WO2025210477A1 - High dimensionality outlier detection and identification - Google Patents

Abstract

Presented herein are techniques for determining whether a healthcare information dataset includes incongruous information. A healthcare information dataset associated with a patient is obtained. The healthcare information dataset is analyzed using a data-derived model to determine whether the healthcare information dataset includes incongruities. One or more actions are performed based on whether the healthcare information dataset includes incongruities.

Description

HIGH DIMENSIONAUITY OUTUIER DETECTION AND IDENTIFICATION

BACKGROUND

Field of the Invention

[oooi] The present invention relates generally to detecting and identifying outliers in datasets.

Related Art

[0002] Medical devices have provided a wide range of therapeutic benefits to recipients over recent decades. Medical devices can include internal or implantable components/devices, external or wearable components/devices, or combinations thereof (e.g., a device having an external component communicating with an implantable component). Medical devices, such as traditional hearing aids, partially or fully-implantable hearing prostheses (e.g., bone conduction devices, mechanical stimulators, cochlear implants, etc.), pacemakers, defibrillators, functional electrical stimulation devices, and other medical devices have been successful in performing lifesaving and/or lifestyle enhancement functions and/or recipient monitoring for a number of years.

[0003] The types of medical devices and the ranges of functions performed thereby have increased over the years. For example, many medical devices, sometimes referred to as “implantable medical devices,” now often include one or more instruments, apparatus, sensors, processors, controllers or other functional mechanical or electrical components that are permanently or temporarily implanted in a recipient. These functional devices are typically used to diagnose, prevent, monitor, treat, or manage a disease/injury or symptom thereof, or to investigate, replace or modify the anatomy or a physiological process. Many of these functional devices utilize power and/or data received from external devices that are part of, or operate in conjunction with, implantable components.

SUMMARY

[0004] In one aspect, a method is provided. The method comprises: receiving medical data associated with a patient; analyzing the medical data using a data-derived model to determine whether the medical data includes incongruities; and performing one or more actions based on whether the medical data includes incongruities.

[0005] In another aspect, another method is provided. The method comprises: receiving a healthcare information dataset; and training a model using the healthcare information dataset to identify incongruities in subsequently received healthcare information datasets. [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 include instructions that, when executed by a processor, cause the processor to: obtain data associated with a user; model the data using a trained model to produce modeled data; and analyze the modeled data using the trained model to determine whether the data includes incongruous data.

[0007] In another aspect, a system is provided. The system comprises: a memory; at least one processor operable coupled to the memory, wherein the at least one processor is configured to: receive a healthcare information dataset associated with a patient; model a structure of the healthcare information dataset using a model to produce a modeled healthcare information dataset; and validate the modeled healthcare information dataset,

BRIEF DESCRIPTION OF THE DRAWINGS

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

[0009] FIG. 1 is a flow diagram illustrating an example process of identifying incongruities in patient data or health information, according to embodiments described herein;

[0010] FIG. 2 is a flow diagram illustrating a method of performing an incongruous impact assessment, according to embodiments described herein;

[0011] FIG. 3 is a flow diagram illustrating an example method of training a Principal Component Analysis (PCA) model to identify incongruous data, according to embodiments described herein;

[0012] FIG. 4 is a flow diagram of a method for calibrating a PCA model, according to embodiments described herein;

[0013] FIG. 5 is a flow diagram illustrating a method of determining whether medical data includes incongruities, according to embodiments described herein;

[0014] FIG. 6 is a flow diagram illustrating a method of training a model to identify incongruities in healthcare information datasets, according to embodiments described herein;

[0015] FIG. 7 is a functional block diagram of a computing system configured to implement aspects of the techniques presented herein; [0016] FIG. 8 is a schematic diagram illustrating a vestibular stimulator system with which aspects of the techniques presented herein can be implemented;

[0017] FIG. 9 is a schematic diagram illustrating a retinal prosthesis system with which aspects of the techniques presented herein can be implemented; and

[0018] FIG. 10A is a schematic diagram illustrating a cochlear implant system with which aspects of the techniques presented herein can be implemented;

[0019] FIG. 10B is a side view of a recipient wearing a sound processing unit of the cochlear implant system of FIG. 10A;

[0020] FIG. 10C is a schematic view of components of the cochlear implant system of FIG. 10A; and

[0021] FIG. 10D is a block diagram of the cochlear implant system of FIG. 10A.

DETAILED DESCRIPTION

[0022] Advanced or complex algorithms developed use artificial intelligence (Al) techniques, such as machine learning (ML) techniques, have their internal structures defined by patterns in the data. Artificial intelligence techniques are valuable in the provision of medical diagnostics and clinical care. For these algorithms to produce safe and efficacious outputs, the data being input into the algorithms, for both the development of the algorithm and clinical use, needs to be of high quality (e.g., accurate and specific to the patient undergoing medical treatment). However, in certain conventional arrangements, the data input in to the algorithms may not be of high quality. Possible causes for the data to not be of high quality include, for example, incomplete medical records, manual data entry errors, misattribution of medical details from one patient to another, anomalies in measurements being made by medical diagnostic devices, incompatibilities between information technology (IT) systems, and inaccuracies of algorithms that process patient health data.

[0023] Some types of poor-quality data can be detected using simple heuristics, such as observing that data is missing, values are beyond certain ranges, or data types are incorrect. Other types of poor-quality data are much more difficult to detect, such as incongruities between different data types and their values. For example, in one scenario, a patient with a poor prognosis can have been issued an active medical device, the medical device’s data logging can have indicated that the medical device was not used, but the outcome data indicated a significant improvement. In this scenario, without proper verification, the incongruity cannot be assumed to be valid. If this data were to be included during the development of an advanced algorithm that relies on machine learning, it can influence the machine learning algorithm to recognize incorrect patterns. In another scenario, if this data were to be input into a clinical decision support tool to aid medical diagnosis or clinical care, the information outputted may not be relevant to the specific patient. The phenomenon described in the example is pervasive and can be extended to many additional possible health factors and measurements.

[0024] As noted, artificial intelligence techniques can be used to analyze patient health or medical data, and the strength of the analysis (e.g., ability correctly and effectively analyze the data) depends, at least on part, on the accuracy/quality of the input data. For example, if data with incorrect details or details that are not related to the correct patient are input to an artificial intelligence system for analysis, the output of the analysis can be inaccurate or erroneous. In addition, patient data tends to have a high dimensionality or complexity and can include many variables. For example, a patient record can include hundreds of patient variables, such as gender, height, weight, genetics, medical conditions, etc. Due to the high dimensionality of patient records, it may not be possible to identify errors or outliers in a medical record merely by inspecting the record (e.g., by a clinician or doctor reviewing the data) or using other conventional techniques. As such, there is a need to ensure that the data, such as patient health or medical data, that is to be analyzed by an artificial intelligence system (e.g., a computing device implementing artificial intelligence techniques for data analysis) is of sufficiently high quality (e.g., accurate and specific to the patient undergoing medical treatment) such that the output of the artificial intelligence system safe and efficacious.

[0025] Therefore, in order to ameliorate the above inadequacies with conventional techniques, presented herein are techniques for determining whether a healthcare information dataset (e.g., a dataset of health data, including patient health, medical data, etc.) includes incongruous information. In operation, a healthcare information dataset associated with a patient is obtained and analyzed using a data-derived model to determine whether the healthcare information dataset (healthcare data) includes incongruities. One or more actions can be performed based on whether or not the healthcare information dataset includes incongruities to generate “clean” or “validated” healthcare information dataset and that can be confidently used in a subsequent process (e.g., a device fitting process, as described below). As used herein, a “validated” healthcare information dataset that has been configured to not include incongruities and/or a modified data set from which incongruities have been remediated (e.g., removed, corrected, etc.). [0026] In certain embodiments, the techniques presented herein provide for modeling patient data/health information (sometimes referred to herein as a “healthcare information dataset”) and analyzing the modeled healthcare information dataset to automatically identify incongruities or outliers in the patient data or health information. If incongruities are identified, those incongruities can, in turn, be provided to a user (e.g., a healthcare professional) for mitigation, remedial activities, or determination that the incongruities are trivial or nonexistent. In some embodiments, the incongruities can be analyzed by a model to, for example, assess the impact of the incongruities on the patient.

[0027] FIG. 1 is a flow diagram illustrating an example process of identifying incongruities in a healthcare information dataset, in accordance with certain embodiments presented herein. The flow of FIG. 1 begins at 102 where a healthcare information dataset that is to be analyzed by an artificial intelligence system is obtained (e.g., retrieved). The healthcare information dataset can include a variety of different types of “health data” (e.g., data that is relevant to the health of one or more patients) that is obtained from a number of different sources. For example, the healthcare information dataset can include measurement data, imaging (scan) data, data obtained via interaction with a patient, etc.

[0028] At 104, the healthcare information dataset is pre-processed to, for example, place the data in a form (e.g., a numerical form) that can be input to a device for analysis. That is, as noted above, a healthcare information dataset can include different types of data, such as, for example, clinician notes and observations, health treatment codes, demographics, appointment schedules, diagnoses, prognoses, diagnostic measurements, results of medical tests, medical device measurements and logs, medical device settings and parameters, location data, clinic preferences in treatment regimes, changes in clinics treating the patients, seasonal or environmental factors (e.g., temperature, humidity, pressure, air pollution index, etc.), life events and mental health trackers/questionnaires, results of scans (e.g., computed tomology (CT) scans, magnetic resonance imaging (MRI) scans, x-rays, etc.), etc. These different types of health data can be converted to a form that can be processed by mathematical, numerical, or computational techniques. The different types of data are pre-processed so that the data can be input to a device or system (e.g., a ML model) for processing or analysis.

[0029] The health data in the healthcare information dataset can be pre-processed in a number of different ways. In one example, the text of clinician notes and observations can be tokenized and the tokenized text can be processed via a natural language processing model to retrieve information such as subject matter and terminologies, sentiment, lexical semantics, etc. In another example, text-based codes and labels can be converted into numeric forms, such as coding male patients as ‘0’ and coding female patients as ‘ 1’ for gender/sex fields, the coding of aetiologies into dummy variables, etc. In another example, date/time fields can be decoded into constituent components such as “year, month, day,” “hour, minute, second,” etc. It is to be appreciated these specific data conversion/pre-processing techniques are merely illustrative and that other techniques could be used in various embodiments presented herein.

[0030] At 106, the pre-processed data in the healthcare information dataset (pre-processed healthcare information dataset or pre-processed healthcare data) is modeled based on or using a model that represents the structure and internal relationships of the types of health data that has been input. Again, as noted above, the underlying health data can be of a variety of different formats or sources that need to be transformed to a uniform datatype interpretable by an algorithm, such as an incongruous data detection algorithm. To identify outliers and high dimensionality outliers, the techniques presented herein use a representation or model of the health data. The representation or model captures patterns, sequences, covariates, nonlinear covariates, autocorrelations, cross correlations, seasonalities, and distributions that would be identifiable in health data of a broad range of the applicable patient population. As described further below in FIG. 2, the model is constructed by inputting a large amount of processed health information into machine learning algorithms that model the structure of the data.

[0031] Different model types can be used to model the health data. For example, the health data can be modeled using a data-derived model, which can include a supervised statistical model or a machine learning model. The data-derived model could be, for example, a Principal Component Analysis (PCA) model, a neural network, or another type of model. For example, the data-derived model may include a large language model (LLM) or a foundational model trained on vast amounts of data such that the structure of the model may encompass general and specific knowledge. The trained LLM or foundational model may be further refined using health data or healthcare data to capture specific structures, patterns, and covariates of this type of data. The health data can be modeled in a different manner based on, for example, the type of model used to model the data and the type of medical data that is being modeled (e.g., cochlear implant patient data, oncology patient data, etc.). In general, the health data is modeled to produce a representation of the dataset that can be analyzed by an algorithm, as described below. In certain examples, it is possible to conduct transfer learning on an LLM to enable it to learn the structure of health data and identify incongruous information (e.g., use an LLM retrained on health data in place of a PCA model). [0032] At 108, the modeled data is analyzed with an algorithm to detect incongruities in the health data. More specifically, the algorithm can be an incongruous data detection algorithm that is initiated or called by a user of a software application, coded logic of a software application or another algorithm, herein known as the parent process. The algorithm analyzes the modeled data by running pattern matching operations over the processed data based on the model. The pattern matching operations return metrics on how similar or dissimilar the input data is from the model’s representation of the health data. The metrics are used to form the basis of the classification of the detection of incongruities.

[0033] As described above, several different types of models can be used to model the health data. In one example, principal component analysis (PCA) can be used to model the structure of the health data/healthcare information dataset. PCA identifies patterns (i.e., principal components) that exist within the dataset that are linearly independent of one another. Each principal component accounts for a degree of a variation of the dataset, and each element in a principal component vector corresponds to an individual variable of the dataset. The set of principle components can be multiplied by weights and summated to reconstruct observations.

[0034] An individual observation can be transformed by the PCA technique to derive a weight per principal component. Deriving the weight per component for all observations of the dataset produces a distribution of weights per principal component. The weight distributions can then be used to discern how similar or dissimilar an observation is from the rest of the dataset for all variables of the dataset. For example, if a weight for a principal component (which accounts for a large amount of variance of the dataset) is at either extremity of the weight distribution, it can be inferred that the observation differs from the rest of the dataset. Alternatively, if the weight is towards the median of the distribution (i.e., the area of highest probability density), it can be inferred that the observation is more similar to the rest of the dataset.

[0035] Incongruous data can be identified using the PCA model by using the formula P^incongruos ) = f(pdf₁(weight₁),pdf₂(weight₂), ... , pdf_c(weight_c)). In this formula, the probability of an observation being incongruous (P^incongruos )) is a function of probability densities (pdf_i(weight_i')), in which pdf_tQ is a probability density function of the 1^th principal component distribution of weights, and weighty is a weight of the PCA transform of the input observation corresponding to the 1^th principal component.

[0036] In this example, when in operation the algorithm is called, health data is input and pre- processed, the health data is PCA transformed, and a probability of the data being incongruous is returned using the above equation. For each weight input into the equation, the weight is altered and the corresponding change in P (incongruous) is recorded. For each subset of weights input into the equation, the weights are altered and the corresponding change in P (incongruous) is recorded. The alteration of a weight or subset of weights that most decrease the P (incongruous) is identified and the corresponding principal components are analyzed to identify variables that most contribute to the component. The candidate variable or variables that are incongruous are returned.

[0037] The above example illustrates identifying candidate incongruous data using a PCA model. Different models can identify candidate incongruous data using different methods.

[0038] After the candidate incongruous data is returned, at 110, an impact assessment is performed. With information input about the nature of the parent processing calling the incongruous data detection algorithm, the algorithm can assess the significance impact of the incongruous data through presenting plausible alternative values and measuring the degree of change of the output of the process. In other words, the impact assessment can determine possible outcomes of a misdiagnosis based on the incongruities. For example, if an extremely high weight is erroneously entered for a patient, a misdiagnosis can be that the patient has type 2 diabetes and an outcome can be that the patient is prescribed insulin, which could negatively impact the health of a patient that does not have type 2 diabetes. The impact assessment identifies how large of an impact the incongruous data can have on the patient (e.g., the patient’s diagnoses, patient’s prognoses, possible future steps for the patient, etc.).

[0039] For each software application or other algorithm that calls the incongruous data detection algorithm, there is a representative model or cost function. The model or cost function has a candidate variable or variables input. The candidate variable or variables and the model are assessed according to a sensitivity analysis. The variable or variables are altered and the corresponding change in the model is recorded. If the change is large or non-normative, it is assumed that the variable or variables have high impact. The cost function directly indicates the impact a variable has or variables have. Variables of medium to high impact are returned. Medium to high impact variables are variables that significantly affect the clinical treatment decisions that take place.

[0040] FIG. 2 is a flow diagram illustrating a method of performing an incongruous impact assessment described in 110. At 202, a set of incongruous variables comprising at least one incongruous variable are obtained (i.e., after the detection of incongruities described above in 108). At 204, a patient’s health (or medical) data is retrieved from, for example, an electronic medical record system and/or clinical decision support tool. At 206, the patient’s health/medical data is pre-processed to convert the data into a form that can be input into and analyzed by algorithms.

[0041] At 208, the incongruous variables of the patient’s health are set to null. At 210, a medical data model is used to impute the values of the incongruous variables to establish a counterfactual scenario (known as counterfactual data). At 212, a clinical decision support algorithm is run using the counterfactual patient health data and an output is recorded. At 214, the clinical decision support algorithm is run on unmodified patient health data and the output is recorded.

[0042] At 216, the patient’s prognoses under each scenario are assessed for a degree of impact on the patient. The counterfactual clinical decision algorithm output is assessed against original unmodified data and the unmodified data clinical decision algorithm output is assessed against counterfactual data. At 218, the prognoses for each assessment are the corresponding medical impacts for the patient are reported to a clinician or other user.

[0043] Returning to FIG. 1, at 112, the detected incongruities and impact assessment of the incongruities are raised to the parent process. In some embodiments, the detected incongruities can be outputted to a user (e.g., a healthcare professional) without performing an incongruous impact assessment. As trusted users (i.e., experienced professionals with domain knowledge) use the incongruous data detection algorithm, data that is classified as incongruous are raised to users to investigate whether the data is valid. The trusted users can review the data with the corresponding impact assessment. The review can be in terms of ensuring that there are no data entry errors, tracking the lineage of the data, following up with patients or clinical professionals, etc. The output of the review can be a flag indicating that the data is either invalid or valid. In some cases, the output of the review can be an indication that the incongruities are trivial or do not affect the patient’s treatment. These flags, along with the data, are in turn used to further refine a calibration dataset such that the algorithm can be retrained to improve performance, as described below in FIG. 4.

[0044] The flow of FIG. 1 ends at 114 where a “clean” or “validated” healthcare information dataset is generated. As noted above, a “clean” or “validated” healthcare information dataset is a dataset that has been confirmed to not include incongruities and/or a modified data set from which incongruities have been remediated (e.g., removed, corrected, etc.). The validated healthcare information dataset can be generated in a number of different manners, including via manual or automated processes. For example, in one embodiment, a healthcare data model, such as a PCA model or LLM, may be called to estimate a correct or more accurate value for the incongruous data to generate the validated healthcare information dataset.

[0045] Reference is now made to FIG. 3. FIG. 3 is a flow diagram illustrating an example method of training a PCA model to identify incongruous data. Although FIG. 3 illustrates the training of a PCA model, other types of models (e.g., LLMs, foundational models, neural networks, etc.) can be trained using patient data to detect incongruities in the patient data.

[0046] At 302, a healthcare information dataset is input. The healthcare information dataset can include health information from many different patients (e.g., thousands of patients). At 304, the data is transformed to ensure that it can be computationally processed. At 306, the PCA model is trained. The model can be trained for analyzing a particular type of medical data. For example, the model can be trained using data from patients with cochlear implants to identify incongruities in medical data associated with cochlear implants. As another example, the model can be trained for use at an oncology clinic with data from oncology patients. In this example, the oncology clinic module can be used for identifying incongruities in medical data associated with oncology patients. With any type of model, the model can be trained using the medical data from a large number of patients (e.g., thousands of patients). The model can be trained using, for example, a machine learning analysis. The trained model can be able to analyze new patient data to identify incongruities or outliers in the new patient data based on the medical data from the patients analyzed during the training.

[0047] At 308, when the model is a PCA model, the distributions of weights for each principal component are derived by transforming the dataset using the trained PCA model. At 310, the trained PCA model and weight distributions are returned. The trained PCA model can be used to analyze the new patient data to detect incongruities in the new patient data. In addition, the trained PCA model can continue to be trained using new patient data.

[0048] Reference is now made to FIG. 4. FIG. 4 is a flow diagram of a method for calibrating a PCA model. When using principal component analysis for detecting incongruities in medical data, some principal components can be irrelevant to the detection of the incongruities. During the calibration of the PCA model, principal components are analyzed to determine whether the principal components should be kept in the PCA model or removed. Irrelevant principal components can be removed from the PCA model so the irrelevant principal components do not interfere with the detection of the incongruities in the medical data. Although FIG. 4 illustrates the calibration of a PCA model, other types of models can be calibrated. For example, if the model is a neural network, calibration can be performed to prune back the network to eliminate unnecessary parts of the network so they do not interfere with the detection of incongruities.

[0049] At 402, the previously trained PCA model is loaded and, at 404, calibration data is loaded. The calibration data comprises known congruous and known incongruous data, and, at 406, data of both types are labelled or coded (e.g., as congruous or incongruous). The data is transformed by the PCA model to yield a weight per principal component for each observation of the calibration data. At 408, a principal component weight distribution to evaluate is selected. At 410, the weights of each principal component are evaluated on a univariate basis by training a classification model (e.g., logistics regression, decision tree, Naive Bayes, etc.). The trained model is evaluated by using accuracy metrics such as sensitivity, specificity, negative predictive value, and positive predictive value. At 412, each principal component that passes the evaluation is recorded.

[0050] At 414, subsets of the principal component weight distributions are selected to evaluate . At 416, subsets of the record of principal components are used to train classification models and the classification models are evaluated using a cross-fold technique. At 418, a defined set of useful principal components is outputted. At 420, the subset of principal components that are most efficacious at distinguishing congruous and incongruous observations are returned. In essence, the calibration technique analyzes each principal component against each congruous subset and each incongruous subset and, if the analysis causes a misclassification, the principal component is removed.

[0051] The trained (and calibrated) model can be used in a number of different ways to detect incongruities in patient healthcare data. In one example, the model can be integrated into clinical software. Clinical software generally requires that users (i.e., clinicians) enter patient health information into a device running the clinical software. To determine whether there is incongruous data in the patient health information, the incongruous data detection algorithm is called after a set of information has been input. The algorithm returns candidate variables that could be incongruous and be of high impact. These variables are raised to the user’s attention to evaluate whether the data has been entered correctly, should be corrected, or should be tagged as being tenuous. [0052] Electronic medical records store patient health information. Clinicians rely on this information to make treatment decisions. In some embodiments, the electronic medical records can be processed continuously by the incongruous data detection algorithm. Variables of certain patient records can be tagged or flagged for further assessment by clinicians or support personnel. Clinicians or support personnel can correct the medical record or take note that the record should be updated at a future appointment. Any edits to the data are logged to ensure that there is a record of the change and what the value was before the change.

[0053] Advanced algorithms access medical records or patient health information via Application Programming Interfaces (APIs). In some embodiments, the incongruous data detection algorithm can be directly integrated into APIs. Integration of the algorithm into the APIs provides the APIs with the ability to evaluate the quality of the data prior to the transmission of data to the advanced algorithm. In one embodiment, when the incongruous data detection algorithm is integrated into an API, the assessment of incongruity and degree of impact can be transmitted with the data. In another embodiment, incongruous data of high impact can be prevented from being transmitted. Flagging data as incongruous or preventing incongruous data of high impact from being transmitted can help to ensure that advanced algorithms do not suggest or make clinical decisions that are not applicable or can cause detrimental outcomes.

[0054] In some embodiments, the incongruous data detection algorithm can be integrated into clinical software to act as clinical decision support. For example, prior to undergoing medical treatment there is a certain expectation of the prognosis of the medical treatment in terms of efficacy and likelihood of occurrence. During the after-care phase, outcome data is collected on patients in the form of tests and clinical observations. As outcome data is being entered into the clinical software or electronic medical record (EMR), the algorithm can be used to detect anomalies. For example, if a patient has a high chance of achieving a good prognosis during pre-treatment phase and the patient is achieving a poor outcome post-treatment, the algorithm can detect an anomaly or incongruity. This incongruity can be raised to the clinician or clinic for intervention.

[0055] Reference is now made to FIG. 5. FIG. 5 is a flow diagram illustrating a method 500 of determining whether medical data includes incongruities. At 502, medical data associated with a patient is received. At 504, the medical data is analyzed using a data-derived model to determine whether the medical data includes incongruities. For example, the medical data can be analyzed using an algorithm associated with a ML model to determine whether the medical data includes incongruities that can be indicative of erroneous data (e.g., an incorrect data entered for a patient, data entered for the wrong patient, etc.).

[0056] At 506, one or more actions are performed based on whether the medical data includes incongruities. For example, the incongruities can be outputted to a user (e.g., a healthcare professional) for further analysis or remediation. As another example, an incongruity impact assessment can be performed on the incongruous data to determine possible impacts of the incongruous data on future treatments and/or prognoses associated with the patient. The incongruity impact assessment can be outputted to a healthcare professional for additional analysis.

[0057] Reference is now made to FIG. 6. FIG. 6 is a flow diagram of a method 600 of training a model to identify incongruities in health information data. At 602, a healthcare information dataset is received. The healthcare information dataset can include health information data for many (e.g., thousands) of patients. The healthcare information dataset can be associated with a particular type of health information (e.g., health information associated with cochlear implant patients, health information associated with oncology patients, etc.).

[0058] At 604, a model is trained using the healthcare information dataset to identify incongruities in subsequently received healthcare information datasets. For example, a healthcare information dataset associated with a patient can be received and the trained model can be used to identify whether the healthcare information dataset includes incongruities.

[0059] FIG. 7 is a block diagram of one example arrangement for an external computing device 710 configured to perform one or more operations in accordance with certain embodiments presented herein. As shown in FIG. 7, in its most basic configuration, the computing device 710 includes at least one processing unit 783 and a memory 784. The processing unit 783 includes one or more hardware or software processors (e.g., Central Processing Units) that can obtain and execute instructions. The processing unit 783 can communicate with and control the performance of other components of the external computing device 710. The memory 784 is one or more software or hardware-based computer-readable storage media operable to store information accessible by the processing unit 783. The memory 784 can store, among other things, instructions executable by the processing unit 783 to implement applications or cause performance of operations described herein, as well as other data. The memory 784 can be volatile memory (e.g., RAM), non-volatile memory (e.g., ROM), or combinations thereof. The memory 784 can include transitory memory or non-transitory memory. The memory 784 can also include one or more removable or non-removable storage devices. In examples, the memory 784 can include RAM, ROM) EEPROM (Electronically-Erasable Programmable Read-Only Memory), flash memory, optical disc storage, magnetic storage, solid state storage, or any other memory media usable to store information for later access. By way of example, and not limitation, the memory 784 can include wired media, such as a wired network or direct- wired connection, and wireless media, such as acoustic, RF, infrared, other wireless media, or combinations thereof. In certain embodiments, the memory 784 comprises logic 795 that, when executed, enables the processing unit 1083 to perform aspects of the techniques presented (e.g., the operations of FIGs. 1, 2, 3, 4, 5 or 6).

[0060] In the illustrated example of FIG. 7, the external computing device 710 further includes a network adapter 786, one or more input devices 787, and one or more output devices 788. The external computing device 710 can include other components, such as a system bus, component interfaces, a graphics system, a power source (e.g., a battery), among other components. The network adapter 786 is a component of the external computing device 710 that provides network access (e.g., access to at least one network 789). The network adapter

786 can provide wired or wireless network access and can support one or more of a variety of communication technologies and protocols, such as Ethernet, cellular, Bluetooth, near-field communication, and RF, among others. The network adapter 786 can include one or more antennas and associated components configured for wireless communication according to one or more wireless communication technologies and protocols. The one or more input devices

787 are devices over which the external computing device 710 receives input from a user. The one or more input devices 787 can include physically-actuatable user-interface elements (e.g., buttons, switches, or dials), a keypad, keyboard, mouse, touchscreen, and voice input devices, among other input devices that can accept user input. The one or more output devices 788 are devices by which the external computing device 710 is able to provide output to a user. The output devices 788 can include a display 790 (e.g., a liquid crystal display (LCD)) and one or more speakers 791, among other output devices for presentation of visual or audible information to the recipient, a clinician, an audiologist, or other user.

[0061] It is to be appreciated that the arrangement for the external computing device 710 shown in FIG. 7 is merely illustrative and that aspects of the techniques presented herein can be implemented at a number of different types of systems/devices including any combination of hardware, software, and/or firmware configured to perform the functions described herein. For example, the external computing device 710 can be a personal computer (e.g., a desktop or laptop computer), a hand-held device (e.g., a tablet computer), a mobile device (e.g., a smartphone), a surgical system, and/or any other electronic device having the capabilities to perform the associated operations described elsewhere herein.

[0062] In certain embodiments, the techniques presented herein can be leveraged for use in customizing or fitting programmable hearing devices and/or medical devices (implantable or not implantable) for a particular recipient (e.g., hearing aids, auditory prostheses, sensory substitution devices that use tactile stimulation, external or implantable insulin pumps, etc). More specifically, hearing devices and/or medical devices generally operate in accordance with a plurality of settings/parameters that can vary for different recipients. As such, when a recipient initially receives a hearing device and/or medical device, and often at various times thereafter, a so-called “fitting” process/procedure is performed to determine the appropriate settings of the device, given the particular recipient. There is an increasing trend to use artificial intelligence systems to perform or at least aid these fitting processes. The techniques presented herein can be incorporated with these so-called “artificial intelligence fitting systems” by validating the data that will be used by the artificial intelligence fitting system.

[0063] For example, in accordance with certain embodiments presented herein, a healthcare information dataset (health data) is data that is clinically relevant to fitting of a hearing device or medical device to a recipient. In example embodiments, this healthcare information dataset (which is clinically relevant to fitting of a hearing device or medical device to a recipient) is analyzed using the techniques presented herein. If any incongruities are identified, those incongruities can be remediated (e.g., automatically by an associated algorithm) to generate a “clean” or “validated” healthcare information dataset. The validated healthcare information dataset can, in turn, be used in a subsequent fitting process (e.g., analyzed by an artificial intelligence fitting systems) for determining one or more settings of a hearing device and/or medical device. The one or more determined settings are then installed/instantiated in the hearing device and/or medical device.

[0064] In some embodiments, the techniques presented herein may be used to validate an output of artificial intelligence algorithms applied to cochlear implantation. For example, given a patient's health information and fitting settings, the techniques described herein may be used to indicate whether the fitting setting are incongruous. In this way, the techniques described herein may enhance artificial intelligence algorithms operating in in the field of cochlear implants. Techniques described herein may additionally be used to verify fitting settings established manually by a clinician. [0065] There are a number of different types of devices in/with which embodiments of the present invention can be used to validate the corresponding healthcare information dataset. For example, the techniques can be incorporated into a process to fit/customize various programmable hearing devices, programmable medical devices (implantable or nonimplantable), etc. to/for a recipient. As used herein, the term “hearing device” is to be broadly construed as any device that acts on an acoustical 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 device systems, combinations or variations thereof, etc.), a device for use by a person with normal hearing (e.g., consumer devices that provide audio streaming, consumer headphones, earphones, and other listening devices), a hearing protection device, etc. In other examples, the techniques presented herein can be implemented by, or used in conjunction with, various implantable or non-implantable medical devices, such as visual devices (i.e., bionic eyes), sensors, pacemakers, drug delivery systems, defibrillators, functional electrical stimulation devices, catheters, seizure devices (e.g., devices for monitoring and/or treating epileptic events), sleep apnea devices, electroporation devices, insulin pumps, etc.

[0066] Example devices that can benefit from technology disclosed herein are described in more detail in FIGS. 8, 9, and 10A-10D. More specifically, FIG. 8 illustrates an example vestibular stimulator system 802 having one or more settings/parameters that can be determined based on validated healthcare information dataset that is validated using the techniques presented herein. As shown, the vestibular stimulator system 802 comprises an implantable component (vestibular stimulator) 812 and an external device/component 804 (e.g., external processing device, battery charger, remote control, etc.). The external device 804 comprises a transceiver unit 860. As such, the external device 804 is configured to transfer data (and potentially power) to the vestibular stimulator 812.

[0067] The vestibular stimulator 812 comprises an implant body (main module) 834, a lead region 836, and a stimulating assembly 816, all configured to be implanted under the skin/tissue (tissue) 815 of the recipient. The implant body 834 generally comprises a hermetically-sealed housing 838 in which RF interface circuitry, one or more rechargeable batteries, one or more processors, and a stimulator unit are disposed. The implant body 834 also includes an intemal/implantable coil 814 that is generally external to the housing 838, but which is connected to the transceiver via a hermetic feedthrough (not shown).

[0068] The stimulating assembly 816 comprises a plurality of electrodes 844(l)-(3) disposed in a carrier member (e.g., a flexible silicone body). In this specific example, the stimulating assembly 816 comprises three (3) stimulation electrodes, referred to as stimulation electrodes 844(1), 844(2), and 844(3). The stimulation electrodes 844(1), 844(2), and 844(3) function as an electrical interface for delivery of electrical stimulation signals to the recipient’s vestibular system.

[0069] The stimulating assembly 816 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.

[0070] In operation, the vestibular stimulator 812, the external device 804, and/or another external device can be configured to implement the techniques presented herein. That is, the vestibular stimulator 812, possibly in combination with the external device 804 and/or another external device, can include an evoked biological response analysis system, as described elsewhere herein.

[0071] FIG. 9 illustrates a retinal prosthesis system 901 having one or more settings/parameters that can be determined based on validated healthcare information dataset that is validated using the techniques presented herein. The retinal prosthesis system 901 comprises an external device 910 configured to communicate with an implantable retinal prosthesis 900 via signals 951. The retinal prosthesis 900 comprises an implanted processing module 925, and a retinal prosthesis sensor-stimulator 990 is positioned proximate the retina of a recipient. The external device 910 and the processing module 925 can communicate via coils 908, 914.

[0072] In an example, sensory inputs (e.g., photons entering the eye) are absorbed by a microelectronic array of the sensor-stimulator 990 that is hybridized to a glass piece 992 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 990 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.

[0073] The processing module 925 includes an image processor 923 that is in signal communication with the sensor-stimulator 990 via, for example, a lead 988 that extends through surgical incision 989 formed in the eye wall. In other examples, processing module 925 is in wireless communication with the sensor-stimulator 990. The image processor 923 processes the input into the sensor-stimulator 990 and provides control signals back to the sensor-stimulator 990 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 990. 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.

[0074] The processing module 925 can be implanted in the recipient and function by communicating with the external device 910, such as a BTE unit, a pair of eyeglasses, etc . The external device 910 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, the sensor-stimulator 990 captures light/images, in which sensor-stimulator 990 is implanted in the recipient.

[0075] FIGs. 10A- 10D illustrate an example cochlear implant system 1002 having one or more settings/parameters that can be determined based on validated healthcare information dataset that is validated using the techniques presented herein. The cochlear implant system 1002 comprises an external component 1004 that is configured to be directly or indirectly attached to the body of the user, and an intemal/implantable component 1012 that is configured to be implanted in or worn on the head of the user. In the examples of FIGs. 10A-10D, the implantable component 1012 is sometimes referred to as a “cochlear implant.” FIG. 10A illustrates the cochlear implant 1012 implanted in the head 1054 of a user, while FIG. 10B is a schematic drawing of the external component 1004 worn on the head 1054 of the user. FIG. 10C is another schematic view of the cochlear implant system 1002, while FIG. 10D illustrates further details of the cochlear implant system 1002. For ease of description, FIGs. 10A-10D will generally be described together. [0076] In the examples of FIGs. 10A-10D, the external component 1004 comprises a sound processing unit 1006, an external coil 1008, and generally, a magnet fixed relative to the external coil 1008. The cochlear implant 1012 includes an implantable coil 1014, an implant body 1034, and an elongate stimulating assembly 1016 configured to be implanted in the user’s cochlea. In one example, the sound processing unit 1006 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 implantable component 1012. In general, an OTE sound processing unit is a component having a generally cylindrically shaped housing 1011 and which is configured to be magnetically coupled to the user’s head 1054 (e.g., includes an integrated external magnet 1050 configured to be magnetically coupled to an intemal/implantable magnet 1052 in the implantable component 1012). The OTE sound processing unit 1006 also includes an integrated external (headpiece) coil 1008 (the external coil 1008) that is configured to be inductively coupled to the implantable coil 1014.

[0077] It is to be appreciated that the OTE sound processing unit 1006 is merely illustrative of the external devices that could operate with implantable component 1012. For example, in alternative examples, the external component 1004 can comprise a behind-the-ear (BTE) sound processing unit configured to be attached to, and worn adjacent to, the recipient’s ear. A BTE sound processing unit comprises a housing that is shaped to be worn on the outer ear of the user. In certain examples, the BTE is connected to a separate external coil assembly via a cable, where the external coil assembly is configured to be magnetically and inductively coupled to the implantable coil 1014, while in other embodiments the BTE includes a coil disposed in or on the housing worn on the outer ear of the user. It is also to be appreciated that alternative external components could be located in the user’s ear canal, worn on the body, etc.

[0078] Although the cochlear implant system 1002 includes the sound processing unit 1006 and the cochlear implant 1012, as described below, the cochlear implant 1012 can operate independently from the sound processing unit 1006, for at least a period, to stimulate the user. For example, the cochlear implant 1012 can operate in a first general mode, sometimes referred to as an “external hearing mode,” in which the sound processing unit 1006 captures sound signals which are then used as the basis for delivering stimulation signals to the user. The cochlear implant 1012 can also operate in a second general mode, sometimes referred as an “invisible hearing” mode, in which the sound processing unit 1006 is unable to provide sound signals to the cochlear implant 1012 (e.g., the sound processing unit 1006 is not present, the sound processing unit 1006 is powered-off, the sound processing unit 1006 is malfunctioning, etc.). As such, in the invisible hearing mode, the cochlear implant 1012 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 1012 in the external hearing mode are provided below, followed by details regarding operation of the cochlear implant 1012 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 1012 could also operate in alternative modes.

[0079] In FIGs. 10A and 10C, the cochlear implant system 1002 is shown with an external device 1010, configured to implement aspects of the techniques presented. The external device 1010, which is shown in greater detail in FIG. 10E, is a computing device, such as a personal computer (e.g., laptop, desktop, tablet), a mobile phone (e.g., smartphone), a remote control unit, etc. The external device 1010 and the cochlear implant system 1002 (e.g., sound processing unit 1006 or the cochlear implant 1012) wirelessly communicate via a bi-directional communication link 1026. The bi-directional communication link 1026 can comprise, for example, a short-range communication, such as Bluetooth link, Bluetooth Low Energy (BLE) link, a proprietary link, etc.

[0080] Returning to the example of FIGs. 10A-10D, the sound processing unit 1006 of the external component 1004 also comprises one or more input devices configured to capture and/or receive input signals (e.g., sound or data signals) at the sound processing unit 1006. The one or more input devices include, for example, one or more sound input devices 1018 (e.g., one or more external microphones, audio input ports, telecoils, etc.), one or more auxiliary input devices 1028 (e.g., audio ports, such as a Direct Audio Input (DAI), data ports, such as a Universal Serial Bus (USB) port, cable port, etc.), and a short-range wireless transmitter/receiver (wireless transceiver) 1020 (e.g., for communication with the external device 1010), each located in, on or near the sound processing unit 1006. 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 short-range wireless transceiver 1020 and/or one or more auxiliary input devices 1028 could be omitted).

[0081] The sound processing unit 1006 also comprises the external coil 1008, a charging coil, a closely-coupled radio frequency transmitter/receiver (RF transceiver) 1022, at least one rechargeable battery 1032, and an external sound processing module 1024. The external sound processing module 1024 can be configured to perform a number of operations that are represented in FIG. 10D by a sound processor 1033. The sound processor 1033 can be formed by one or more processors (e.g., one or more Digital Signal Processors (DSPs), one or more uC cores, etc.), firmware, software, etc. arranged to perform operations described herein. That is, the sound processor 1033 can be implemented as a firmware element, partially or fully implemented with digital logic gates in one or more application-specific integrated circuits (ASICs), partially or fully in software, etc. Although FIG. 10D illustrates the sound processor 1033 as being implemented/performed at the external sound processing module 1024, it is to be appreciated that this element (e.g., functional operations) could also or alternatively be implemented/performed as part of the implantable sound processing module 1058, as part of the external device 1010, etc.

[0082] Returning to the example of FIGs. 10A-10D, the implantable component 1012 comprises an implant body (main module) 1034, a lead region 1036, and the stimulating assembly 1016, all configured to be implanted under the skin (tissue) 1015 of the user. The implant body 1034 generally comprises a hermetically-sealed housing 1038 that includes, in certain examples, at least one power source 1025 (e.g., one or more batteries, one or more capacitors, etc.), in which the RF interface circuitry 1040 and a stimulator unit 1042 are disposed. The implant body 1034 also includes the intemal/implantable coil 1014 that is generally external to the housing 1038, but which is connected to the RF interface circuitry 1040 via a hermetic feedthrough (not shown in FIG. 10D).

[0083] As noted, the stimulating assembly 1016 is configured to be at least partially implanted in the user’s cochlea. The stimulating assembly 1016 includes a plurality of longitudinally spaced intra-cochlear electrical stimulating contacts (electrodes) 1044 that collectively form a contact array (electrode array) 1046 for delivery of electrical stimulation (current) to the recipient’s cochlea. The stimulating assembly 1016 extends through an opening in the recipient’s cochlea (e.g., cochleostomy, the round window, etc.) and has a proximal end connected to stimulator unit 1042 via lead region 1036 and a hermetic feedthrough (not shown in FIG. 10D). Lead region 1036 includes a plurality of conductors (wires) that electrically couple the electrodes 1044 to the stimulator unit 1042. The implantable component 1012 also includes an electrode outside of the cochlea, sometimes referred to as the extra-cochlear electrode (ECE) 1039.

[0084] As noted, the cochlear implant system 1002 includes the external coil 1008 and the implantable coil 1014. The external magnet 1050 is fixed relative to the external coil 1008 and the intemal/implantable magnet 1052 is fixed relative to the implantable coil 1014. The external magnet 1050 and the intemal/implantable magnet 1052 fixed relative to the external coil 1008 and the intemal/implantable coil 1014, respectively, facilitate the operational alignment of the external coil 1008 with the implantable coil 1014. This operational alignment of the coils enables the external component 1004 to transmit data and power to the implantable component 1012 via a closely-coupled wireless link 1048 formed between the external coil 1008 with the implantable coil 1014. In certain examples, the closely-coupled wireless link 1048 is an 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. 10D illustrates only one example arrangement.

[0085] As noted above, the sound processing unit 1006 includes the external sound processing module 1024. The external sound processing module 1024 is configured to process the received input audio signals (received at one or more of the input devices, such as sound input devices 1018 and/or auxiliary input devices 1028) and convert the received input audio signals into output control signals for use in stimulating a first ear of a recipient or user (i.e., the external sound processing module 1024 is configured to perform sound processing on input signals received at the sound processing unit 1006). Stated differently, the one or more processors (e.g., processing element(s) implementing firmware, software, etc.) in the external sound processing module 1024 are configured to execute sound processing logic in memory to convert the received input audio signals into output control signals (stimulation signals) that represent electrical stimulation for delivery to the recipient.

[0086] As noted, FIG. 10D illustrates an embodiment in which the external sound processing module 1024 in the sound processing unit 1006 generates the output control signals. In an alternative embodiment, the sound processing unit 1006 can send less processed information (e.g., audio data) to the implantable component 1012, and the sound processing operations (e.g., conversion of input sounds to output control signals 1056) can be performed by a processor within the implantable component 1012.

[0087] In FIG. 10D, according to an example embodiment, output control signals (stimulation signals) are provided to the RF transceiver 1022, which transcutaneously transfers the output control signals (e.g., in an encoded manner) to the implantable component 1012 via the external coil 1008 and the implantable coil 1014. That is, the output control signals (stimulation signals) are received at the RF interface circuitry 1040 via the implantable coil 1014 and provided to the stimulator unit 1042. The stimulator unit 1042 is configured to utilize the output control signals to generate electrical stimulation signals (e.g., current signals) for delivery to the user’s cochlea via one or more of the stimulating contacts 1044. In this way, cochlear implant system 1002 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 recipient to perceive one or more components of the input audio signals (the received sound signals).

[0088] As detailed above, in the external hearing mode, the cochlear implant 1012 receives processed sound signals from the sound processing unit 1006. However, in the invisible hearing mode, the cochlear implant 1012 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. 10D, an example embodiment of the cochlear implant 1012 can include a plurality of implantable sound sensors 1065(1), 1065(2) that collectively form a sensor array 1060, and an implantable sound processing module 1058. Similar to the external sound processing module 1024, the implantable sound processing module 1058 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.

[0089] In the invisible hearing mode, the implantable sound sensors 1065(1), 1065(2) of the sensor array 1060 are configured to detect/capture input sound signals 1066 (e.g., acoustic sound signals, vibrations, etc.), which are provided to the implantable sound processing module 1058. The implantable sound processing module 1058 is configured to convert received input sound signals 1066 (received at one or more of the implantable sound sensors 1065(1), 1065(2)) into output control signals 1056 for use in stimulating the first ear of a recipient or user (i.e., the implantable sound processing module 1058 is configured to perform sound processing operations). Stated differently, the one or more processors (e.g., processing element(s) implementing firmware, software, etc.) in the implantable sound processing module 1058 are configured to execute sound processing logic in memory to convert the received input sound signals 1066 into output control signals 1056 that are provided to the stimulator unit 1042. The stimulator unit 1042 is configured to utilize the output control signals 1056 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.

[0090] 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 1002 could operate differently in different embodiments. For example, in one alternative implementation of the external hearing mode, the cochlear implant 1012 could use signals captured by the sound input devices 1018 and the implantable sound sensors 1065(1), 1065(2) of sensor array 1060 in generating stimulation signals for delivery to the user.

[0091] According to the techniques of the present disclosure, the external sound processing module 1024 can also include an inertial measurement unit (IMU) 1070. The IMU 1070 is configured to measure the inertia of the user's head, that is, motion of the user's head. As such, the IMU 1070 comprises one or more sensors 1075 each configured to sense one or more of rectilinear or rotatory motion in the same or different axes. Examples of sensors 1075 that can be used as part of inertial measurement unit 1070 include accelerometers, gyroscopes, inclinometers, compasses, and the like. Such sensors can be implemented in, for example, micro electromechanical systems (MEMS) or with other technology suitable for the particular application.

[0092] As also illustrated in FIG. 10D, in certain examples, a second IMU 1080 including one or more sensors 1085 is incorporated into implantable sound processing module 1058 of implant body 1034. The second IMU 1080 can serve as an additional or alternative inertial measurement unit to the IMU 1070 of external sound processing module 1024. Like sensors 1075, sensors 1085 can each be configured to sense one or more of rectilinear or rotatory motion in the same or different axes. Examples of sensors 1085 that can be used as part of inertial measurement unit 1080 include accelerometers, gyroscopes, inclinometers, compasses, and the like. Such sensors can be implemented in, for example, MEMS or with other technology suitable for the particular application. For hearing devices that include an implantable sound processing module, such as implantable sound processing module 1058, that includes an IMU, such as the IMU 1080, the techniques presented herein can be implemented without an external processor. Accordingly, a hearing device that includes an implant body 1034 and lacks an external component 1004 can be configured to implement the techniques presented herein. [0093] 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.

[0094] 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.

[0095] 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.

[0096] 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.

[0097] 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.

[0098] 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.

[0099] 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

CLAIMS What is claimed is:

1. A method comprising : receiving medical data associated with a patient; analyzing the medical data using a data-derived model to determine whether the medical data includes incongruities; and performing one or more actions based on whether the medical data includes incongruities.

2. The method of claim 1, wherein performing one or more actions based on whether the medical data includes incongruities comprises: generating validated medical data.

3. The method of claim 2, further comprising: fitting a medical device to a recipient using the validated medical data.

4. The method of claim 1, wherein the data-derived model includes a machine learning model.

5. The method of claim 1, wherein the data-derived model includes a principal component analysis (PCA) model.

6. The method of claim 1, wherein the data-derived model includes a neural network.

7. The method of claim 1, 2, 3, 4, 5, or 6, further comprising: processing the medical data based on a type of the medical data and a type of the data- derived model.

8. The method of claim 1, 2, 3, 4, 5, or 6, wherein analyzing the medical data includes performing pattern matching based on the data-derived model.

9. The method of claim 1, 2, 3, 4, 5, or 6, wherein, when the medical data includes one or more incongruities, performing the one or more actions includes : performing an incongruity impact assessment.

10. The method of claim 9, wherein performing the incongruity impact assessment includes: analyzing an incongruous variable associated with the one or more incongruities to determine a degree of a medical impact of the one or more incongruities on the patient.

11. The method of claim 1, 2, 3, 4, 5, or 6, wherein analyzing the medical data includes: transforming a structure of the medical data based on a type of the data-derived model to produce modeled data; and analyzing the modeled data to determine whether the medical data includes incongruities.

12. The method of claim 11, wherein analyzing the modeled data includes: using the data-derived model to perform pattern matching on the modeled data based on other medical data from a plurality of other patients.

13. A method comprising : receiving a healthcare information dataset; and training a model using the healthcare information dataset to identify incongruities in subsequently received healthcare information datasets.

14. The method of claim 13, wherein training the model includes training the model based on a type of health information associated with the healthcare information dataset.

15. The method of claim 13 or 14, wherein training the model includes training the model based on a type of the model.

16. The method of claim 13 or 14, wherein the model is a machine learning model.

17. The method of claim 13 or 14, wherein the model is a principal component analysis

(PCA) model.

18. The method of claim 17, further comprising: transforming the healthcare information dataset using the PC A model; deriving weight distributions for each principal component of the healthcare information dataset; and outputting the PCA model and the weight distributions.

19. The method of claim 13 or 14, further comprising calibrating the model to remove unnecessary elements from the model.

20. The method of claim 13 or 14, wherein the healthcare information dataset includes heath information associated with a plurality of patients.

21. The method of claim 20, wherein the healthcare information dataset and the subsequently received healthcare information datasets are associated with a same type of health information.

22. One or more non-transitory computer readable storage media comprising instructions that, when executed by a processor, cause the processor to: obtain data associated with a user; model the data using a trained model to produce modeled data; and analyze the modeled data using the trained model to determine whether the data includes incongruous data.

23. The one or more non-transitory computer readable storage media of claim 22, wherein the instructions further cause the processor to: determine that the data includes incongruous data; and perform an impact assessment analysis on the incongruous data.

24. The one or more non-transitory computer readable storage media of claim 23, wherein, when performing the impact assessment analysis, the instructions further cause the processor to determine a degree of a medical impact of the incongruous data on the user.

25. The one or more non-transitory computer readable storage media of claim 24, further comprising instructions that further cause the processor to: generating validated medical data based on the impact assessment analysis and the degree of a medical impact.

26. The one or more non-transitory computer readable storage media of claim 22, 23, 24, or 25, wherein the trained model is a machine learning model.

27. The one or more non-transitory computer readable storage media of claim 22, 23, 24, or 25, wherein the trained model is a principal component analysis (PCA) model.

28. The one or more non-transitory computer readable storage media of claim 22, 23, 24, or 25, wherein the data and the trained model are associated with a same type of healthcare information.

29. The one or more non-transitory computer readable storage media of claim 22, 23, 24, or 25, further comprising instructions that further cause the processor to: generate validated medical data based on the analyzing of the modeled data.

30. The one or more non-transitory computer readable storage media of claim 22, 23, 24, or 25, further comprising instructions that further cause the processor to: configure a medical device for a recipient using the validated medical data.

31. A system comprising : a memory; and at least one processor operable coupled to the memory, wherein the at least one processor is configured to: receive a healthcare information dataset associated with a patient; model a structure of the healthcare information dataset using a model to produce a modeled healthcare information dataset; and validate the modeled healthcare information dataset.

32. The system of claim 31, to validate the modeled healthcare information dataset, the at least one processor is configured to: analyze the modeled healthcare information to determine whether the healthcare information dataset includes incongruous information; and determine an impact of the incongruous information on the patient when the healthcare information dataset includes incongruous information.

33. The system of claim 31 or 32, wherein the model is a machine learning model.

34. The system of claim 31 or 32, wherein the healthcare information dataset and the model are associated with a particular type of healthcare information.

35. The system of claim 31 or 32, wherein the model is a principal component analysis (PCA) model.

36. The system of claim 31 or 32, wherein the model is trained using healthcare information datasets associated with a plurality of patients.