Objective The aims of this study were (i) to determine the magnitude of the interaural level differences (ILDs) that remain after cochlear implant (CI) signal processing and (ii) to relate the ILDs to the pattern of errors for sound source localization on the horizontal plane. Design The listeners were 16 bilateral CI patients fitted with MED-EL cochlear implants and 34 normal hearing listeners. The stimuli were wideband, high-pass and low-pass noise signals. ILDs were calculated by passing signals, filtered by head-related transfer functions (HRTFs), to a Matlab simulation of MED-EL signal processing. Results For the wideband signal and high-pass signals, maximum ILDs of 15–17dB in the input signal were reduced to 3–4dB after CI signal processing. For the low-pass signal, ILDs were reduced to 1–2dB. For wideband and high-pass signals, the largest ILDs for +/− 15 degree speaker locations were between .4 and .7dB; for the +/− 30 degree locations between .9 and 1.3dB; for the 45 degree locations between 2.4 and 2.9dB, for the +/− 60 degree locations, between 3.2 and 4.1dB and for the +/− 75 degree locations between 2.7 and 3.4dB. All of the CI patients in all stimulus conditions showed poorer localization than the normal hearing listeners. Localization accuracy for CI patients was best for the wideband and high-pass signals and was poorest for the low-pass signal. Conclusions Localization accuracy was related to the magnitude of the ILD cues available to the normal hearing listeners and CI patients. The pattern of localization errors for the CI patients was related to the magnitude of the ILD differences among loudspeaker locations. The error patterns for the wideband and high-pass signals, suggest that, for the conditions of this experiment, patients, on average, sorted signals on the horizontal plane into four sectors – on each side of the midline, one sector including 0, 15 and possibly 30 degrees, and a sector from 45 degrees to 75 degrees. Resolution within a sector was relatively poor.

No abstract

Widespread adoption of artificial intelligence has yet to occur in the hearing field. Hearing technologies, such as cochlear implants (CIs), provide limited benefits of noise reduction, even with current state-of-the-art signal processing strategies. Recent developments in machine learning have produced deep neural network (DNN) models achieving remarkable performance in speech enhancement and source separation tasks. However, there are currently no commercially available CI audio processors that utilize DNN models for noise suppression. Furthermore, the current research community lacks a computational tool to match the complexity of natural auditory processing. To address these gaps, we implemented two DNN models: a recurrent neural network (RNN)—a lightweight template model for speech enhancement, and the SepFormer—the current top-performing speech-separation model in the literature. The DNN models resulted in significant improvements in terms of objective evaluation metrics, as well as intelligibility scores obtained with CI users at different signal-to-noise ratios. Given their flexibility and good performance on complex tasks, these models can also be used to generate hypotheses about speech-in-noise perception and serve as richer substitutes for models commonly used in research. This work serves as a proof-of-concept and a guide for the next steps towards integrating DNN technology into hearing technologies and research.

The fluorescence from the diseased sites was found to be higher than the normal site. Clinical and fluorescence data suggest that a dose of 25 mg/kg may be satisfactory for diagnostic purposes and would have minimal side effects.