarly, where steps of a process are disclosed, those steps are described for purposes of illustrating the present methods and systems and are not intended to limit the disclosure to a particular sequence of steps. For example, the steps can be performed in differing order, two or more steps can be performed concurrently, additional steps can be performed, and disclosed steps can be excluded without departing from the present disclosure. Further, the disclosed processes can be repeated.

[0156] Although specific aspects were described herein, the scope of the technology is not limited to those specific aspects. One skilled in the art will recognize other aspects or improvements that are within the scope of the present technology. Therefore, the specific structure, acts, or media are disclosed only as illustrative aspects. The scope of the technology is defined by the following claims and any equivalents therein.

[0157] It is also to be appreciated that the embodiments presented herein are not mutually exclusive and that the various embodiments can be combined with another in any of a number of different manners.

Claims

Atty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC1CLAIMSWhat is claimed is:

1. A method, comprising:obtaining a plurality of past operational parameters associated with configuring a recipient device for a recipient in one or more previous configuration sessions;generating, via a trained probabilistic regression network, a plurality of operational parameters for the recipient device based on the plurality of past operational parameters; and configuring the recipient device with the plurality of operational parameters.

2. The method of claim 1, wherein the trained probabilistic regression network is configured to generate a probability function for estimating a probability of a predicted operational parameter vector.

3. The method of claim 2, wherein the predicted operational parameter vector includes one or more predicted observed operational parameters or one or more predicted unobserved operational parameters.

4. The method of claim 2 or 3, further comprising:obtaining a plurality of historical operational parameters and a plurality of recipientspecific parameters from a parameter database; andtraining, based on the plurality of historical operational parameters and the plurality of recipient-specific parameters, a probabilistic regression network to generate the trained probabilistic regression network.

5. The method of claim 4, wherein the plurality of historical operational parameters is associated with a plurality of electrodes of one or more recipient devices, and wherein the plurality of historical operational parameters comprises one or more of: threshold levels associated with the plurality of electrodes, comfort levels associated with the plurality of electrodes, or frequency maps associated with the plurality of electrodes.

6. The method of claim 4, wherein the plurality of recipient-specific parameters comprises one or more of: sex, age, electrode impedance values, speech test outcomes, anatomical features, or one or more objective measures.Atty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC17. The method of claim 6, wherein the one or more objective measures comprises one or more of: a phoneme discrimination test, a speech test, or electrically evoked compound action potentials (eCAP) values.

8. The method of claim 2, 3, 5, 6, or 7, further comprising:obtaining one or more adjustments to the plurality of operational parameters from one or more experts, wherein the one or more adjustments are determined based on an evaluation of the probability function; andupdating the plurality of operational parameters based on the one or more adjustments.

9. The method of claim 8, wherein the one or more experts include a medical practitioner or the recipient.

10. The method of claim 4, wherein the plurality of historical operational parameters and the plurality of recipient-specific parameters are associated with a plurality of recipients from a recipient population.

11. The use of a device according to anyone of claims 1-10 in a cochlear implant, sleep disorder device, a seizure device, a balance or movement disorder device, a tinnitus management device, or a visual device.

12. A system according to claims 1-10, wherein the system is a cochlear implant system, a sleep disorder system, a seizure system, a balance or movement disorder system, a tinnitus management system, or a visual system.

13. A method, comprising:obtaining a plurality of historical stimulation parameters associated with configuring one or more medical devices;generating, via a probabilistic regression network, a plurality of stimulation parameters for a recipient medical device based on the plurality of historical stimulation parameters; andconfiguring the recipient medical device with the plurality of stimulation parameters.Atty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC114. The method of claim 13, wherein the probabilistic regression network is a trained probabilistic regression network.

15. The method of claim 14, wherein the plurality of stimulation parameters includes one or more predicted observed operational parameters or one or more predicted unobserved operational parameters.

16. The method of claim 15, further comprising:obtaining a plurality of recipient-specific parameters associated with one or more recipients of the one or more medical devices; andtraining, based on the plurality of historical stimulation parameters and the plurality of recipient-specific parameters, the probabilistic regression network to generate the trained probabilistic regression network.

17. The method of claim 16, wherein generating, via the probabilistic regression network, the plurality of stimulation parameters based on the plurality of historical stimulation parameters comprises:determining, via the probabilistic regression network, a probability function of a vector representing the plurality of stimulation parameters based on the plurality of historical stimulation parameters, the plurality of recipient-specific parameters, and a plurality of observed stimulation parameters for the recipient medical device; andupdating the plurality of stimulation parameters based on the probability function.

18. The method of claim 17, wherein updating the plurality of stimulation parameters based on the probability function comprises:obtaining one or more adjustments to the plurality of stimulation parameters from one or more experts, wherein the one or more adjustments are determined based on an evaluation of the probability function; andupdating the plurality of stimulation parameters based on the one or moreadjustments.Atty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC119. The method of claim 13, 14, 15, 16, 17, or 18, wherein the plurality of historical stimulation parameters includes one or more parameters associated with configuring the one or more medical devices or the recipient medical device at one or more previous configuration sessions.

20. The method of claim 19, wherein the plurality of historical stimulation parameters is associated with a plurality of electrodes of the one or more medical devices or the recipient medical device, and wherein the plurality of historical stimulation parameters comprises one or more of threshold levels associated with the plurality of electrodes, comfort levels associated with the plurality of electrodes, or frequency maps associated with the plurality of electrodes.

21. The method of claim 16, wherein the plurality of recipient-specific parameters comprises one or more of sex, age, electrode impedance values, speech test outcomes, anatomical features, or one or more objective measures.

22. The method of claim 21, wherein the one or more objective measures comprises one or more of a phoneme discrimination test, a speech test, or electrically evoked compound action potentials (eCAP) values.

23. The method of claim 13, 14, 15, 16, 17, or 18, wherein the recipient medical device is worn by or implanted in a recipient.

24. The use of a device according to anyone of claims 13-23 in a cochlear implant, sleep disorder device, a seizure device, a balance or movement disorder device, a tinnitus management device, or a visual device.

25. A system according to claims 13-23, wherein the system is a cochlear implant system, a sleep disorder system, a seizure system, a balance or movement disorder system, a tinnitus management system, or a visual system.Atty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC126. One or more non-transitory computer readable storage media comprising instructions that, when executed by a processor, cause the processor to:predict, via a first machine learning model, a plurality of predicted operational parameters for a recipient device;generate, via a second machine learning model, a plurality of operational parameters based on the plurality of predicted operational parameters; andconfigure the recipient device based on the plurality of operational parameters.

27. The one or more non-transitory computer readable storage media of claim 26, wherein the second machine learning model includes a probabilistic regression network.

28. The one or more non-transitory computer readable storage media of claim 27, wherein the probabilistic regression network uses a conditional multivariate gaussian process.

29. The one or more non-transitory computer readable storage media of claim 28, wherein the instructions are operable to cause the processor to generate, via the second machine learning model, the plurality of operational parameters based on the plurality of predicted operational parameters by:setting a first mean value of the plurality of predicted operational parameters as a mean value of the conditional multivariate gaussian process;determining a covariance value of the conditional multivariate gaussian process; and generating the plurality of operational parameters based on the mean value and the covariance value of the conditional multivariate gaussian process.

30. The one or more non-transitory computer readable storage media of claim 29, wherein determining the covariance value of the conditional multivariate gaussian process comprises:training the first machine learning model to generate a second plurality of predicted operational parameters;determining a difference between a plurality of ground truth operational parameters and the second plurality of predicted operational parameters; anddetermining the covariance value based on the difference.Atty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC131. The one or more non-transitory computer readable storage media of claim 30, wherein the plurality of ground truth operational parameters is obtained from a database storing data associated with a plurality of recipients, the data includes operational parameters for configuring one or more recipient devices associated with the plurality of recipients and recipient-specific parameters.

32. The one or more non-transitory computer readable storage media of claim 26, 27, 28, 29, 30, or 31, further comprising instructions that, when executed by a processor, cause the processor to:obtain one or more adjustments to the plurality of operational parameters from one or more experts; andupdate the plurality of operational parameters based on the one or more adjustments.

33. The one or more non-transitory computer readable storage media of claim 32, wherein the one or more experts include a medical practitioner or a recipient of the recipient device.

34. The one or more non-transitory computer readable storage media of claim 26, 27, 28, 29, 30, or 31, wherein the recipient device is worn by or implanted in a recipient.

35. The one or more non-transitory computer readable storage media of claim 31, wherein the data stored in the database are associated with one or more previous configuration sessions of one or more recipients from a recipient population.Atty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC136. A system, comprising:a memory; andat least one processor operable coupled to the memory, wherein the at least one processor is configured to:generate, via a machine learning model, a plurality of operational parameters for a device;determine a plurality of uncertainty values associated with the plurality of operational parameters;update a graphical representation of the plurality of operational parameters based on the plurality of uncertainty values;select one or more electrodes from a plurality of electrodes associated with the device based on the graphical representation; andconfigure the one or more electrodes of the device based on the plurality of operational parameters.

37. The system of claim 36, wherein the graphical representation is configured to display a respective operational parameter of the plurality of operational parameters corresponding to a respective electrode of the plurality of electrodes, and display a respective uncertainty value of the plurality of uncertainty values corresponding to the respective operational parameter and the respective electrode.

38. The system of claim 37, wherein the graphical representation is configured to visually emphasize a subset of the plurality of electrodes based on the plurality of uncertainty values.

39. The system of claim 38, wherein selecting the one or more electrodes from the plurality of electrodes associated with the device based on the graphical representation comprises:selecting the one or more electrodes from the subset of the plurality of electrodes that have been visually emphasized.

40. The system of claim 36, 37, or 38, wherein the plurality of uncertainty values is determined based on a difference between the plurality of operational parameters and a plurality of ground truth operational parameters.Atty. Docket No. 3065.0869i Client Ref. No. CID04036WOPC141. The system of claim 40, wherein the plurality of ground truth operational parameters is obtained from a database storing data associated with a plurality of recipients, the data includes operational parameters for configuring one or more medical devices associated with the plurality of recipients and recipient-specific parameters.

42. The system of claim 36, 37, or 38, wherein the machine learning model is a trained probabilistic regression network.

43. The system of claim 36, 37, or 38, wherein the plurality of uncertainty values and a plurality of mean values associated with the plurality of operational parameters are iteratively updated based on iterative adjustments to the plurality of operational parameters.

44. The system of claim 43, wherein the iterative adjustments are received from one or more experts.

45. The system of claim 44, wherein the graphical representation of the plurality of operational parameters is displayed, via a graphical user interface, to the one or more experts.

Images