Systems and methods for serial monitoring of hearing and forecasting changes in hearing

Abstract

Disclosed are various embodiments for forecasting hearing change. Various embodiments of the present disclosure can receive a first audiogram of a patient and determine a patient-individualized predicted region for subsequent audiograms. Various embodiments can receive a second audiogram subsequent to the patient being exposed an audiopathic damaging factor or an intervention designed to improve auditory function. Various embodiments can provide a risk assessment of auditory injury and / or a treatment recommendation. Conversely, various embodiments can provide an estimate of the likelihood that auditory function has improved, based on a determination that at least a portion of the second audiogram does not fall within the patient-individualized predicted region.

Description

ATTORNEY DOCKET NO.: 37759.0635P1PATENTSYSTEMS AND METHODS FOR SERIAL MONITORING OF HEARING AND FORECASTING CHANGES IN HEARING CROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to U. S. Provisional Application No. 63 / 788,571. titled " SYSTEMS AND METHODS FOR SERIAL MONITORING OF HEARING AND FORECASTING CHANGES IN HEARING", filed April 14, 2025, which is hereby incorporated by reference in its entirety.STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT

[0002] This invention was made with government support under the U. S. Department of Veterans Affairs (VA) Rehabilitation Research, Development, and Translation (RRD& T) Merit Award (ERA Commons # I01 RX003127; VA #C3127R), the VA RRD& T SPiRE Award (ERA Commons # I21 RX003483; VA C3483P), the VA RRD& T Research Career Scientist Award (ERA Commons # RX22019), and the VA RRD& T Center Award (ERA Commons # I50 RX002361; VA #C2361C), each provided by the U. S. Department of Veterans Affairs. The government has certain rights in the invention.BACKGROUND

[0003] Hearing loss represents a significant health concern affecting millions of individuals worldwide, linked to reduced social participation, diminished workplace productivity, depression, and cognitive decline. Various factors can cause or contribute to hearing loss, including age, noise exposure, and exposure to ototoxic substances. Certain medical treatments, while effective for their intended therapeutic purposes, can have adverse effects on auditory function. For example, cisplatin is a potent chemotherapeutic agent effective against several cancers, but is also a known ototoxin, with incidence of hearing shifts reported in up to 50-60% of patients. These ototoxic effects, which damage outer hair cells within the organ of Corti and marginal cells within the stria vascularis, are often irreversible and typically progress from the cochlear base toward the apex with each treatment.

[0004] Predicting when cochlear damage from ototoxic agents results in hearing loss is challenging because of variability in individual susceptibility, influenced by risk factorsATTORNEY DOCKET NO.: 37759.0635P1 including patient age, pre-existing hearing loss, cumulative drug dosage, and concurrent treatments such as radiation therapy.

[0005] Existing approaches to predicting hearing changes have generally focused on forecasting a single criterion outcome, such as a binary determination of whether hearing loss has occurred or a simplified metric such as a pure tone average. Such outcomes may be influenced by various sources of bias including regression to the mean, multiple correlations in the data, and information loss. Additionally, the clinical value of each outcome depends on the objectives and priorities of the end user, whether oncologists comparing treatment options, audiologists identifying incipient ototoxic effects, or patients prioritizing pretreatment counseling.

[0006] Further challenges exist in handling incomplete audiometric data, including " No Response" thresholds when a patient's hearing loss exceeds equipment measurement capability. There is a need for improved systems and methods that can forecast individualized hearing changes across the full audiometric profile, accommodate multiple clinical use cases, and appropriately handle incomplete or censored audiometric data.SUMMARY

[0007] This summary is provided to introduce a selection of concepts in a simplified form that are further described below in the detailed description. This summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter.

[0008] The present disclosure relates to a method for monitoring hearing changes in patients undergoing auditory damaging treatments. The method receives a first audiogram comprising sound levels at multiple pitches, determines a patient-individualized predicted region for expected shifts based on population hearing data, and provides treatment recommendations when a subsequent audiogram falls outside the predicted region.

[0009] The present disclosure also relates to a method for assessing changes in auditory function following exposure to audiopathic factors. The method receives a first audiogram, determines a patient-individualized predicted region, and provides an assessment representing a likelihood of change in auditory function when a subsequent audiogram does not fall within the predicted region.

[0010] The present disclosure further relates to a method for forecasting hearing changes based on auditory function changing factors. The method receives a first audiogram, receivesATTORNEY DOCKET NO.: 37759.0635P1 data corresponding to an auditory function changing factor, forecasts a predicted hearing change, and provides a treatment recommendation based on the forecast.

[0011] The foregoing general description of the illustrative embodiments and the following detailed description thereof are merely exemplary aspects of the teachings of this disclosure and are not restrictive.BRIEF DESCRIPTION OF FIGURES

[0012] Non-limiting and non-exhaustive examples are described with reference to the following figures.

[0013] The accompanying drawings, which are incorporated in and constitute a part of the present description serve to explain the principles of the apparatuses and systems described herein:

[0014] FIG. 1 shows an example system;

[0015] FIG.2 show's a flow chart of a first example method;

[0016] FIGs.3-6 show' example audiograms for patients; and

[0017] FIG.7 shows a flowchart of a second example method.DETAILED DESCRIPTION

[0018] The following description sets forth exemplary aspects of the present disclosure. It should be recognized, however, that such description is not intended as a limitation on the scope of the present disclosure. Rather, the description also encompasses combinations and modifications to those exemplary' aspects described herein.

[0019] As used in the specification and the appended claims, the singular forms “a,” “an,” and “the” include plural referents unless the context clearly dictates otherwise. Ranges can be expressed herein as from “about” one particular value, and / or to “about” another particular value. When such a range is expressed, another configuration includes from the one particular value and / or to the other particular value. When values are expressed as approximations, by use of the antecedent “about,” it will be understood that the particular value forms another configuration. It will be further understood that the endpoints of each of the ranges are significant both in relation to the other endpoint, and independently of the other endpoint.ATTORNEY DOCKET NO.: 37759.0635P1

[0020] “Optional’" or “optionally” means that the subsequently described event or circumstance may or may not occur, and that the description includes cases where said event or circumstance occurs and cases where it does not.

[0021] Throughout the description and claims of this specification, the word “comprise” and variations of the word, such as “comprising” and “comprises,” means “including but not limited to,” and is not intended to exclude other components, integers or steps. “Exemplary’” means “an example of’ and is not intended to convey an indication of a preferred or ideal configuration. “Such as” is not used in a restrictive sense, but for explanatory purposes.

[0022] It is understood that when combinations, subsets, interactions, groups, etc. of components are described that, while specific reference of each various individual and collective combinations and permutations of these cannot be explicitly described, each is specifically contemplated and described herein. This applies to all parts of this application including, but not limited to, steps in described methods. Thus, if there are a variety of additional steps that can be performed it is understood that each of these additional steps can be performed with any specific configuration or combination of configurations of the described methods.

[0023] As will be appreciated by one skilled in the art, the methods and systems can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the methods and systems can take the form of a computer program product on a computer-readable storage medium having computer-readable program instructions (e.g., computer software) embodied in the storage medium. More particularly, the present methods and systems can take the form of web-implemented computer software. Any suitable computer-readable storage medium can be utilized including hard disks, CD-ROMs, optical storage devices, magnetic storage devices, memristors, Non-Volatile Random Access Memory (NVRAM), Random Access Memory (RAM), flash memory, or a combination thereof.

[0024] Throughout this application reference is made to block diagrams and flowcharts. It will be understood that each block of the block diagrams and flowcharts, and combinations of blocks in the block diagrams and flowcharts, respectively, can be implemented by processor-executable instructions. These processor-executable instructions can be loaded onto a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the processor-executable instructionsATTORNEY DOCKET NO.: 37759.0635P1 which execute on the computer or other programmable data processing apparatus create a device for implementing the functions specified in the flowchart block or blocks.

[0025] These processor-executable instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the processor-executable instructions stored in the computer-readable memory produce an article of manufacture including processor-executable instructions for implementing the function specified in the flowchart block or blocks. The processor-executable instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the processor-executable instructions that execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks.

[0026] Accordingly, blocks of the block diagrams and flowcharts support combinations of devices for performing the specified functions, combinations of steps for performing the specified functions and program instruction means for performing the specified functions. It will also be understood that each block of the block diagrams and flowcharts, and combinations of blocks in the block diagrams and flowcharts, can be implemented by special purpose hardware-based computer systems that perform the specified functions or steps, or combinations of special purpose hardware and computer instructions.

[0027] This detailed description can refer to a given entity performing some action. It should be understood that this language can in some cases mean that a system (e.g., a computer) owned and / or controlled by the given entity is actually performing the action.

[0028] Various embodiments of the present disclosure are directed to serial monitoring of hearing and forecasting changes in hearing based on audiograms. In at least some embodiments, serial monitoring and forecasting of the audiogram can be used for hearing conservation aimed at identifying and predicting hearing change (e.g., loss, improvement, etc.). In some other embodiments, serial monitoring and forecasting of the audiogram can be used for identifying or predicting improvements in hearing (e.g, following surgical, therapeutic, rehabilitative interventions, etc.). These use cases are relevant not only to the care of individual patients in various clinical settings, they can optimize the design and execution of clinical trials involving groups of patients, and can be directed toward researchATTORNEY DOCKET NO.: 37759.0635P1 applications seeking to understand how various patient factors (e.g., age, genetics, etc. interact with a specific exposure or an interventional factor to impact hearing.

[0029] Audiometric serial monitoring is the act of evaluating changes in hearing thresholds. In various embodiments, audiometric serial monitoring can be performed by a trained professional (e.g., tester, audiologist, etc.) by evaluating the changes in hearing thresholds as shown by the audiogram. An audiogram is a graphical representation of an individual's sensitivity to pure tones, which are single-frequency sounds. Often, an audiogram can depict the individual’s sensitivity to pure tones based on audiometric thresholds at specific tones. An audiometric threshold quantifies the minimum sound level at which a specific tone can be detected by the subject at least 50% of the time. During an audiometric evaluation, a professional trained to test a subject’s hearing can present a series of pure tones through earphones or speakers, at varying sound levels (often measured in decibels) and test frequencies (often measured in Hertz), to determine the subject’s audiometric thresholds to sound that is perceived through vibrations in the air (‘‘air-conduction thresholds”). In at least some embodiments, the trained professional (e.g., tester, audiologist, etc.) can present tones through vibrations directed toward bones in the subject’s skull (“bone-conduction thresholds”). Both air-conduction thresholds and bone-conduction thresholds can be used to diagnose the presence and severity of hearing change. Comparing air-conduction thresholds and bone-conduction thresholds may be way to determine the type of hearing change (e.g., sensorineural versus conductive impairment, etc.}.

[0030] Audiograms depict the audiometric threshold for a tone of a specific frequency. Typically, an audiogram includes measurements at approximately seven or eight different frequencies. The selection of frequencies for testing is determined by the specific use case. For example, higher frequencies are often emphasized in hearing conservation programs due to their susceptibility to the damaging effects from ototoxic drugs, noise, and aging, given the physiological characteristics of the cochlea. In various embodiments, the test frequencies can include standard audiometric frequencies, such as 0.25, 0.5, 1, 2, 3, 4, 6, and 8 kHz. Extended high-frequency audiometry may additionally include frequencies above 8 kHz, such as 9, 10, 11.2, 12.5, 14, and 16 kHz, which can be particularly useful for early detection of ototoxic damage since high-frequency hearing loss often precedes changes at conventional audiometric frequencies. The cochlea is tonotopically organized, with high-frequency sounds processed at the basal end and low-frequency sounds processed at the apical end. Because ototoxic agents and noise exposure typically damage the basal region of the cochlea first,ATTORNEY DOCKET NO.: 37759.0635P1 monitoring higher frequencies can provide earlier warning of auditory damage before it progresses to frequencies critical for speech understanding.

[0031] Audiometric thresholds are recorded as a series of integers. In various embodiments, the thresholds are often recorded in increments of five decibels (e.g., 5, 10, 15, 20, 25 dB, etc.). Higher values indicate poorer hearing sensitivity, as these values represent the minimum sound level required for the subject to detect the tone. Typically, each ear is tested separately, resulting in a set of measurements for both the left and right ears. Although the audiometric threshold could theoretically increase infinitely until the patient detects the tone, safety limits are often observed during the audiometric testing process to prevent potential damage to the subject's hearing. An audiometer (the device frequently used for obtaining these measurements) often includes technical constraints on the maximum sound level it can produce. Excessive sound levels could damage the auditory system. These safety limits can introduce complexities in standardizing audiometric assessment methodologies, as persons who are incredibly hard of hearing might still have some hearing at incredibly high sound levels.

[0032] Trained professionals (e.g., tester, audiologist, etc.) can identify changes in a patient's hearing by comparing audiogram results over time. The rationale is that pure tone sensitivity, as measured by the audiogram, is susceptible to damage from audiopathic exposures, such as noise or ototoxic medications, and reflect changes associated with normal aging and certain disease conditions. Additionally, improving pure tone sensitivity, as measured by the audiogram, is often a surgical or therapeutic goal. A change in pure-tone sensitivity' is taken as evidence of a potential change in auditory function. Depending on the use case, such changes could motivate follow-up care and / or removal from the audiopathic exposure, or could constitute proof that a surgical or therapeutic treatment was successful. However, various problems exist with the existing methodologies to identify hearing changes.

[0033] A first problem that exists with the existing methodologies is that there is a lack of a ‘“gold standard” in the industry for identifying changes in auditory function. Since there is no gold standard for reduced or improved auditory function and thus no way to evaluate the accuracy of the various existing criteria, it is up to the audiologist(s), tester(s), and / or trained professional(s) to choose among serial monitoring criteria based on clinical objectives, convention, intuition, invasiveness, time, expense, or any other priority. Priorities often differ among the end users, such as the trained professional (e.g., audiologist, tester, etc.), primaryATTORNEY DOCKET NO.: 37759.0635P1 care clinician, employer, and patient. The various monitoring criteria cannot simultaneously achieve the objectives of all stakeholders, resulting in inefficient care.

[0034] A second problem is that the trained professional (e.g., tester, audiologist, etc. ) must also decide how to handle thresholds that exceed audiometer test limits, called “No Response” (NR), or how to handle missing thresholds due to patient non-response. The latter is particularly challenging in pediatric applications, while the former often occurs in older populations of patients or in patients with severe hearing loss. When a patient's hearing threshold exceeds the maximum output capability of the audiometer at a given frequency, the trained professional cannot determine the actual threshold value, only that it exceeds the equipment's detection limit. This creates a censored data problem where valuable information about the patient's hearing is lost or improperly characterized. Existing approaches to handling NR thresholds are inadequate. One common approach is to impute the NR threshold to the detection limit plus an arbitrary constant (e.g., 5 dB), but this implies an observation that was never actually made and introduces artificial certainty about the patient's hearing status. Another approach is to treat the NR measurement as completely missing, but this discards the valuable information that the threshold exceeds the detection limit. Neither approach properly accounts for the uncertainty inherent in censored measurements. Missing thresholds due to patient non-response present similar challenges, particularly in pediatric populations where young children may be unable to reliably respond to audiometric testing, or in populations with cognitive impairments that affect test compliance. The manner in which NR and missing thresholds are handled will directly impact clinical judgments, potentially leading to false referrals when arbitrary imputation methods overstate certainty, or missed detections when valuable censored information is discarded.

[0035] A third problem is a result of a bias due to regression to the mean. Regression to the mean is the fact that, barring any real changes, bigger than average baselines are always expected to get smaller, and smaller than average baselines are always expected to get bigger. This is necessarily true in many homeostatic systems, real or imaginary. In the absence of a change in auditory function, regression to the mean guarantees that, on average, a “large” or “high” baseline threshold will be followed by a “smaller” or “lower” one, and that a “small” baseline threshold will be followed by a “larger” one. This is expected regardless of any auditory function change that may have occurred. This clearly confounds any attempt to judge audiometric changes in terms of potential injury or improvement to the auditory system, because any observed changes are at least partially due to regression to the mean. AATTORNEY DOCKET NO.: 37759.0635P1 proper approach is to statistically condition the expected follow-up measurement on the previously observed baseline. The clinical expectation about a patient at follow-up naturally depends on what was observed at baseline, and a proper statistical expectation for a patient at follow-up must also depend on the previously observed baseline. Regression to the mean induces bias in all existing serial monitoring criteria.

[0036] The first problem impacts most every facet of audiology or medicine. The second and third problems occur in most hearing monitoring criteria because standard methods of evaluating changes in pure tone sensitivity are based on the computed difference between baseline and follow-up audiograms. While intuitive, the computed difference approach will cause bias and loss of information. The trained professional (e.g, tester, audiologist, etc.) must perform the differencing computations to determine if a given criterion has been met. Subsequently, the trained professional must communicate the results to the patient and other stakeholders in their care (e.g., family, care team, employer, etc.). There is a need for rapid or even real-time communication of these results, particularly when results indicate the need for care coordination, for example to eliminate or reduce the audiopathic exposure, or promote timely access to treatment. An unbiased, rapid and transparent way to communicate serial monitoring results would promote more efficient care and limit significant variability, leading to inconsistencies in clinical interpretations. This variability is particularly concerning for populations such as veterans, who often experience noise-related hearing loss, age-related hearing loss, workplace-related hearing loss, and / or medication-related hearing loss. In patients who already have some hearing loss, further losing hearing from one or more exposures can severely impact their quality of life, underscoring the need for consistent and reliable assessment methodologies.

[0037] At least one solution to the various problems includes a serial monitoring embodiment of the present disclosure. The serial monitoring methodology enables stakeholders (e.g., trained professionals, audiologists, individual doctors or medical practitioners, healthcare teams, and patients, etc.) to evaluate whether a patient has experienced a significant or alarming change in hearing over the course of exposure to auditory changing factors (e.g., auditory damaging factors, auditory improving factors, etc.), or over the course of an intervention to improve hearing outcomes. Potentially damaging factors include treatment with chemotherapy drugs, such as Cisplatin and Carboplatin, which are known to affect hearing, as well as antibiotics, like aminoglycosides, which have long been associated with hearing damage. Potentially improving factors can include auditoryATTORNEY DOCKET NO.: 37759.0635P1 surgeries and various possible medications. Additionally, this methodology can be applied to monitor hearing changes due to noise exposure, such as workplace noise, or toxic exposures (e.g., jet fuel), thereby extending the usage of the methodology to employees and military personnel, as well as patients.

[0038] For purposes of this disclosure, it should be assumed that serial monitoring occurs under the assumption that pure tone sensitivity does not change between the baseline and follow-up time point, which is often called the " Homeostasis Hypothesis,” and audiometric serial monitoring is conducted to evaluate whether or not the Homeostasis Hypothesis is true. If the follow-up audiogram is unusual with respect to the expectations of the Homeostasis Hypothesis, then the trained professional has evidence against the assumption that pure tone sensitivity has remained constant over the course of exposure. Follow-up action is therefore warranted.

[0039] Various embodiments of the serial monitoring method include measuring a person's audiogram (a pre-exposure audiogram), which reflects sensitivities to sound, before the onset of treatment or workplace exposure. Subsequent measurements (one or more postexposure audiograms) are then compared to the initial audiogram, allowing participants or stakeholders to determine whether a change has occurred. The serial monitoring method also addresses various technical issues related to measurement and statistical challenges in managing these measurements.

[0040] At least one benefit of the serial monitoring method is that the model, which is used to determine whether there is a change in a patient’s hearing, does not require modification for different types of exposures. Instead the serial monitoring approach may include a pre-exposure measurement, which occurs before the exposure begins, and at least one post-exposure measurement, which occurs after the exposure begins. In at least some embodiments, the measurements can be temporally spaced out from at a predetermined amount of time (e.g., a week, a month, multiple months, a year, etc.). It should be noted that the greater time between when the measurements are taken may affect the result of the serial monitoring method due to additional factors, such as aging and additional exposures to loud noises can affect hearing. In various embodiments, post-exposure measurements are taken three months or six months from the pre-exposure measurement.

[0041] In various embodiments, the analysis can be performed using Gaussian processes. Gaussian processes provide an alternative to traditional grading or binary scales that are prone to the biases, as discussed above. This methodology is suitable for patients andATTORNEY DOCKET NO.: 37759.0635P1 employees at risk of audiopathic injury from any type of exposure (e.g., noise, bactericidal or antineoplastic therapies, etc.) as long as baseline audiometry is available. This methodology is also suitable for patients and employees who undergo an intervention to improve hearing outcomes (e.g., surgical, therapeutic, or rehabilitative, etc., as long as baseline audiometry is available. Further, no standard guidelines are required for determining pure tone test frequencies, testing intervals, or procedures for treating patients with auditory function changes. Such decisions are specific to each exposure and are left to the serial monitoring program. Accordingly, various embodiments avoid bias and loss of information that affects current approaches, and can serve a wider range of clinical objectives and stakeholder priorities than standard criteria currently in use.

[0042] The clinical problem for the trained professional is that of deciding whether a follow-up audiogram measured on a patient demonstrates evidence that pure tone sensitivity has degraded and that the Homeostasis Hypothesis is false. The statistical problem is that of defining (1) the relationship between pure tone sensitivity, a theoretical construct that we cannot observe directly, and its representation as the audiogram, and (2) defining the expected relationship between baseline and follow-up audiograms under the assumption that the Homeostasis Hypothesis is correct.

[0043] To determine whether the audiogram is consistent with expectations given by the baseline audiogram assuming that the Homeostasis Hypothesis is true, comparisons are made with follow-up audiogram (post-exposure audiogram) to a multivariate Normal distribution calculated from the pre-exposure audiogram and population data. These calculations can take into account age and sex specific population information to better establish a baseline expectation, along with the pre-exposure audiogram, to determine the multivariate Normal distribution. To facilitate clinical applications, the predictive distribution is typically distilled into one or more prediction intervals. A follow-up threshold that lies outside the prediction interval is unexpected and worthy of further consideration, either by clinical referral or even ignoring the result. This decision is left to the trained professional.

[0044] At least another solution presented in various embodiments of the disclosure focuses on forecasting hearing change based on a patient’s audiogram and a possible exposure. For example, one possible exposure that could affect a patient’s hearing is the use of chemotherapy drugs, particularly Cisplatin, which is commonly used for cancers of the head and neck. Cisplatin is widely known to damage hearing. Such a solution would provide a forecast for the oncology team, the patient, and the audiology team regarding the expectedATTORNEY DOCKET NO.: 37759.0635P1 changes in hearing, specific to the patient, given the prescribed Cisplatin treatment. Before treatment begins, the patient's audiogram and Cisplatin treatment regimen, including dosage and timing, as well as concurrent radiation and other patient-specific features, are input into the forecasting model. This model can also incorporate genotyping and other relevant data. Based on this input information, the model forecasts the expected changes in hearing that the patient will experience over the course of treatment. This methodology allows for anticipation of changes before they occur.

[0045] The goal of forecasting is to enable the oncology team to modify the treatment regimen if there is a serious risk of debilitating hearing loss or to allow the audiology team to adjust their monitoring schedule. This provides information about potential toxicity associated with Cisplatin treatment, specifically regarding changes in hearing. Another goal of forecasting could be to demonstrate possible improvements to hearing based on a specified therapy.

[0046] The forecasting methodology must be adapted to specific types of exposure. In one example, the forecasting methodology is developed for Cisplatin, a cancer treatment that is known to cause hearing loss. However, additional treatments could also use such a forecasting methodology, such as cystic fibrosis patients receiving regular antibiotic treatment with aminoglycosides. Such forecasting methodologies could also address hearing changes from noise exposure, such as blast exposures or firearm training noise. Additionally, such forecasting methodologies could also address hearing improvements for various therapies, treatments, and / or surgeries.

[0047] In various embodiments, the data corresponding to the auditory function changing factor may include information regarding concurrent radiation to the cochlea. The forecasting model may incorporate a binary indicator corresponding to the presence or absence of radiation exposure to the cochlea up to a given timepoint. In some embodiments, the model may account for the combined effects of cumulative drug dosage and concurrent radiation on pure-tone thresholds. For example, higher cumulative doses of cisplatin may result in progressively greater magnitudes of hearing loss, and concurrent radiation may have a marked impact on hearing changes, particularly at higher frequencies. In at least one embodiment, model results indicate that after a cumulative cisplatin dose of 280 mg / m2, 8 kHz thresholds may increase by an average of 12 dB with concurrent radiation versus 8 dB without concurrent radiation. The forecasting model may use a parameter vector representing the average change in pure tone threshold in each ear and at each test frequency per unit ofATTORNEY DOCKET NO.: 37759.0635P1 cumulative drug exposure, and a separate parameter vector representing the effect of radiation on pure tone thresholds.

[0048] A single, pre-exposure audiogram can be used for the forecasting methodology. Various models of the present disclosure can use a predetermined number of frequencies in the audiogram measured in both the left and right ears. While the forecasting algorithm can function with fewer frequencies, various embodiments developed and tested utilize between five and thirteen frequencies.

[0049] The process involves taking an initial measurement of the patient's hearing and reviewing the treatment. For example, the treatment can be the administration of a medication, such as Cisplatin, which specifies the dosage and timing. For example, patients with head and neck cancers commonly receive Cisplatin as 40 milligrams per meter squared weekly for seven or eight weeks. Given the prescription and the pretreatment audiogram, the model performs matrix manipulations to forecast the audiogram at various points during treatment. This involves matrix multiplications and inversions using mathematical parameters estimated from a large population of patients who have received such a treatment (e.g.,Cisplatin). The model uses this prior experience to generate expectations for new patients based on their treatment schedule and baseline features.

[0050] In various embodiments, the forecasting model may generate predictions at multiple future timepoints during a course of treatment. For example, for a patient scheduled to receive multiple infusions of a chemotherapy drug, the model may forecast the audiogram after each planned infusion, thereby providing the clinical care team with a trajectory of expected hearing changes over the course of treatment. The timepoints for which forecasts are to be made may be dictated by the needs of the clinical care team or the wishes of the patient. In some embodiments, the prediction regions may widen and shift toward higher threshold values as the cumulative drug dose increases, reflecting the progressive ototoxic effect of the treatment. The upper boundary of the prediction region may represent a bestcase scenario above which a specified percentage of follow-up thresholds are expected to lie, while the lower boundary may correspond to thresholds above which the audiologist is a specified percentage certain the follow-ups will lie. and thus may be interpreted as a worstcase hearing loss prediction for the patient.

[0051] The relationship between drug exposure and changes in hearing is determined through regression analysis of existing data, allowing for accurate forecasts. The model employs a Bayesian approach and multivariate normal regression modeling. New patientATTORNEY DOCKET NO.: 37759.0635P1 audiograms can be incorporated into the forecasting model to improve accuracy for future patients.

[0052] In various embodiments, the forecasting model may be implemented as a Bayesian multivariate linear mixed model. Audiograms may be forecasted for specific timepoints relative to an initial treatment, conditional on the patient's pre-exposure audiogram, patient age, and prescribed treatment regimen. The model may employ a Gaussian process framework, using machine learning techniques to efficiently predict probability distributions for pure-tone thresholds at multiple frequencies in both ears. By assuming the output values are drawn from a multivariate Gaussian distribution, the model may provide a statistically-based method for forecasting and identifying hearing changes that is not prone to the biases described above. The model architecture may allow the exposure variables to vary, facilitating application to different exposures and patient populations. In some embodiments, removing the treatment variables from the forecasting model yields a model that can be used to screen hearing during treatment to identify statistically significant hearing changes.

[0053] In various embodiments, the following steps may be performed to generate predictions using the forecasting model. First, the patient's age, baseline audiogram for the left and right ears, and the cumulative drug exposure and presence or absence of concurrent radiation exposure for the timepoint at which predictions are to be made may be gathered. Second, the thresholds in the baseline left and right ear audiograms may be scaled (e.g, divided by a constant such as ten) to facilitate computation, and stacked into a single ordered vector. Third, the cumulative drug dose may be similarly scaled (e.g., divided by one-hundred). Fourth, a centered patient age may be computed (e.g., by subtracting a reference age and dividing by a scaling factor), and the population mean audiogram may be computed as a function of patient age and test frequency, including quadratic terms and interactions. Fifth, the covariance among elements of the patient-specific random effects may be computed using a separable squared exponential covariance function between ears and between test frequencies. Sixth, the residual variance matrix may be computed as a diagonal matrix of frequency-specific residual variances. Seventh, the expected follow-up audiogram and covariance may be computed. In some embodiments, when the baseline audiogram is completely missing, the predictive distribution may be based on the population mean audiogram and the treatment exposure alone. In some embodiments, when the baseline audiogram is partially missing due to No Response or missing thresholds, the missingATTORNEY DOCKET NO.: 37759.0635P1 elements may be imputed using a truncated multivariate Normal distribution conditioned on the observed thresholds, and the augmented baseline audiogram may then be used to compute the predictive distribution.

[0054] With reference to FIG. 1, shown is a network environment 100 according to various embodiments. The network environment 100 can include a computing environment 103 and a client device 106, which can be in data communication with each other via a network 109.

[0055] The network 109 can include wide area networks (WANs), local area networks (LANs), personal area networks (PANs), or a combination thereof. These networks can include wired or wireless components or a combination thereof. Wired networks can include Ethernet networks, cable networks, fiber optic networks, and telephone networks such as dial-up, digital subscriber line (DSL), and integrated services digital network (ISDN) networks. Wireless networks can include cellular networks, satellite networks, Institute of Electrical and Electronic Engineers (IEEE) 802.11 wireless networks (i.e., WI-FI®), BLUETOOTH® networks, microwave transmission networks, as well as other networks relying on radio broadcasts. The network 109 can also include a combination of two or more networks 109. Examples of networks 109 can include the Internet, intranets, extranets, virtual private networks (VPNs), and similar networks.

[0056] The computing environment 103 can include one or more computing devices. For example, the computing devices can be configured to perform computations on behalf of other computing devices or applications. As another example, such computing devices can host and / or provide content to other computing devices in response to requests for content. In various embodiments, the computing environment can include a processor 112, a memory 115, an input / output (IO) interface 118, and / or a network interface 121, in data connection with each other over a bus 124 or over the network 109. Moreover, the computing environment 103 can employ a plurality of computing devices that can be arranged in one or more server banks or computer banks or other arrangements. Such computing devices can be located in a single installation or can be distributed among many different geographical locations. For example, the computing environment 103 can include a plurality of computing devices that together can include a hosted computing resource, a grid computing resource or any other distributed computing arrangement. In some cases, the computing environment 103 can correspond to an elastic computing resource where the allotted capacity of processing, network, storage, or other computing-related resources can vary’ over time.ATTORNEY DOCKET NO.: 37759.0635P1

[0057] The bus 124 can include a circuit for connecting the bus 124, the processor 112, the memory 115. the network interface 121, and the input / output interface 118 to each other and for delivering communication (e.g., a control message and / or data) between the bus 124, the processor 112, the memory 115, the network interface 121, and the input / output interface 118

[0058] The processor 112 can include one or more of a Central Processing Unit (CPU), an Application Processor (AP), and a Communication Processor (CP). The processor 112 can control, for example, at least one of the bus 124, the memory 115, the network interface 121, and the input / output interface 118 of the computing environment 103 and / or can execute an arithmetic operation or data processing for communication. The processing (or controlling) operation of the processor 112 according to various embodiments is described in detail with reference to the following drawings.

[0059] The processor-executable instructions executed by the processor 112 can be stored and / or maintained by the memory 115. The memory 115 can include a volatile and / or non-volatile memory. The memory 115 can comprise random-access memory (RAM), flash memory, solid state or inertial disks, or any combination thereof. The memory 115 can store, for example, a command or data related to at least one of the bus 124, the processor 112, the memory 115, the network interface 121, and the input / output interface 118 of the computing environment 103. As an example, the memory 115 can store a software and / or a program. The program can include, for example, a kernel 124, a middleware 127, an Application Programming Interface (API) 130, and / or an Audiogram Analysis Application (or an “application”) 133, or the like, configured for controlling one or more functions of the server computing device 101 and / or an external device. At least one part of the kernel 124, middleware 127, or API 130 can be referred to as an Operating System (OS). The memory 115 can include a computer-readable recording medium having a program recorded therein to perform the method according to various embodiment by the processor 112.

[0060] The kernel 124 can control or manage, for example, system resources (e.g, the bus 124, the processor 112, the memory 115, etc.) used to execute an operation or function implemented in other programs (e.g., the middleware 127, the API 130, or the audiogram analysis application 133). Further, the kernel 124 can provide an interface capable of controlling or managing the system resources by accessing individual constitutional elements of the computing environment 103 in the middleware 127, the API 130, and / or the audiogram analysis application 133.ATTORNEY DOCKET NO.: 37759.0635P1

[0061] The middleware 127 can perform, for example, a mediation role so that the API 130 or the audiogram analysis application 133 can communicate with the kernel 124 to exchange data. Further, the middleware 127 can handle one or more task requests received from the audiogram analysis application 133 according to a priority. For example, the middleware 127 can assign a priority of using the system resources (e.g, the bus 124, the processor 112, or the memory 115) of the computing environment 103 to at least one of the audiogram analysis application 133. For example, the middleware 127 can process the one or more task requests according to the priority assigned to the at least one of the application programs, and thus can perform scheduling or load balancing on the one or more task requests.

[0062] The Application Programming Interface (API) 130 can include at least one interface or function (e.g, instruction), for example, for file control, window control, video processing, or character control, as an interface capable of controlling a function provided by the audiogram analysis application 133 in the kernel 124 or the middleware 127.

[0063] The audiogram analysis application 133 can include logic (e.g.. hardware, software, firmware, etc.) that can be implemented to perform various analysis of audiograms. Examples of functionality performed by the audiogram analysis application 133 is further described in the description of FIG.2 and FIG. 7.

[0064] The computing environment 103 can include one or more databases. In one example, the one or more databases can comprise one or more relational databases that use structured query language (SQL) for storing and processing data. In another example, the one or more databases can comprise one or more non-relational databases that use non-structured query language (NoSQL) for storing and processing data. One such database can be a data store 136 stored in the memory’ 115. The data store 136 can be representative of a plurality of data stores 136, which can include relational databases or non-relational databases such as object-oriented databases, hierarchical databases, hash tables or similar key-value data stores, as well as other data storage applications or data structures. Moreover, combinations of these databases, data storage applications, and / or data structures may be used together to provide a single, logical, data store 136. The data stored in the data store 136 can be associated with the operation of the various applications or functional entities as previously described. This data can include audiograms 139, population audiogram data 142, and potentially other data.

[0065] The input / output interface 118 can include an interface for delivering an instruction or data input from a user (e.g., an operator of the computing environment 103) orATTORNEY DOCKET NO.: 37759.0635P1 from a different external device (e.g., client device 106 or other computing devices) to the different elements of the computing environment 103. The input / output interface 118 can further include an interface for outputting one or more user interfaces to the user. For example, the input / output interface 118 can comprise a display, such as a touch screen display, and / or one or more physical input interfaces (e.g., keyboard, mouse, etc.) configured to receive user inputs. Further, the input / output interface 118 can output an instruction or data received from one or more elements of the computing environment 103 to one or more external devices (e.g, client device 106 or other computing devices).

[0066] The network interface 121 can establish, for example, communication between the computing environment 103 and one or more external devices (e.g.. client device 106 or other computing devices). For example, the network interface 121 can communicate with the one or more external devices (e.g, the client device 106 or other computing devices) by being connected to the network 109 through wireless communication or wired communication. The network interface 121 can be configured to communicate with the one or more external devices (e.g. the client device 106 or other computing devices) via the network 109 (e.g, Internet, LAN, etc.). In an example, the network interface 121 can be configured to access the network 109 via a wireless communication interface such as a cellular communication protocol. The cellular communication protocol can comprise at least one of Long-Term Evolution (LTE), LTE Advance (LTE-A), Code Division Multiple Access (CDMA), Wideband CDMA (WCDMA), Universal Mobile Telecommunications System (UMTS). Wireless Broadband (WiBro), Global System for Mobile Communications (GSM), and the like. In an example, the wireless communication interface can be configured to use a near-distance communication. The near-distance communication interface can include for example, at least one of Wireless Fidelity (WiFi), Bluetooth, Bluetooth Low Energy (BLE), Near Field Communication (NFC), Global Navigation Satellite System (GNSS), and the like. According to a usage region or a bandwidth or the like, the GNSS can include, for example, at least one of Global Positioning System (GPS), Global Navigation Satellite System (GLONASS), BeiDou Navigation Satellite System (BDS), Galileo, the European global satellite-based navigation system, and the like. Hereinafter, the " GPS and the " GNSS” can be used interchangeably in the present document.

[0067] The client device 106 can be representative of a plurality of client devices that can be coupled to the network 109. The client device 106 can include a processor-based system such as a computer system. Such a computer system can be embodied in the form of aATTORNEY DOCKET NO.: 37759.0635P1 personal computer (e.g., a desktop computer, a laptop computer, or similar device), a mobile computing device (e.g., personal digital assistants, cellular telephones, smartphones, web pads, tablet computer systems, music players, portable game consoles, electronic book readers, and similar devices), media playback devices (e.g., media streaming devices, BluRay® players, Digital Video Disc (DVD) players, set-top boxes, and similar devices), a videogame console, or other devices with like capability. The client device 106 can include one or more displays, such as Liquid Crystal Displays (LCDs), gas plasma-based flat panel displays, Organic Light Emitting Diode (OLED) displays, electrophoretic ink ('‘E-ink”) displays, projectors, or other types of display devices. In some instances, the display can be a component of the client device 106 or can be connected to the client device 106 through a wired or wireless connection.

[0068] The client device 106 can be configured to execute various applications such as a client application 145, or other applications. The client application 145 can be executed in a client device 106 to access network content served up by the computing environment 103, or other servers, thereby rendering a user interface on the display. To this end, the client application 145 can be a file browser, an Internet-connected browser, a dedicated application, or other executable, and the user interface can include a network page, an application screen, or other user mechanism for obtaining user input. The client device 106 can be configured to execute applications beyond the client application 145, such as email applications, social networking applications, word processors, spreadsheets, or other applications. The client application 145 can be utilized to obtain and generate audiograms 139 and send the audiograms 139 to the computing environment 103. The client device 106 may be connected to specialized equipment to measure hearing in a patient. For example, the client device 106 may be connected to audiometers, tympanometers, optoacoustic emission devices, auditory’ brainstem response equipment, and / or soundproof booths to measure hearing in patients.

[0069] The networked environment 100 shown in FIG. 1 may provide several technical benefits for audiogram analysis and hearing change forecasting. The computing environment 103 may enable automated, real-time audiogram analysis that reduces variability arising from manual interpretation by trained professionals. The networked architecture may enable remote access so that clinicians at different locations can access the same analysis capabilities and population audiogram data 142. The data store 136 may enable accumulation of population audiogram data 142 over time, which can improve prediction accuracy as additional patient data is incorporated into the forecasting models. The separation of theATTORNEY DOCKET NO.: 37759.0635P1 client device 106 and the computing environment 103 may enable specialized audiometric equipment to interface with computationally intensive statistical models without requiring local processing power at the point of care, thereby facilitating deployment in clinical settings with limited computational resources.

[0070] Referring next to FIG. 2, shown is a flowchart that provides one example of the operation of a portion of the audiogram analysis application 133 according to various embodiments of the present disclosure. The flowchart of FIG. 2 provides merely an example of the many different types of functional arrangements that can be employed to implement the operation of the depicted portion of the audiogram analysis application 133. As an alternative, the flowchart of FIG. 2 can be viewed as depicting an example of elements of a method implemented within the network environment 100.

[0071] Beginning with block 203, the audiogram analysis application 133 can receive a first audiogram of a patient. In various embodiments, the client device 106 can generate or obtain data to generate a first audiogram (also called a pre-exposure audiogram). An audiogram is a graphical representation of an individual's sensitivity to pure tones, which are single-frequency sounds. Often, an audiogram can depict the individual’s sensitivity to pure tones based on audiometric thresholds at specific tones. An audiometric threshold quantifies the minimum sound level at which a specific tone can be detected by the subject at least 50% of the time. During an audiometric evaluation performed by a client device 106, a professional trained to test a subject’s hearing can present a series of pure tones through earphones or speakers, at varying sound levels (often measured in decibels) and test frequencies (often measured in Hertz), to determine the subject’s audiometric thresholds to sound that is perceived through vibrations in the air (‘‘air-conduction thresholds’’). In at least some embodiments, the trained professional (e.g.. tester, audiologist, etc.) can present tones through vibrations directed toward bones in the subject’s skull (’‘bone-conduction thresholds”). Both air-conduction thresholds and bone-conduction thresholds can be used to diagnose the presence and severity of hearing change. Comparing air-conduction thresholds and bone-conduction thresholds may be way to determine the type of hearing change (e.g.. sensorineural versus conductive impairment, etc.). Audiograms depict the audiometric threshold for a tone of a specific frequency. Typically, an audiogram includes measurements at approximately seven or eight different frequencies. The selection of frequencies for testing is determined by the specific use case. For example, higher frequencies are often emphasized in hearing conservation programs due to their susceptibility to the damaging effects fromATTORNEY DOCKET NO.: 37759.0635P1 ototoxic drugs, noise, and aging, given the physiological characteristics of the cochlea.Audiometric thresholds are recorded as a series of integers. In various embodiments, the thresholds are often recorded in increments of five decibels (e.g., 5, 10, 15, 20, 25 dB, etc.). Higher values indicate poorer hearing sensitivity, as these values represent the minimum sound level required for the subject to detect the tone. Typically, each ear is tested separately, resulting in a set of measurements for both the left and right ears. Although the audiometric threshold could theoretically increase infinitely until the patient detects the tone, safety limits are often observed during the audiometric testing process to prevent potential damage to the subject's hearing, an audiometer (the device frequently used for obtaining these measurements) often includes technical constraints on the maximum sound level it can produce. Excessive sound levels could damage the auditory system. These safety limits can introduce complexities in standardizing audiometric assessment methodologies, as persons who are incredibly hard of hearing might still have some hearing at incredibly high sound levels.

[0072] The client application 145 can be utilized to obtain and generate audiograms 139 and send the audiograms 139 to the computing environment 103. The client device 106 may be connected to specialized equipment to measure hearing in a patient. For example, the client device 106 may be connected to audiometers, tympanometers, optoacoustic emission devices, auditory brain stem response equipment, and / or soundproof booths to measure hearing in patients.

[0073] For calculations, we can identify the first audiogram as audiogram Y0, which measure the sound level for which a patient can hear at a plurality of pitches. For example, the audiogram Y0can comprise left ear and right ear thresholds at 0.5hz, Ihz, 2hz, 3hz, 4hz, 6hz, 8hz, such that Y0has 14 measurements of the sound level for which the patient can hear.

[0074] Next, at block 206, the audiogram analysis application 133 can determine a patient-individualized predicted region. To determine whether the audiogram is consistent with expectations given by the baseline audiogram assuming that the Homeostasis Hypothesis is true, comparisons are made with follow-up audiogram (post-exposure audiogram) to a multivariate Normal distribution calculated from the pre-exposure audiogram and population data. These calculations can take into account age and sex specific population information to better establish a baseline expectation, along with the pre-exposure audiogram, to determine the multivariate Normal distribution. To facilitate clinical applications, the predictive distribution is typically distilled into one or more prediction intervals. A follow-upATTORNEY DOCKET NO.: 37759.0635P1 threshold that lies outside the prediction interval is unexpected and worthy of further consideration, either by clinical referral or even ignoring the result. This decision is left to the trained professional.

[0075] Various embodiments use a pre-exposure baseline audiogram as an unbiased estimate of pure tone sensitivity in the absence of any audiopathic injury caused by the exposure. Having first observed the first audiogram (the pre-exposure audiogram) it's natural to think of the follow-up audiogram as a test of stability within the auditory system that generated the baseline audiogram (z.e. the follow-up audiogram is a test that the Homeostasis Hypothesis is correct). This motivates computing the conditional distribution of the follow-up audiogram given the observed baseline under homeostasis.

[0076] For explanation purposes, we shall assume that there will be a second audiogram received at some point. For calculations, we can identify the second audiogram as audiogram Yt, which measure the sound level for which a patient can hear at a plurality of pitches. For example, the audiogram Ytcan comprise left ear and right ear thresholds at 0.5hz, Ihz, 2hz, 3hz, 4hz, 6hz, 8hz, such that Ythas 14 measurements of the sound level for which the patient can hear. The ordering of ears and frequencies in Y must be consistent for intelligibility. Yo and Ytcorrespond to audiometry at baseline and follow-up.

[0077] We assume that pure tone sensitivity 6 in each ear e and across the frequency spectrum f at baseline time 0 and follow-up time t are Gaussian Processes with covariance functions Ko and Kt, and population gender- and age-specific mean function p(e,f)

[0078] <50(e ) ~(; P( (e ), / 0)

[0080] This model contains the important assumption that the time that passes between baseline and follow-up measurements (i.e. from 0 to t) is less than the amount of time that is required before the population age-specific mean functionchanges. In other words, the model assumes that monitoring occurs over months, during which time normal presbycusis is effectively unmeasurable, and not decades, when many accumulated factors unrelated to the exposure of concern can induce hearing changes.

[0081] The pure tone sensitivity 8(e, f) is measured by the audiogram at test frequencies defined by the clinical protocol. Viewed in this way, the baseline and follow-up audiograms are each an error-susceptible sample from the pure tone sensitivity7processes So(e,f) and 3t(e, j). Moving forward, we use So and 6tto represent the functions So(e, ) and St(e, J) evaluatedATTORNEY DOCKET NO.: 37759.0635P1 in the ears and pure tone frequencies specified by the testing protocol. The process mean is defined the same way.

[0082] By virtue of the Gaussian process model, Yo and Ytare multivariate normal random variables with respective meansand dtand residual covariance matrices ∑0and ∑t

[0083] Yo| 5o~ o, So ) and

[0084] Yt| 5t~y (5t, 2t)

[0085] The Homeostasis Hypothesis states that pure tone sensitivity has not changed, i.e. that dt=do. An important, but often tacit assumption is that variance components, such as £ and K (see below) are assumed constant over the monitoring period, so that any contradiction to the Homeostasis Hypothesis is taken as evidence of audiopathic injury and not to changes model variance components. With these assumptions, the joint distribution of the audiograms at baseline and follow-up time points conditional on do according to the Homeostasis Hypothesis is:

[0086] (hereinafter “Expression 1”)

[0087] We don't know the baseline pure tone sensitivity do, so we integrate Expression 1 with respect to the distribution of do. This gives the unconditional joint distribution of Ytand Yo as:

[0088] (hereinafter " Expression 2”)

[0089] where is a matrix of evaluations of the covariance function K at the frequencies and ears specified by the testing protocol. The conventional squared exponential covariance model between binary ear indicators (0 = left, 1 = right) e and e* and log frequencies and f* is used for this purpose:

[0090] K(e, e*, f, f* = <p2• exp -a • (e - e*)2- / ? • ( / - / *)2),

[0091] which implies between-ear correlation of and between-octave correlation ofWe also assume that is a diagonal matrix with constant diagonal elements G2

[0092] There may be at least two uses of Expression 2 in serial monitoring. First, embodiments can compute the distribution of the difference between the baseline and followup audiograms as:

[0093] (Yt- Y0)~ N(0, 2 • ∑)

[0094] and then develop prediction regions based on this model. Application of this approach is hampered by missing data or NR at either baseline or follow-up, and regressionATTORNEY DOCKET NO.: 37759.0635P1 to the mean is in effect so that unusually large differences are often incorrectly interpreted as evidence of physiological change.

[0095] Instead, we use the pre-exposure baseline audiogram as an unbiased estimate of pure tone sensitivity in the absence of any audiopathic injury caused by the exposure. Having first observed the baseline audiogram, prior to any exposure, it is natural to think of the follow-up audiogram as a test of stability within the auditory system that generated the baseline audiogram (i.e. the follow-up audiogram is a test that the Homeostasis Hypothesis is correct). This motivates computing the conditional distribution of the follow-up audiogram given the observed baseline under Homeostasis. The multivariate Normal model of Ytand Yo implies that the conditional distribution E| Yo is also multivariate Normal with expected value:

[0096] / j. + K •• (To—10 (hereinafter Expression 3)

[0097] and covariance:

[0098] —K * (2 + K)~ • Kf(hereinafter Expression 4).

[0099] Continuing to block 209, the audiogram analysis application 133 can receive a second audiogram of the patient. In various embodiments, the client device 106 can generate or obtain data to generate a second audiogram (also called a post-exposure audiogram). The second audiogram is obtained subsequent to the patient being exposed to an audiopathic factor or undergoing treatment with an auditory damaging medication. The timing of the second audiogram relative to the first audiogram may vary depending on the clinical context and the nature of the exposure. For example, in ototoxicity monitoring programs for patients receiving cisplatin chemotherapy, the second audiogram may be obtained after one or more treatment cycles, such as after four to six weeks of treatment. In workplace noise monitoring programs, the second audiogram may be obtained at regular intervals, such as annually or after a specified period of occupational noise exposure. In monitoring programs for patients receiving aminoglycoside antibiotics, the second audiogram may be obtained during the course of treatment or shortly after completion of the antibiotic regimen. The second audiogram comprises a second plurality of sound levels that correspond to the same plurality of pitches tested in the first audiogram, enabling direct comparison between the baseline and follow-up measurements. The client application 145 can be utilized to obtain and generate audiograms 139 and send the audiograms 139 to the computing environment 103, for which the audiogram analysis application 133 can receive a second audiogram of the patient. In various embodiments, the second audiogram may include thresholds at standard audiometricATTORNEY DOCKET NO.: 37759.0635P1 frequencies as well as extended high frequencies above 8 kHz, which may be particularly sensitive to early ototoxic damage. The second audiogram may also include No Response thresholds at one or more frequencies if the patient's hearing loss exceeds the audiometer's detection limits, and such censored data is handled appropriately by the model as described above. In at least some embodiments, the second audiogram may be obtained using the same audiometric equipment and testing protocols as the first audiogram to minimize variability attributable to differences in measurement methodology.

[0100] Next, at block 212, the audiogram analysis application 133 can determine that the second audiogram does not fall within the patient-individualized predicted region. The goal is to evaluate whether the follow-up audiogram is consistent with expectations given by the baseline audiogram assuming that the Homeostasis Hypothesis is true. This can be performed by the audiogram analysis application 133 by comparing the follow-up audiogram (the postexposure audiogram or the second audiogram) to the multivariate Normal distribution parameterized by Expression 3 and Expression 4, as determined at block 206. To facilitate clinical applications, the predictive distribution is typically distilled into one or more prediction intervals. A follow-up threshold that lies outside the prediction interval is unexpected and worthy of further consideration, either by clinical referral or even ignoring the result. This decision is left to the trained professional.

[0101] Ninety percent pointwise prediction intervals for the follow-up threshold at each frequency are given by Expression 3 ± 1.64 times the square root of the diagonal elements of Expression 4 (a 95% prediction interval substitutes 1.96 for 1.64 and so forth). The result is a vector of lower and upper 90% prediction limits for each ear and test frequency within which each follow-up threshold is predicted to lie. These 90% pointwise prediction intervals are called “pointwise” because they provide 90% prediction intervals for that specific test ear and frequency “point” only. To clarify, a 90% pointwise prediction interval for one ear and test frequency is an interval such that 9 in 10 follow-up thresholds in that ear and test frequency will be within the interval. However, it is typical to monitor multiple frequencies in both ears rather than single frequencies in any one ear. A 90% simultaneous prediction region is one in which 9 in 10 audiograms are completely inside the region.

[0102] In at least some embodiments, a 90% simultaneous prediction region for the entire left and right ear audiogram could be utilized, and not for each ear and frequency individually. 9 in 10 follow-up audiograms should lie entirely within the prediction region if the Homeostasis Hypothesis is correct. Any frequency in either ear with a threshold outsideATTORNEY DOCKET NO.: 37759.0635P1 the interval is cause for concern. The simultaneous prediction region is computed using a “Volume Tube” methodology. The idea is to numerically expand the width of all the pointwise prediction intervals until exactly 90% of the predicted audiograms are completely contained within the adjusted intervals in both ears and at each test frequency. Let ntj, lj, and Uj denote the expected value, upper and lower 90% pointwise prediction limits for thehear-by -frequency combination. We first simulate a large number of audiograms from multivariate Normal parameterized by Expressions 3 and 4. A 90% simultaneous prediction region is found by numerically searching for a constant c > 1 that adjusts the lower and upper prediction limits at each frequency by mj-c(m-lp and mj+c(uj-mj) so that 90% of the simulated audiograms completely lie within the adjusted intervals at all frequencies in both ears.

[0103] The predictive distribution of the follow-up audiogram given its baseline is given by the parameters in Expressions 3 and 4 and requires as inputs the age-sex specific population mean audiogram, the baseline audiogram Yo and estimates of a, <p, a, and. The population mean thresholds for men and women can be taken from ISO 1999-2013. We use these population mean estimates to center the distribution of pure tone sensitivities, though the model allows for considerable variation with respect to the population. Unless there are NR thresholds in the baseline response, the model parameters are easily estimated by maximizing the marginal likelihood in Expression 2. We prefer a Bayesian approach so as to easily propagate uncertainty about the parameter estimates into the predictions. This is done through MCMC evaluation of the joint posterior distribution of the model parameters and using those same MCMC evaluations to compute the predictive distribution of Ytgiven Y«.These predictions are then used in the volume-tube methodology for computing prediction regions for the entire audiogram.

[0104] A pure tone threshold that exceeds the audiometer's test limits is called “No Response” (NR) in audiology and more generally called “Right-Censored” at the detection limit d in statistics. This feature is commonly observed in time-to-event data such as patient survivorship in biomedical research or equipment reliability in manufacturing. There are several approaches to handling NR thresholds in hearing research, such as imputing the NR threshold to d plus 5 dB or some other constant. Another approach is to treat the NR measurement as completely missing. Neither of these approaches is appealing because imputation by adding an arbitrary constant implies an observation (the NR limit + 5 dB) that was never made, which implies greater certainty about pure tone thresholds than the trainedATTORNEY DOCKET NO.: 37759.0635P1 professional can legitimately claim. This will increase the false-referral rate beyond the nominal levels dictated by the monitoring protocol. Conversely, treating the NR measurement as completely missing isn't a valid approach either, since the trained professional knows that the pure tone threshold exceeds the detection limit d. Thresholds that exceed the audiometer detection limit provide valuable information for making accurate inferences about K andso that more accurate predictions about the follow-up audiogram can be made.

[0105] We approach NR thresholds using censored-data models. Expression 1 represents the likelihood ff, ([>, a, and fi conditional on d Without creating additional notation, the Gaussian Process model for 8(e,j) implies that 6 is also a multivariate normal random variable, <5 ~ N(p, K). In the absence of any NR thresholds, we eliminate dependence on <5 by marginalizing the likelihood in Expression I giving Expression 2. Elowever, when there are one or more NR in the audiogram we factor the likelihood into scalar contributions from thresholds that we observe as N(y; S, a2) and into scalar contributions from NR thresholds as S, a2). This latter expression is one minus the Normal cumulative distribution function evaluated at the audiometer's detection limit d.

[0106] There is no closed form integral of this factored likelihood with respect to the distribution of <5 meaning that the simplicity achieved with a complete baseline audiogram is lost. However, we can use MCMC to evaluate the joint distribution of <5 and the parameters o, <p, a, and conditional on the baseline audiogram. Each of these MCMC evaluations generate a predicted follow-up audiogram according to Expressions 3 and 4. The 90% pointwise prediction interval are the 5thand 95thpercentiles of the generated predictions at each frequency and ear. The volume tube methodology is applied to these predictive distributions to achieve 90% prediction regions over the entire audiogram. The result is a shaded region (See FIGs.3-6) that expresses the clinical expectation that 9 in 10 follow-up audiograms will lie completely within the shaded region if the Homeostasis Hypothesis is correct.

[0107] The width of the interval can be changed, depending on the clinical application. Chemotherapy monitoring may demand a very low false referral rate so as not to withhold life-saving anti-cancer therapy. Larger apparent changes are admissible before alerting the trained professional to deleterious side effects of the therapy. A 95% reference interval may be preferable in this instance instead of the 90% intervals used in the above examples.Workplace noise damage monitoring may prefer a higher false-referral rate to avoid financial liability. Smaller threshold changes in the noise exposure context will therefore provoke a response from administrators, so that an 80% reference interval may be preferred. TheseATTORNEY DOCKET NO.: 37759.0635P1 considerations illustrate the relationship between the consequences of a false-referral and the desired width of the reference interval. If false-referrals provoke little harm, then a narrower interval is acceptable, but if the ramifications of a false-referral are serious, then wider reference intervals are desirable.

[0108] Next, at block 215, the audiogram analysis application 133 can provide a risk assessment, a damage assessment, and / or a change assessment. In various embodiments, the audiogram analysis application 133 can provide feedback to the trained professional based on the analysis performed in block 212. A risk assessment can be provided in response to block 212 to determine the risk that any change in hearing observed at the subsequent audiogram is likely to be an alarming change requiring follow up by the medical team. A damage assessment can be provided to demonstrate an amount of damage that has been acquired by the patient or individual from a first audiogram to a follow up audiogram. A change assessment can not only demonstrate damage or risk, but rather demonstrate improvements to hearing observed at the subsequent audiogram, which can be a result of auditory improvement therapies, such as surgeries or auditory improvement medications.

[0109] In various embodiments, the assessment provided by the audiogram analysis application 133 may include a probability of meeting one or more audiogram-based hearing change criteria. For example, the assessment may include a probability of meeting American Speech-Language Hearing Association (ASHA) criteria, Common Terminology Criteria for Adverse Events (CTCAE) criteria (e.g., Grade 1, Grade 2, etc. or other classification criteria. Because the model predicts the complete audiogram in both ears, any traditional audiogram-based criterion may be derived from the predicted audiograms without reference to any other prediction algorithms or approaches. In some embodiments, the MCMC algorithm draws repeatedly from the joint posterior distribution of the model parameters, and each posterior draw generates a predicted follow-up audiogram for a given baseline audiogram, patient age, and treatment history. Predicted changes to the audiogram, pure tone averages, or any other transformation of the audiogram may be computed directly from the observed baseline and predicted follow-up audiograms. The fraction of the posterior predicted audiograms that yield a positive outcome for a given criterion (e.g, ASHA shift or CTCAE Grade 1) may be used as an estimate of the probability of that event at follow-up. In this way, the model may simultaneously output results in formats appropriate for multiple end users, including oncologists, audiologists, patients, and caregivers.ATTORNEY DOCKET NO.: 37759.0635P1

[0110] Finally, at block 218, the audiogram analysis application 133 can provide a treatment recommendation. Treatment changes may be advised by the relevant stakeholders. In various embodiments, the audiogram analysis application 133 can provide one or more treatment recommendations to the client device 106. In at least some embodiments, the treatment recommendation includes a reduction in a dosage of the auditory damaging medication for the patient, such as cisplatin or IV Tobramycin. The amount of dosage can be tailored to the individual patient's condition to balance the negative effects and the beneficial purposes of taking the medication. In various embodiments, the treatment recommendation may include modifications to the timing or frequency of medication administration, such as extending the interval between infusions to allow for potential recovery of auditory function or to reduce cumulative ototoxic exposure. The audiogram analysis application 133 may also recommend substitution of an alternative medication with a lower ototoxic profile when clinically appropriate, enabling the oncology or infectious disease team to consider therapeutic alternatives that may achieve similar efficacy with reduced risk of hearing damage. In some embodiments, the treatment recommendation may include enhanced monitoring protocols, such as more frequent audiometric evaluations or the addition of extended high-frequency testing to detect early signs of ototoxic damage before it progresses to frequencies critical for speech understanding. The audiogram analysis application 133 may further recommend referral to audiology services for baseline counseling, hearing aid evaluation, or cochlear implant candidacy assessment when the predicted hearing change indicates a high likelihood of significant functional impairment. In various embodiments, the treatment recommendation may be formatted differently for different end users, with oncologists receiving information focused on treatment modifications and risk-benefit considerations, audiologists receiving detailed frequency-specific predictions and monitoring recommendations, and patients receiving simplified summaries of expected outcomes along with counseling resources and communication strategies. The audiogram analysis application 133 may also recommend protective interventions, such as the administration of otoprotective agents when available, or the use of hearing protection devices during concurrent noise exposure. In at least some embodiments, the treatment recommendation may include coordination w ith other members of the care team, such as speech-language pathologists for communication strategy development or social workers for psychosocial support related to anticipated hearing changes.ATTORNEY DOCKET NO.: 37759.0635P1

[0111] The method shown in FIG. 2 may provide several technical benefits for serial monitoring of hearing changes. The method may provide an automated, unbiased approach to serial monitoring that overcomes regression to the mean bias inherent in conventional difference-based approaches that simply compute the difference between baseline and followup audiograms. The use of patient-individualized predicted regions conditioned on the observed baseline audiogram may provide statistically rigorous expectations for follow-up measurements that account for the specific hearing profile of each patient. The method may handle censored and missing data through censored-data models rather than arbitrary imputation or discarding valuable information, thereby providing more accurate predictions when thresholds exceed audiometer detection limits. The method may simultaneously serve multiple clinical objectives and stakeholder priorities through a single unified statistical framework, enabling oncologists, audiologists, patients, and caregivers to each receive results in formats appropriate for their needs.

[0112] Continuing to FIG. 3, show n is a diagram that illustrates model results of a case study of utilizing methods described by FIG. 2. Given the patient's baseline audiogram as an input, we compute the predictive distribution of the follow-up audiogram under the assumption that the Homeostasis Hypothesis is correct. Correlations between thresholds in each ear and at neighboring frequencies, as well as population-based pure-tone threshold data are used to narrow the patient-specific predictive distribution. This predictive distribution is expressed as a simultaneous prediction region, which can be easily interpreted: 9 out of 10 follow-up audiograms on the patient will be entirely within the region if the Homeostasis Hypothesis is correct. A follow-up audiogram that exceeds the audiogram prediction region at any frequency in either ear is evidence that the Homeostasis Hypothesis is false and that pure tone sensitivity has changed.

[0113] Specifically, FIG.3 relates to serial monitoring an adult male, 77 years old, being treated for Oropharyngeal cancer with 40 mg / m2, w eekly of cisplatin. Solid lines show the pre-treatment baseline audiogram (top row ) and 6-w eek follow-up (bottom row ) after four infusions of cisplatin equaling a cumulative dose 200 mg / m2. The dashed line is the ISO 1999-2013 (ISO, 2013) age-sex specific mean threshold obtained from population-based pure-tone threshold data. The “X” indicates No Response at that test frequency. The shaded area is a 90% prediction region for this patient's audiogram under the assumption that pure tone sensitivity hasn't changed from their baseline test. Follow-up thresholds exceeding the prediction region (circled) are “significant” changes to the audiogram that warrant follow-up.ATTORNEY DOCKET NO.: 37759.0635P1

[0114] FIGs. 4-6 illustrate model results in the context of additional case studies following the format of FIG. 3. The prediction region shown for these patients were generated by inputting the baseline audiogram thresholds, age, and sex into Expressions 3 and 4, and following the Volume Tube methodology (See block 206 of FIG. 2). FIG. 4 shows results for a patient with Cystic Fibrosis who was treated with IV Tobramycin for a bacterial lung infection. FIG. 5 shows results for a patient with cancer who was treated with cisplatin, and FIG.6 shows results for an individual exposed to workplace noise over a five-year period. Note that this subject did not provide baseline 3 and 6 kHz thresholds, though the model structure still allows predictions at these frequencies.

[0115] Regarding FIG. 4, depicted are model results of serial monitoring of a 42 year-old male with a history of pancreatic insufficiency and diabetes mellitus related to his Cystic Fibrosis. The patient presented to emergency unit with a pulmonary infection and was admitted to the hospital where they received between 6.75 to 12.5 ml of IV Tobramycin. Visit A: baseline audiogram. Visit B: audiogram taken 1 week after the baseline exam during their course on IV Tobramycin therapy as an inpatient. Visit C: first follow-up visit conducted about 1 month after end of IV course of treatment. X = no response, NR. Circled thresholds are significant changes.

[0116] Regarding FIG. 5, depicted are model results of serial monitoring of a 66 year old patient with a tumor of the larynx who received cisplatin-based chemotherapy prior to surgical removal. The patient received five infusions of cisplatin for a total of 240 mg / m2. X = no response, NR. Circled thresholds are significant changes.

[0117] Regarding FIG. 6, depicted are model results of serial monitoring of a 47 year old patient who worked as a chemical engineer and reported exposure to loud machine noise. His initial audiogram was completed at age 47, with a follow up occurring at age 52. Baseline thresholds are missing at 3 and 6 kHz. 3 kHz thresholds are missing at five-year follow-up. X = no response, NR. Circled thresholds are significant changes.

[0118] Referring next to FIG. 7, shown is a flowchart that provides one example of the operation of a portion of the audiogram analysis application 133 according to various embodiments of the present disclosure. The flowchart of FIG. 7 provides merely an example of the many different types of functional arrangements that can be employed to implement the operation of the depicted portion of the audiogram analysis application 133. As an alternative, the flowchart of FIG. 7 can be viewed as depicting an example of elements of a method implemented within the network environment 100.ATTORNEY DOCKET NO.: 37759.0635P1

[0119] Beginning with block 703. the audiogram analysis application 133 can receive a first audiogram. In various embodiments, the client device 106 can generate or obtain data to generate a first audiogram (also called a pre-exposure audiogram). An audiogram is a graphical representation of an individual's sensitivity to pure tones, which are single-frequency sounds. Often, an audiogram can depict the individual's sensitivity to pure tones based on audiometric thresholds at specific tones. An audiometric threshold quantifies the minimum sound level at which a specific tone can be detected by the subject at least 50% of the time. During an audiometric evaluation performed by a client device 106, a professional trained to test a subject’s hearing can present a series of pure tones through earphones or speakers, at varying sound levels (often measured in decibels) and test frequencies (often measured in hertz), to determine the subject's audiometric thresholds to sound that is perceived through vibrations in the air (‘’air-conduction thresholds”). In at least some embodiments, the trained professional (e.g., tester, audiologist, etc.) can present tones through vibrations directed toward bones in the subject's skull (“bone-conduction thresholds”). Both air-conduction thresholds and bone-conduction thresholds can be used to diagnose the presence and severity of hearing change. Comparing air-conduction thresholds and bone-conduction thresholds may be way to determine the type of hearing change (e.g., sensorineural versus conductive impairment, etc.). Audiograms depict the audiometric threshold for a tone of a specific frequency. Typically, an audiogram includes measurements at approximately seven or eight different frequencies. The selection of frequencies for testing is determined by the specific use case. For example, higher frequencies are often emphasized in hearing conservation programs due to their susceptibility to the damaging effects from ototoxic drugs, noise, and aging, given the physiological characteristics of the cochlea.Audiometric thresholds are recorded as a series of integers. In various embodiments, the thresholds are often recorded in increments of five decibels (e.g, 5, 10, 15, 20, 25 dB, etc.). Higher values indicate poorer hearing sensitivity, as these values represent the minimum sound level required for the subject to detect the tone. Typically, each ear is tested separately, resulting in a set of measurements for both the left and right ears. Although the audiometric threshold could theoretically increase infinitely until the patient detects the tone, safety limits are often observed during the audiometric testing process to prevent potential damage to the subject's hearing. An audiometer (the device frequently used for obtaining these measurements) often includes technical constraints on the maximum sound level it can produce. Excessive sound levels could damage the auditory system. These safety limits canATTORNEY DOCKET NO.: 37759.0635P1 introduce complexities in standardizing audiometric assessment methodologies, as persons who are incredibly hard of hearing might still have some hearing at incredibly high sound levels.

[0120] The client application 145 can be utilized to obtain and generate audiograms 139 and send the audiograms 139 to the computing environment 103. The client device 106 may be connected to specialized equipment to measure hearing in a patient. For example, the client device 106 may be connected to audiometers, which are electronic devices used to present calibrated pure tones at specific frequencies and intensities to determine a patient's hearing thresholds. The client device 106 may also be connected to tympanometers, which measure the mobility of the tympanic membrane and middle ear system by varying air pressure in the ear canal to assess middle ear function and identify conditions such as fluid accumulation or eustachian tube dysfunction. Additionally, the client device 106 may be connected to otoacoustic emission devices, which detect sounds generated by the outer hair cells of the cochlea in response to acoustic stimulation, providing an objective measure of cochlear function that does not require patient response. The client device 106 may further be connected to auditory brainstem response equipment, which measures electrical activity along the auditory pathway from the cochlea through the brainstem in response to acoustic stimuli, enabling objective assessment of hearing sensitivity and neural integrity without requiring behavioral responses from the patient. The client device 106 may also interface with soundproof booths or sound-treated rooms that provide acoustically isolated environments to ensure accurate threshold measurements by minimizing ambient noise interference during audiometric testing. In various embodiments, the client device 106 may receive data directly from these specialized devices through wired connections (e.g, USB, serial ports, etc.) or wireless connections (e.g., Bluetooth. WiFi, etc.), enabling automated capture and transmission of audiometric data to the computing environment 103 for analysis by the audiogram analysis application 133.

[0121] Next, at block 706, the audiogram analysis application 133 can receive data corresponding to an audiopathic damaging factor or an interventional factor that could improve auditory function. To identify hearing change caused by various auditory function change factors, the audiogram analysis application 133 could receive comprehensive data on noise exposure, medication use, and other relevant factors. For instance, it w ould track the decibel levels, duration, and frequency of loud noise exposure, as w ell as the dosage, frequency, and duration of hearing-damaging medications. Additionally, the applicationATTORNEY DOCKET NO.: 37759.0635P1 would consider the type of medication, as certain drugs like specific antibiotics and chemotherapy agents are known to cause hearing damage. This data, combined with an individual's medical history, including any previous hearing issues and family history of hearing loss, would provide a robust foundation for analysis. Environmental and lifestyle factors also play a crucial role in assessing hearing change risk. Occupational exposure to high noise levels, such as in construction or manufacturing jobs, and participation in noisy recreational activities, like attending concerts or visiting shooting ranges, are significant contributors. The audiogram analysis application 133 could also receive audiometric data from hearing tests and speech recognition scores to evaluate the individual's hearing ability. Furthermore, lifestyle choices, such as the use of hearing protection and habits like smoking and alcohol consumption, would be factored in, as they can influence hearing health.Additionally, surgical, therapeutic, or rehabilitative interventions may improve a patient’s hearing.

[0122] Continuing to block 709, the audiogram analysis application 133 can forecast a predicted hearing change based on the first audiogram and the data. Utilizing the data gathered from testing various patients using the methods described in FIG. 2, a predictive model can be generated based on the population of patients (such as patients described in FIGs. 3-6) having similar qualities to the present patient, which can be identified based on the data corresponding to the auditory function change factor. These calculations can take into account age and sex specific population information to better establish a baseline expectation, along with the pre-exposure audiogram, to determine the predicted hearing change. In various embodiments, the forecasting model may be implemented as a Bayesian multivariate linear mixed model that employs a Gaussian process framework. The model may use machine learning techniques to efficiently predict probability distributions for pure-tone thresholds at multiple frequencies in both ears. By assuming the output values are drawn from a multivariate Gaussian distribution, the model may' provide a statistically-based method for forecasting hearing changes that accounts for correlations between thresholds at neighboring frequencies and between ears. The forecasting process may involve matrix manipulations including matrix multiplications and inversions using mathematical parameters estimated from a large population of patients who have received similar treatments or exposures. For example, when forecasting hearing changes for a patient scheduled to receive cisplatin chemotherapy, the model may incorporate the patient's baseline audiogram, the prescribed dosage and timing of infusions, the presence or absence of concurrent radiationATTORNEY DOCKET NO.: 37759.0635P1 therapy, and patient-specific features such as age. The model may generate predictions at multiple future timepoints during the course of treatment, providing the clinical care team with a trajectory of expected hearing changes. In some embodiments, the prediction regions may widen and shift toward higher threshold values as the cumulative drug dose increases, reflecting the progressive ototoxic effect of the treatment. The relationship between drug exposure and changes in hearing may be determined through regression analysis of existing data from prior patients, allowing for accurate forecasts that are specific to the individual patient's circumstances.

[0123] Next at block 712, the audiogram analysis application 133 can provide a risk assessment, a damage assessment, and / or a change assessment based on the predicted hearing change. In various embodiments, the audiogram analysis application 133 can provide feedback to the trained professional based on the analysis performed in block 709. A risk assessment can be provided in response to block 709 to determine a potential risk that any change in hearing will occur as a result of exposure to the hearing affecting factor. The risk assessment may quantify the probability that the patient will experience clinically significant hearing changes at specific frequencies or across the entire audiogram following the planned exposure or treatment. In some embodiments, the risk assessment may include probabilities of meeting specific hearing change criteria, such as American Speech-Language Hearing Association (ASHA) criteria or Common Terminology Criteria for Adverse Events (CTCAE) criteria at various grades. The risk assessment may be computed by determining the fraction of posterior predicted audiograms that yield a positive outcome for a given criterion, thereby providing the clinical care team with actionable probability estimates for treatment planning purposes. A damage assessment can be provided to demonstrate an amount of damage that can be expected by the patient or individual as a result of hearing affecting factor. The damage assessment may include predicted threshold shifts at each test frequency in both ears, expressed as the expected change in decibels from the baseline audiogram. In various embodiments, the damage assessment may provide upper and low er bounds representing best-case and worst-case scenarios for hearing loss, allowing the clinical care team to understand the range of possible outcomes. The damage assessment may also include predictions of changes to pure tone averages or other summary metrics commonly used in clinical practice to characterize overall hearing ability. In some embodiments, the damage assessment may be presented as a predicted follow-up audiogram with associated prediction intervals, enabling visualization of the expected hearing profile following exposure. A changeATTORNEY DOCKET NO.: 37759.0635P1 assessment can not only demonstrate damage or risk, but rather demonstrate improvements to hearing as a result of the hearing affecting factor, which can be a result of auditory improvement therapies, such as surgeries or auditory improvement medications. The change assessment may quantify the expected magnitude of improvement at each test frequency and provide probability estimates for achieving clinically meaningful improvements in hearing thresholds. In various embodiments, the change assessment may compare predicted postintervention audiograms against pre-intervention baselines to estimate the likelihood that the intervention will achieve its therapeutic goals. The assessments provided at block 712 may be formatted for different end users, including oncologists who may prioritize information about treatment toxicity, audiologists who may focus on detailed frequency-specific predictions, and patients who may benefit from simplified summaries of expected hearing outcomes.

[0124] Next, at block 715, the audiogram analysis application 133 can provide a treatment recommendation based on the predicted hearing change and / or the risk assessment. Treatment changes may be advised by the relevant stakeholders. In various embodiments, the audiogram analysis application 133 can provide one or more treatment recommendations to the client device 106. In at least some embodiments, the treatment recommendation includes a reduction in a dosage of the auditory damaging medication for the patient, such as cisplatin or IV Tobramycin. The amount of dosage can be tailored to the individual patient's condition to balance the negative effects and the beneficial purposes of taking the medication. In various embodiments, the treatment recommendation may include modifications to the timing or frequency of medication administration, such as extending the interval between infusions to allow for potential recovery of auditory function or to reduce cumulative ototoxic exposure. The audiogram analysis application 133 may also recommend substitution of an alternative medication with a lower ototoxic profile when clinically appropriate, enabling the oncology or infectious disease team to consider therapeutic alternatives that may achieve similar efficacy with reduced risk of hearing damage. In some embodiments, the treatment recommendation may include enhanced monitoring protocols, such as more frequent audiometric evaluations or the addition of extended high-frequency testing to detect early signs of ototoxic damage before it progresses to frequencies critical for speech understanding. The audiogram analysis application 133 may further recommend referral to audiology services for baseline counseling, hearing aid evaluation, or cochlear implant candidacy assessment when the predicted hearing change indicates a high likelihood of significant functional impairment. In various embodiments, the treatment recommendation may beATTORNEY DOCKET NO.: 37759.0635P1 formatted differently for different end users, with oncologists receiving information focused on treatment modifications and risk-benefit considerations, audiologists receiving detailed frequency-specific predictions and monitoring recommendations, and patients receiving simplified summaries of expected outcomes along with counseling resources and communication strategies. The audiogram analysis application 133 may also recommend protective interventions, such as the administration of otoprotective agents when available, or the use of hearing protection devices during concurrent noise exposure. In at least some embodiments, the treatment recommendation may include coordination with other members of the care team, such as speech-language pathologists for communication strategy development or social workers for psychosocial support related to anticipated hearing changes.

[0125] In various embodiments, the forecasting model may be developed using training data and validated with independent data. For example, in at least one embodiment, the model was developed using training data collected from a plurality of patients over a period of years and validated with data from an independent group of patients. Forecasted thresholds may achieve root mean squared prediction errors within the precision of the audiometric measurement. In at least one embodiment, root mean squared prediction errors of 4.3 to 8.5 dB were achieved across frequencies in training data, and 5.2 to 10.6 dB across frequencies in validation data. The model may accurately classify ears as having changed hearing, with area under the receiver operating characteristic curve (AUC) values of 0.78 for both ASHA and CTCAE Grade 1 predictions in validation data. These performance metrics may demonstrate that the forecasting model provides predictions within the precision of the pure-tone threshold measurement itself, supporting the clinical use of the model for personalized pre-treatment counseling and guiding timely interventions for patients receiving ototoxic medications.

[0126] The forecasting method shown in FIG. 7 may provide several technical benefits for predicting hearing changes before they occur. The method may provide patient-individualized predictions of hearing changes before they occur, enabling proactive rather than reactive clinical intervention. The Bayesian multivariate linear mixed model may generate complete predicted audiograms across all test frequencies in both ears rather than predicting a single criterion outcome such as a binary determination of hearing loss or a simplified metric such as a pure tone average. The model may simultaneously output multiple classification criteria (e.g., ASHA, CTCAE Grade 1, CTCAE Grade 2) from a single predictive distribution without requiring separate prediction algorithms for each criterion.ATTORNEY DOCKET NO.: 37759.0635P1 The model may be validated with independent data, achieving prediction errors within the precision of the audiometric measurement itself, thereby supporting the clinical use of the model for personalized pre-treatment counseling, and guiding timely interventions for patients receiving ototoxic medications.

[0127] A number of implementations have been described. Nevertheless, it will be understood that various modifications may be made without departing from the spirit and scope of the disclosure. Accordingly, other implementations are within the scope of the following claims.

[0128] Various embodiments of the present disclosure are further detailed by the following clauses of this detailed description:

[0129] Clause 1 describes a method of receiving a first audiogram of a patient comprising a first plurality of sound levels that correspond to a plurality’ of pitches, each pitch of the plurality of pitches having a unique frequency; determining, based on the first plurality of sound levels that correspond to the plurality' of pitches and population-based hearing data, a patient-individualized predicted region representing expected shifts for a subsequent audiogram; receiving, subsequent to the patient being treated with an auditory' damaging treatment, a second audiogram of the patient comprising a second plurality of sound levels that correspond to the plurality of pitches; and providing, based on determining that at least one sound level of the second plurality' of sound levels that correspond to the plurality of pitches is not w ithin the patient-individualized predicted region, a treatment recommendation.

[0130] Clause 2 describes the method of clause 1, w herein the auditory' damaging treatment is an auditory' damaging medication and the treatment recommendation includes a reduction in a dosage of the auditory damaging medication for the patient.

[0131] Clause 3 describes the method of clause 2, wherein the auditory’ damaging medication includes cisplatin.

[0132] Clause 4 describes the method of any one of clauses 1-3, wherein at least one sound level of the second plurality of sound levels is greater than a corresponding sound level of the first plurality' of sound levels.

[0133] Clause 5 describes the method of any one of clauses 1-4, wherein determining the patient-individualized predicted region is further based on a multivariate normal model.ATTORNEY DOCKET NO.: 37759.0635P1

[0134] Clause 6 describes the method of any one of clauses 1-5, wherein the first plurality of sound levels of the first audiogram includes a first ear plurality of sound levels and a second ear plurality of sound levels, wherein each of the first ear plurality of sound levels and the second ear plurality of sound levels are tested on the plurality of pitches.

[0135] Clause 7 describes the method of any one of clauses 1-6, wherein the patient-individualized predicted region is a simultaneous prediction region computed using a volume tube methodology.

[0136] Clause 8 describes the method of any one of clauses 1-7, further comprising handling No Response thresholds using censored-data models.

[0137] Clause 9 describes a method of receiving a first audiogram of a patient; determining, based on the first audiogram, a patient-individualized predicted region; receiving, subsequent to the patient experiencing an audiopathic changing factor, a second audiogram; and providing, based on a determination that at least a portion of the second audiogram does not fall within the patient-individualized predicted region, an assessment representing a likelihood of change in auditory’ function.

[0138] Clause 10 describes the method of clause 9, wherein the audiopathic changing factor is workplace noise.

[0139] Clause 11 describes the method of clauses 9 or 10, wherein the audiopathic changing factor is an audiopathic improving surgery’.

[0140] Clause 12 describes the method of any one of clauses 9-11, wherein the patient-individualized predicted region represents a width of an interval around the first audiogram of the patient.

[0141] Clause 13 describes the method of clause 12, wherein the interval represents the percentile of patients within a population for which a homeostasis hypothesis is correct.

[0142] Clause 14 describes the method of any one of clauses 9-13, wherein the assessment includes a probability of meeting American Speech-Language Hearing Association (ASHA) criteria or Common Terminology Criteria for Adverse Events (CTCAE) criteria.

[0143] Clause 15 describes a method of receiving a first audiogram of a patient comprising a first plurality of sound levels that correspond to a plurality of pitches, each pitchATTORNEY DOCKET NO.: 37759.0635P1 of the plurality of pitches having a unique frequency; receiving data corresponding to an auditory function changing factor; forecasting, based on the first audiogram and the data corresponding to the auditory function changing factor, a predicted hearing change; and providing, based on the predicted hearing change, a treatment recommendation.

[0144] Clause 16 describes the method of clause 15, wherein the data corresponding to the auditory function changing factor includes at least a dosage of an auditory affecting medication.

[0145] Clause 17 describes the method of clause 16, wherein the auditory affecting medication is cisplatin.

[0146] Clause 18 describes the method of any one of clauses 15-17, wherein the treatment recommendation is to reduce a dosage of an auditory affecting medication.

[0147] Clause 19 describes the method of any one of clauses 15-18, wherein the auditory function changing factor includes concurrent radiation to the cochlea.

[0148] Clause 20 describes the method of any one of clauses 15-19, wherein forecasting includes computing a predictive distribution using Markov Chain Monte Carlo (MCMC) evaluation.

Claims

ATTORNEY DOCKET NO.: 37759.0635P1CLAIMSWhat is claimed is:

1. A method, comprising:receiving a first audiogram of a patient comprising a first plurality of sound levels that correspond to a plurality of pitches, each pitch of the plurality of pitches having a unique frequency;determining, based on the first plurality of sound levels that correspond to the plurality' of pitches and population-based hearing data, a patient-individualized predicted region representing expected shifts for a subsequent audiogram: receiving, subsequent to the patient being treated with an auditory damaging treatment, a second audiogram of the patient comprising a second plurality of sound levels that correspond to the plurality' of pitches; andproviding, based on determining that at least one sound level of the second plurality’ of sound levels that correspond to the plurality of pitches is not within the patient-individualized predicted region, a treatment recommendation.

2. The method of claim 1, wherein the auditory damaging treatment is an auditory damaging medication and the treatment recommendation includes a reduction in a dosage of the auditory damaging medication for the patient.

3. The method of claim 2, wherein the auditory’ damaging medication includes cisplatin.

4. The method of claim 1, wherein at least one sound level of the second plurality of sound levels is greater than a corresponding sound level of the first plurality of sound levels.

5. The method of claim 1, wherein determining the patient-individualized predicted region is further based on a multivariate normal model.

6. The method of claim 1, wherein the first plurality of sound levels of the first audiogram includes a first ear plurality' of sound levels and a second ear plurality of sound levels, wherein each of the first ear plurality7of sound levels and the second ear plurality of sound levels are tested on the plurality of pitches.ATTORNEY DOCKET NO.: 37759.0635P1 7. The method of claim 1, wherein the patient-individualized predicted region is a simultaneous prediction region computed using a volume tube methodology.

8. The method of claim 1, wherein the method further comprises handling No Response thresholds using censored-data models.

9. A method, comprising:receiving a first audiogram of a patient;determining, based on the first audiogram, a patient-individualized predicted region; receiving, subsequent to the patient experiencing an audiopathic changing factor, a second audiogram; andproviding, based on a determination that at least a portion of the second audiogram does not fall within the patient-individualized predicted region, an assessment representing a likelihood of change in auditory function.

10. The method of claim 9, wherein the audiopathic changing factor is workplace noise.

11. The method of claim 9, wherein the audiopathic changing factor is an audiopathic improving surgery.

12. The method of claim 9, wherein the patient-individualized predicted region represents a width of an interval around the first audiogram of the patient.

13. The method of claim 12, wherein the interval represents the percentile of patients within a population for which a homeostasis hypothesis is correct.

14. The method of claim 9, wherein the assessment includes a probability' of meeting American Speech-Language Hearing Association (ASHA) criteria or Common Terminology Criteria for Adverse Events (CTCAE) criteria.

15. A method, comprising:receiving a first audiogram of a patient comprising a first plurality’ of sound levels that correspond to a plurality of pitches, each pitch of the plurality of pitches having a unique frequency;receiving data corresponding to an auditory’ function changing factor;ATTORNEY DOCKET NO.: 37759.0635P1 forecasting, based on the first audiogram and the data corresponding to the auditory function changing factor, a predicted hearing change; andproviding, based on the predicted hearing change, a treatment recommendation.

16. The method of claim 15, wherein the data corresponding to the auditor}’ function changing factor includes at least a dosage of an auditory affecting medication.

17. The method of claim 16, wherein the auditory affecting medication is cisplatin.

18. The method of claim 15, wherein the treatment recommendation is to reduce a dosage of an auditory affecting medication.

19. The method of claim 15, wherein the auditory’ function changing factor includes concurrent radiation to the cochlea.

20. The method of claim 15, wherein forecasting includes computing a predictive distribution using Markov Chain Monte Carlo (MCMC) evaluation.

Images