Active transfer learning for audiogram estimation.

Computational audiology (CA) has grown over the last few years with the improvement of computing power and the growth of machine learning (ML) models. There are today several audiogram databases which have been used to improve the accuracy of CA models as well as reduce testing time and diagnostic complexity. However, these CA models have mainly been trained on single populations. This study integrated contextual and prior knowledge from audiogram databases of multiple populations as informative priors to estimate audiograms more precisely using two mechanisms: (1) a mapping function drawn from feature-based homogeneous Transfer Learning (TL) also known as Domain Adaptation (DA) and (2) Active Learning (Uncertainty Sampling) using a stream-based query mechanism. Simulations of the Active Transfer Learning (ATL) model were tested against a traditional adaptive staircase method akin to the Hughson-Westlake (HW) method for the left ear at frequencies ω=0.25,0.5,1,2,4,8 kHz, resulting in accuracy and reliability improvements. ATL improved HW tests from a mean of 41.3 sound stimuli presentations and reliability of ±9.02 dB down to 25.3±1.04 dB. Integrating multiple databases also resulted in classifying the audiograms into 18 phenotypes, which means that with increasing data-driven CA, higher precision is achievable, and a possible re-conceptualisation of the notion of phenotype classifications might be required. The study contributes to CA in identifying an ATL mechanism to leverage existing audiogram databases and CA models across different population groups. Further studies can be done for other psychophysical phenomena using ATL.

Contrast response function estimation with nonparametric Bayesian active learning.

Multidimensional psychometric functions can typically be estimated nonparametrically for greater accuracy or parametrically for greater efficiency. By recasting the estimation problem from regression to classification, however, powerful machine learning tools can be leveraged to provide an adjustable balance between accuracy and efficiency. Contrast sensitivity functions (CSFs) are behaviorally estimated curves that provide insight into both peripheral and central visual function. Because estimation can be impractically long, current clinical workflows must make compromises such as limited sampling across spatial frequency or strong assumptions on CSF shape. This article describes the development of the machine learning contrast response function (MLCRF) estimator, which quantifies the expected probability of success in performing a contrast detection or discrimination task. A machine learning CSF can then be derived from the MLCRF. Using simulated eyes created from canonical CSF curves and actual human contrast response data, the accuracy and efficiency of the machine learning contrast sensitivity function (MLCSF) was evaluated to determine its potential utility for research and clinical applications. With stimuli selected randomly, the MLCSF estimator converged slowly toward ground truth. With optimal stimulus selection via Bayesian active learning, convergence was nearly an order of magnitude faster, requiring only tens of stimuli to achieve reasonable estimates. Inclusion of an informative prior provided no consistent advantage to the estimator as configured. MLCSF achieved efficiencies on par with quickCSF, a conventional parametric estimator, but with systematically higher accuracy. Because MLCSF design allows accuracy to be traded off against efficiency, it should be explored further to uncover its full potential.

Cognition and Brain System Segregation in Pediatric Brain Tumor Patients Treated with Proton Therapy.

Purpose: Pediatric brain tumor patients often experience significant cognitive sequelae. Resting-state functional MRI (rsfMRI) provides a measure of brain network organization, and we hypothesize that pediatric brain tumor patients treated with proton therapy will demonstrate abnormal brain network architecture related to cognitive outcome and radiation dosimetry. Participants and Methods: Pediatric brain tumor patients treated with proton therapy were enrolled on a prospective study of cognitive assessment using the NIH Toolbox Cognitive Domain. rsfMRI was obtained in participants able to complete unsedated MRI. Brain system segregation (BSS), a measure of brain network architecture, was calculated for the whole brain, the high-level cognition association systems, and the sensory-motor systems. Results: Twenty-six participants were enrolled in the study for cognitive assessment, and 18 completed rsfMRI. There were baseline cognitive deficits in attention and inhibition and processing speed prior to radiation with worsening performance over time in multiple domains. Average BSS across the whole brain was significantly decreased in participants compared with healthy controls (1.089 and 1.101, respectively; P = 0.001). Average segregation of association systems was significantly lower in participants than in controls (P < 0.001) while there was no difference in the sensory motor networks (P = 0.70). Right hippocampus dose was associated with worse attention and inhibition (P < 0.05) and decreased segregation in the dorsal attention network (P < 0.05). Conclusion: Higher mean dose to the right hippocampus correlated with worse dorsal attention network segregation and worse attention and inhibition cognitive performance. Patients demonstrated alterations in brain network organization of association systems measured with rsfMRI; however, somatosensory system segregation was no different from healthy children. Further work with preradiation rsfMRI is needed to assess the effects of surgery and presence of a tumor on brain network architecture.

Cognitive deficits and altered functional brain network organization in pediatric brain tumor patients.

Survivors of pediatric brain tumors experience significant cognitive deficits from their diagnosis and treatment. The exact mechanisms of cognitive injury are poorly understood, and validated predictors of long-term cognitive outcome are lacking. Resting state functional magnetic resonance imaging allows for the study of the spontaneous fluctuations in bulk neural activity, providing insight into brain organization and function. Here, we evaluated cognitive performance and functional network architecture in pediatric brain tumor patients. Forty-nine patients (7-18 years old) with a primary brain tumor diagnosis underwent resting state imaging during regularly scheduled clinical visits. All patients were tested with a battery of cognitive assessments. Extant data from 139 typically developing children were used as controls. We found that obtaining high-quality imaging data during routine clinical scanning was feasible. Functional network organization was significantly altered in patients, with the largest disruptions observed in patients who received propofol sedation. Awake patients demonstrated significant decreases in association network segregation compared to controls. Interestingly, there was no difference in the segregation of sensorimotor networks. With a median follow-up of 3.1 years, patients demonstrated cognitive deficits in multiple domains of executive function. Finally, there was a weak correlation between decreased default mode network segregation and poor picture vocabulary score. Future work with longer follow-up, longitudinal analyses, and a larger cohort will provide further insight into this potential predictor.

Unicorn, Hare, or Tortoise? Using Machine Learning to Predict Working Memory Training Performance.

People differ considerably in the extent to which they benefit from working memory (WM) training. Although there is increasing research focusing on individual differences associated with WM training outcomes, we still lack an understanding of which specific individual differences, and in what combination, contribute to inter-individual variations in training trajectories. In the current study, 568 undergraduates completed one of several N-back intervention variants over the course of two weeks. Participants' training trajectories were clustered into three distinct training patterns (high performers, intermediate performers, and low performers). We applied machine-learning algorithms to train a binary tree model to predict individuals' training patterns relying on several individual difference variables that have been identified as relevant in previous literature. These individual difference variables included pre-existing cognitive abilities, personality characteristics, motivational factors, video game experience, health status, bilingualism, and socioeconomic status. We found that our classification model showed good predictive power in distinguishing between high performers and relatively lower performers. Furthermore, we found that openness and pre-existing WM capacity to be the two most important factors in distinguishing between high and low performers. However, among low performers, openness and video game background were the most significant predictors of their learning persistence. In conclusion, it is possible to predict individual training performance using participant characteristics before training, which could inform the development of personalized interventions.

Contrast Response Function Estimation with Nonparametric Bayesian Active Learning.

Multidimensional psychometric functions can typically be estimated nonparametrically for greater accuracy or parametrically for greater efficiency. By recasting the estimation problem from regression to classification, however, powerful machine learning tools can be leveraged to provide an adjustable balance between accuracy and efficiency. Contrast Sensitivity Functions (CSFs) are behaviorally estimated curves that provide insight into both peripheral and central visual function. Because estimation can be impractically long, current clinical workflows must make compromises such as limited sampling across spatial frequency or strong assumptions on CSF shape. This paper describes the development of the Machine Learning Contrast Response Function (MLCRF) estimator, which quantifies the expected probability of success in performing a contrast detection or discrimination task. A machine learning CSF can then be derived from the MLCRF. Using simulated eyes created from canonical CSF curves and actual human contrast response data, the accuracy and efficiency of the MLCSF was evaluated in order to determine its potential utility for research and clinical applications. With stimuli selected randomly, the MLCSF estimator converged slowly toward ground truth. With optimal stimulus selection via Bayesian active learning, convergence was nearly an order of magnitude faster, requiring only tens of stimuli to achieve reasonable estimates. Inclusion of an informative prior provided no consistent advantage to the estimator as configured. MLCSF achieved efficiencies on par with quickCSF, a conventional parametric estimator, but with systematically higher accuracy. Because MLCSF design allows accuracy to be traded off against efficiency, it should be explored further to uncover its full potential. Precis: Machine learning classifiers enable accurate and efficient contrast sensitivity function estimation with item-level prediction for individual eyes.

Computational Audiology: New Approaches to Advance Hearing Health Care in the Digital Age.

The global digital transformation enables computational audiology for advanced clinical applications that can reduce the global burden of hearing loss. In this article, we describe emerging hearing-related artificial intelligence applications and argue for their potential to improve access, precision, and efficiency of hearing health care services. Also, we raise awareness of risks that must be addressed to enable a safe digital transformation in audiology. We envision a future where computational audiology is implemented via interoperable systems using shared data and where health care providers adopt expanded roles within a network of distributed expertise. This effort should take place in a health care system where privacy, responsibility of each stakeholder, and patients' safety and autonomy are all guarded by design.

Precision clinical trials: a framework for getting to precision medicine for neurobehavioural disorders.

The goal of precision medicine (individually tailored treatments) is not being achieved for neurobehavioural conditions such as psychiatric disorders. Traditional randomized clinical trial methods are insufficient for advancing precision medicine because of the dynamic complexity of these conditions. We present a pragmatic solution: the precision clinical trial framework, encompassing methods for individually tailored treatments. This framework includes the following: (1) treatment-targeted enrichment, which involves measuring patients' response after a brief bout of an intervention, and then randomizing patients to a full course of treatment, using the acute response to predict long-term outcomes; (2) adaptive treatments, which involve adjusting treatment parameters during the trial to individually optimize the treatment; and (3) precise measurement, which involves measuring predictor and outcome variables with high accuracy and reliability using techniques such as ecological momentary assessment. This review summarizes precision clinical trials and provides a research agenda, including new biomarkers such as precision neuroimaging, transcranial magnetic stimulation-electroencephalogram digital phenotyping and advances in statistical and machine-learning models. Validation of these approaches - and then widespread incorporation of the precision clinical trial framework - could help achieve the vision of precision medicine for neurobehavioural conditions.

Hidden Hearing Loss: Mixed Effects of Compensatory Plasticity.

Hidden hearing loss manifests as speech perception difficulties with normal hearing thresholds. A new study shows that neural compensation induced by this disorder may actually improve speech perception under narrow conditions within an overall profile of degradation.

Dynamically Masked Audiograms With Machine Learning Audiometry.

OBJECTIVES: When one ear of an individual can hear significantly better than the other ear, evaluating the worse ear with loud probe tones may require delivering masking noise to the better ear to prevent the probe tones from inadvertently being heard by the better ear. Current masking protocols are confusing, laborious, and time consuming. Adding a standardized masking protocol to an active machine learning audiogram procedure could potentially alleviate all of these drawbacks by dynamically adapting the masking as needed for each individual. The goal of this study is to determine the accuracy and efficiency of automated machine learning masking for obtaining true hearing thresholds. DESIGN: Dynamically masked automated audiograms were collected for 29 participants between the ages of 21 and 83 (mean 43, SD 20) with a wide range of hearing abilities. Normal-hearing listeners were given unmasked and masked machine learning audiogram tests. Listeners with hearing loss were given a standard audiogram test by an audiologist, with masking stimuli added as clinically determined, followed by a masked machine learning audiogram test. The hearing thresholds estimated for each pair of techniques were compared at standard audiogram frequencies (i.e., 0.25, 0.5, 1, 2, 4, 8 kHz). RESULTS: Masked and unmasked machine learning audiogram threshold estimates matched each other well in normal-hearing listeners, with a mean absolute difference between threshold estimates of 3.4 dB. Masked machine learning audiogram thresholds also matched well the thresholds determined by a conventional masking procedure, with a mean absolute difference between threshold estimates for listeners with low asymmetry and high asymmetry between the ears, respectively, of 4.9 and 2.6 dB. Notably, out of 6200 masked machine learning audiogram tone deliveries for this study, no instances of tones detected by the nontest ear were documented. The machine learning methods were also generally faster than the manual methods, and for some listeners, substantially so. CONCLUSIONS: Dynamically masked audiograms achieve accurate true threshold estimates and reduce test time compared with current clinical masking procedures. Dynamic masking is a compelling alternative to the methods currently used to evaluate individuals with highly asymmetric hearing, yet can also be used effectively and efficiently for anyone.

Visual Field Estimation by Probabilistic Classification.

The gold standard clinical tool for evaluating visual dysfunction in cases of glaucoma and other disorders of vision remains the visual field or threshold perimetry exam. Administration of this exam has evolved over the years into a sophisticated, standardized, automated algorithm that relies heavily on specifics of disease processes particular to common retinal disorders. The purpose of this study is to evaluate the utility of a novel general estimator applied to visual field testing. A multidimensional psychometric function estimation tool was applied to visual field estimation. This tool is built on semiparametric probabilistic classification rather than multiple logistic regression. It combines the flexibility of nonparametric estimators and the efficiency of parametric estimators. Simulated visual fields were generated from human patients with a variety of diagnoses, and the errors between simulated ground truth and estimated visual fields were quantified. Error rates of the estimates were low, typically within 2 dB units of ground truth on average. The greatest threshold errors appeared to be confined to the portions of the threshold function with the highest spatial frequencies. This method can accurately estimate a variety of visual field profiles with continuous threshold estimates, potentially using a relatively small number of stimuli.

Talking Points: A Modulating Circle Increases Listening Effort Without Improving Speech Recognition in Young Adults.

Speech recognition is improved when the acoustic input is accompanied by visual cues provided by a talking face (Erber in Journal of Speech and Hearing Research, 12(2), 423-425, 1969; Sumby & Pollack in The Journal of the Acoustical Society of America, 26(2), 212-215, 1954). One way that the visual signal facilitates speech recognition is by providing the listener with information about fine phonetic detail that complements information from the auditory signal. However, given that degraded face stimuli can still improve speech recognition accuracy (Munhall, Kroos, Jozan, & Vatikiotis-Bateson in Perception & Psychophysics, 66(4), 574-583, 2004), and static or moving shapes can improve speech detection accuracy (Bernstein, Auer, & Takayanagi in Speech Communication, 44(1-4), 5-18, 2004), aspects of the visual signal other than fine phonetic detail may also contribute to the perception of speech. In two experiments, we show that a modulating circle providing information about the onset, offset, and acoustic amplitude envelope of the speech does not improve recognition of spoken sentences (Experiment 1) or words (Experiment 2). Further, contrary to our hypothesis, the modulating circle increased listening effort despite subjective reports that it made the word recognition task seem easier to complete (Experiment 2). These results suggest that audiovisual speech processing, even when the visual stimulus only conveys temporal information about the acoustic signal, may be a cognitively demanding process.

Precision medicine and the cursed dimensions.

Our intuition regarding "average" is rooted in one-dimensional thinking, such as the distribution of height across a population. This intuition breaks down in higher dimensions when multiple measurements are combined: fewer individuals are close to average for many measurements simultaneously than for any single measurement alone. This phenomenon is known as the curse of dimensionality. In medicine, diagnostic sophistication generally increases through the addition of more predictive factors. Disease classes themselves become more dissimilar as a result, increasing the difficulty of incorporating (i.e., averaging) multiple patients into a single class for guiding treatment of new patients. Failure to consider the curse of dimensionality will ultimately lead to inherent limits on the degree to which precision medicine can extend the advances of evidence-based medicine for selecting suitable treatments. One strategy to compensate for the curse of dimensionality involves incorporating predictive observation models into the patient workup.

Talking points: A modulating circle reduces listening effort without improving speech recognition.

Speech recognition is improved when the acoustic input is accompanied by visual cues provided by a talking face (Erber in Journal of Speech and Hearing Research, 12(2), 423-425 1969; Sumby & Pollack in The Journal of the Acoustical Society of America, 26(2), 212-215, 1954). One way that the visual signal facilitates speech recognition is by providing the listener with information about fine phonetic detail that complements information from the auditory signal. However, given that degraded face stimuli can still improve speech recognition accuracy (Munhall et al. in Perception & Psychophysics, 66(4), 574-583, 2004), and static or moving shapes can improve speech detection accuracy (Bernstein et al. in Speech Communication, 44(1/4), 5-18, 2004), aspects of the visual signal other than fine phonetic detail may also contribute to the perception of speech. In two experiments, we show that a modulating circle providing information about the onset, offset, and acoustic amplitude envelope of the speech does not improve recognition of spoken sentences (Experiment 1) or words (Experiment 2), but does reduce the effort necessary to recognize speech. These results suggest that although fine phonetic detail may be required for the visual signal to benefit speech recognition, low-level features of the visual signal may function to reduce the cognitive effort associated with processing speech.

Conjoint psychometric field estimation for bilateral audiometry.

Behavioral testing in perceptual or cognitive domains requires querying a subject multiple times in order to quantify his or her ability in the corresponding domain. These queries must be conducted sequentially, and any additional testing domains are also typically tested sequentially, such as with distinct tests comprising a test battery. As a result, existing behavioral tests are often lengthy and do not offer comprehensive evaluation. The use of active machine-learning kernel methods for behavioral assessment provides extremely flexible yet efficient estimation tools to more thoroughly investigate perceptual or cognitive processes without incurring the penalty of excessive testing time. Audiometry represents perhaps the simplest test case to demonstrate the utility of these techniques. In pure-tone audiometry, hearing is assessed in the two-dimensional input space of frequency and intensity, and the test is repeated for both ears. Although an individual's ears are not linked physiologically, they share many features in common that lead to correlations suitable for exploitation in testing. The bilateral audiogram estimates hearing thresholds in both ears simultaneously by conjoining their separate input domains into a single search space, which can be evaluated efficiently with modern machine-learning methods. The result is the introduction of the first conjoint psychometric function estimation procedure, which consistently delivers accurate results in significantly less time than sequential disjoint estimators.

Online Machine Learning Audiometry.

OBJECTIVES: A confluence of recent developments in cloud computing, real-time web audio and machine learning psychometric function estimation has made wide dissemination of sophisticated turn-key audiometric assessments possible. The authors have combined these capabilities into an online (i.e., web-based) pure-tone audiogram estimator intended to empower researchers and clinicians with advanced hearing tests without the need for custom programming or special hardware. The objective of this study was to assess the accuracy and reliability of this new online machine learning audiogram method relative to a commonly used hearing threshold estimation technique also implemented online for the first time in the same platform. DESIGN: The authors performed air conduction pure-tone audiometry on 21 participants between the ages of 19 and 79 years (mean 41, SD 21) exhibiting a wide range of hearing abilities. For each ear, two repetitions of online machine learning audiogram estimation and two repetitions of online modified Hughson-Westlake ascending-descending audiogram estimation were acquired by an audiologist using the online software tools. The estimated hearing thresholds of these two techniques were compared at standard audiogram frequencies (i.e., 0.25, 0.5, 1, 2, 4, 8 kHz). RESULTS: The two threshold estimation methods delivered very similar threshold estimates at standard audiogram frequencies. Specifically, the mean absolute difference between threshold estimates was 3.24 ± 5.15 dB. The mean absolute differences between repeated measurements of the online machine learning procedure and between repeated measurements of the Hughson-Westlake procedure were 2.85 ± 6.57 dB and 1.88 ± 3.56 dB, respectively. The machine learning method generated estimates of both threshold and spread (i.e., the inverse of psychometric slope) continuously across the entire frequency range tested from fewer samples on average than the modified Hughson-Westlake procedure required to estimate six discrete thresholds. CONCLUSIONS: Online machine learning audiogram estimation in its current form provides all the information of conventional threshold audiometry with similar accuracy and reliability in less time. More importantly, however, this method provides additional audiogram details not provided by other methods. This standardized platform can be readily extended to bone conduction, masking, spectrotemporal modulation, speech perception, etc., unifying audiometric testing into a single comprehensive procedure efficient enough to become part of the standard audiologic workup.

Correction to: Bayesian active probabilistic classification for psychometric field estimation.

The original version of this article neglected to mention a conflict of interest. DLB has a patent pending on technology described in this manuscript.