Abstract
In classical computational neuroscience, analytical model descriptions are derived from neuronal recordings to mimic the underlying biological system. These neuronal models are typically ...slow to compute and cannot be integrated within large-scale neuronal simulation frameworks. We present a hybrid, machine-learning and computational-neuroscience approach that transforms analytical models of sensory neurons and synapses into deep-neural-network (DNN) neuronal units with the same biophysical properties. Our DNN-model architecture comprises parallel and differentiable equations that can be used for backpropagation in neuro-engineering applications, and offers a simulation run-time improvement factor of 70 and 280 on CPU or GPU systems respectively. We focussed our development on auditory neurons and synapses, and show that our DNN-model architecture can be extended to a variety of existing analytical models. We describe how our approach for auditory models can be applied to other neuron and synapse types to help accelerate the development of large-scale brain networks and DNN-based treatments of the pathological system.
Sound processing in the human auditory system is complex and highly non-linear, whereas hearing aids (HAs) still rely on simplified descriptions of auditory processing or hearing loss to restore ...hearing. Even though standard HA amplification strategies succeed in restoring audibility of faint sounds, they still fall short of providing targeted treatments for complex sensorineural deficits and adverse listening conditions. These shortcomings of current HA devices demonstrate the need for advanced hearing-loss compensation strategies that can effectively leverage the non-linear character of the auditory system. Here, we propose a differentiable deep-neural-network (DNN) framework that can be used to train DNN-based HA models based on biophysical auditory-processing differences between normal-hearing and hearing-impaired systems. We investigate different loss functions to accurately compensate for impairments that include outer-hair-cell (OHC) loss and cochlear synaptopathy (CS), and evaluate the benefits of our trained DNN-based HA models for speech processing in quiet and in noise. Our results show that auditory-processing enhancement was possible for all considered hearing-loss cases, with OHC loss proving easier to compensate than CS. Several objective metrics were considered to estimate the expected speech intelligibility after processing, and these simulations hold promise in yielding improved understanding of speech-in-noise for hearing-impaired listeners who use our DNN-based HA processing. Since our framework can be tuned to the hearing-loss profiles of individual listeners, we enter an era where truly individualised and DNN-based hearing-restoration strategies can be developed and be tested experimentally.
Current hearing aids mostly provide sound amplification fittings based on individual hearing thresholds or perceived loudness, even though it is known that sensorineural hearing damage is ...functionally complex, and requires different treatment strategies. To meet this demand, we propose an optimisation framework for the design of individualised hearingaid signal processing based on simulated (hearing-impaired) auditory-nerve responses. The framework is fully differentiable, thus the backpropagation algorithm can be used to train DNN-based hearing-aid models that optimally process sound to restore hearing in impaired cochleae. The auditory models within the framework can be tuned to the precise hearing-loss profile of a listener to yield trully individualised restoration strategies. Our simulations show that the trained hearing-aid models were able to enhance the auditory-nerve responses of hearing-impaired cochleae, and this provides a promising outlook for embedding our framework within future hearing aids and augmented-hearing applications.
•Ageing can result in synaptic damage to the auditory nerve (CS).•Hearing aids do not compensate for the functional consequences of CS.•Model-based hearing-enhancement strategies restored ...temporal-envelope processing.•Peripheral coding and speech intelligibility were enhanced in NH listeners.•CS-compensating algorithms could improve current hearing aids and hearables.
It is well known that ageing and noise exposure are important causes of sensorineural hearing loss, and can result in damage of the outer hair cells or other structures of the inner ear, including synaptic damage to the auditory nerve (AN), i.e., cochlear synaptopathy (CS). Despite the suspected high prevalence of CS among people with self-reported hearing difficulties but seemingly normal hearing, conventional hearing-aid algorithms do not compensate for the functional deficits associated with CS. Here, we present and evaluate a number of auditory signal-processing strategies designed to maximally restore AN coding for listeners with CS pathologies. We evaluated our algorithms in subjects with and without suspected age-related CS to assess whether physiological and behavioural markers associated with CS can be improved. Our data show that after applying our algorithms, envelope-following responses and perceptual amplitude-modulation sensitivity were consistently enhanced in both young and older listeners. Speech-in-noise intelligibility showed small improvements after processing but mostly for young normal-hearing participants, with median improvements of up to 8.3%. Since our hearing-enhancement strategies were designed to optimally drive the AN fibres, they were able to improve temporal-envelope processing for listeners both with and without suspected CS. Our proposed algorithms can be rapidly executed and can thus extend the application range of current hearing aids and hearables, while leaving sound amplification unaffected.
Although hearing aids (HAs) can compensate for elevated hearing thresholds using sound amplification, they often fail to restore auditory perception in adverse listening conditions. To achieve robust ...treatment outcomes for diverse HA users, we use a differentiable framework that can compensate for impaired auditory processing based on a biophysically realistic and personalisable auditory model. Here, we present a deep-neural-network (DNN) HA processing strategy that can provide individualised sound processing for the audiogram of a listener using a single model architecture. The DNN architecture was trained to compensate for different audiogram inputs and was able to enhance simulated responses and intelligibility even for audiograms that were not part of training. Our multi-purpose HA model can be used for different individuals and can process audio inputs of 3.2 ms in <0.5 ms, thus paving the way for precise DNN-based treatments of hearing loss that can be embedded in hearing devices.
This thesis concerns the development, analysis, and computer implementation of mesh generation algorithms encountered in finite element modeling in health care and aerospace. The finite element ...method can reduce a continuous system to a discrete idealization that can be solved in the same manner as a discrete system, provided the continuum is discretized into a finite number of simple geometric shapes (e.g., triangles in two dimensions or tetrahedrons in three dimensions). In health care, namely anatomic modeling, a discretization of the biological object is essential to compute tissue deformation for physics-based simulations. This thesis proposes an efficient procedure to convert 3-dimensional imaging data into adaptive lattice-based discretizations of well-shaped tetrahedra or mixed elements (i.e., tetrahedra, pentahedra and hexahedra). This method operates directly on segmented images, thus skipping a surface reconstruction that is required by traditional Computer-Aided Design (CAD)-based meshing techniques and is convoluted, especially in complex anatomic geometries. Our approach utilizes proper mesh gradation and tissue-specific multi-resolution, without sacrificing the fidelity and while maintaining a smooth surface to reflect a certain degree of visual reality. Image-to-mesh conversion can facilitate accurate computational modeling for biomechanical registration of Magnetic Resonance Imaging (MRI) in image-guided neurosurgery. Neuronavigation with deformable registration of preoperative MRI to intraoperative MRI allows the surgeon to view the location of surgical tools relative to the preoperative anatomical (MRI) or functional data (DT-MRI, fMRI), thereby avoiding damage to eloquent areas during tumor resection. This thesis presents a deformable registration framework that utilizes multi-tissue mesh adaptation to map preoperative MRI to intraoperative MRI of patients who have undergone a brain tumor resection. Our enhancements with mesh adaptation improve the accuracy of the registration by more than 5 times compared to rigid and traditional physics-based non-rigid registration, and by more than 4 times compared to publicly available B-Spline interpolation methods. The adaptive framework is parallelized for shared memory multiprocessor architectures. Performance analysis shows that this method could be applied, on average, in less than two minutes, achieving desirable speed for use in a clinical setting. The last part of this thesis focuses on finite element modeling of CAD data. This is an integral part of the design and optimization of components and assemblies in industry. We propose a new parallel mesh generator for efficient tetrahedralization of piecewise lin- ear complex domains in aerospace. CAD-based meshing algorithms typically improve the shape of the elements in a post-processing step due to high complexity and cost of the operations involved. On the contrary, our method optimizes the shape of the elements throughout the generation process to obtain a maximum quality and utilizes high performance computing to reduce the overheads and improve end-user productivity. The proposed mesh generation technique is a combination of Advancing Front type point placement, direct point insertion, and parallel multi-threaded connectivity optimization schemes. The mesh optimization is based on a speculative (optimistic) approach that has been proven to perform well on hardware-shared memory. The experimental evaluation indicates that the high quality and performance attributes of this method see substantial improvement over existing state-of-the-art unstructured grid technology currently incorporated in several commercial systems. The proposed mesh generator will be part of an Extreme-Scale Anisotropic Mesh Generation Environment to meet industries expectations and NASA's CFD vision for 2030.
Sound processing in the human auditory system is complex and highly non-linear, whereas hearing aids (HAs) still rely on simplified descriptions of auditory processing or hearing loss to restore ...hearing. Even though standard HA amplification strategies succeed in restoring audibility of faint sounds, they still fall short of providing targeted treatments for complex sensorineural deficits and adverse listening conditions. These shortcomings of current HA devices demonstrate the need for advanced hearing-loss compensation strategies that can effectively leverage the non-linear character of the auditory system. Here, we propose a differentiable deep-neural-network (DNN) framework that can be used to train DNN-based HA models based on biophysical auditory-processing differences between normal-hearing and hearing-impaired systems. We investigate different loss functions to accurately compensate for impairments that include outer-hair-cell (OHC) loss and cochlear synaptopathy (CS), and evaluate the benefits of our trained DNN-based HA models for speech processing in quiet and in noise. Our results show that auditory-processing enhancement was possible for all considered hearing-loss cases, with OHC loss proving easier to compensate than CS. Several objective metrics were considered to estimate the expected speech intelligibility after processing, and these simulations hold promise in yielding improved understanding of speech-in-noise for hearing-impaired listeners who use our DNN-based HA processing. Since our framework can be tuned to the hearing-loss profiles of individual listeners, we enter an era where truly individualised and DNN-based hearing-restoration strategies can be developed and be tested experimentally.