Age-Related Changes in the Relationship Between Auditory Brainstem Responses and Envelope-Following Responses

Hearing thresholds and wave amplitudes measured using auditory brainstem responses (ABRs) to brief sounds are the predominantly used clinical measures to objectively assess auditory function. However, frequency-following responses (FFRs) to tonal carriers and to the modulation envelope (envelope-following responses or EFRs) to longer and spectro-temporally modulated stimuli are rapidly gaining prominence as a measure of complex sound processing in the brainstem and midbrain. In spite of numerous studies reporting changes in hearing thresholds, ABR wave amplitudes, and the FFRs and EFRs under neurodegenerative conditions, including aging, the relationships between these metrics are not clearly understood. In this study, the relationships between ABR thresholds, ABR wave amplitudes, and EFRs are explored in a rodent model of aging. ABRs to broadband click stimuli and EFRs to sinusoidally amplitude-modulated noise carriers were measured in young (3–6 months) and aged (22–25 months) Fischer-344 rats. ABR thresholds and amplitudes of the different waves as well as phase-locking amplitudes of EFRs were calculated. Age-related differences were observed in all these measures, primarily as increases in ABR thresholds and decreases in ABR wave amplitudes and EFR phase-locking capacity. There were no observed correlations between the ABR thresholds and the ABR wave amplitudes. Significant correlations between the EFR amplitudes and ABR wave amplitudes were observed across a range of modulation frequencies in the young. However, no such significant correlations were found in the aged. The aged click ABR amplitudes were found to be lower than would be predicted using a linear regression model of the young, suggesting altered gain mechanisms in the relationship between ABRs and FFRs with age. These results suggest that ABR thresholds, ABR wave amplitudes, and EFRs measure complementary aspects of overlapping neurophysiological processes and the relationships between these measurements changes asymmetrically with age. Hence, measuring all three metrics provides a more complete assessment of auditory function, especially under pathological conditions like aging.     

From the Cover: Convolutional networks for fast,energy-efficient neuromorphic computing

Deep networks are now able to achieve human-level performance on a broad spectrum of recognition tasks. Independently, neuromorphic computing has now demonstrated unprecedented energy-efficiency through a new chip architecture based on spiking neurons, low precision synapses, and a scalable communication network. Here, we demonstrate that neuromorphic computing, despite its novel architectural primitives, can implement deep convolution networks that (i) approach state-of-the-art classification accuracy across eight standard datasets encompassing vision and speech, (ii) perform inference while preserving the hardware’s underlying energy-efficiency and high throughput, running on the aforementioned datasets at between 1,200 and 2,600 frames/s and using between 25 and 275 mW (effectively >6,000 frames/s per Watt), and (iii) can be specified and trained using backpropagation with the same ease-of-use as contemporary deep learning. This approach allows the algorithmic power of deep learning to be merged with the efficiency of neuromorphic processors, bringing the promise of embedded, intelligent, brain-inspired computing one step closer.The human brain is capable of remarkable acts of perception while consuming very little energy. The dream of brain-inspired computing is to build machines that do the same, requiring high-accuracy algorithms and efficient hardware to run those algorithms. On the algorithm front, building on classic work on backpropagation (1), the neocognitron (2), and convolutional networks (3), deep learning has made great strides in achieving human-level performance on a wide range of recognition tasks (4). On the hardware front, building on foundational work on silicon neural systems (5), neuromorphic computing, using novel architectural primitives, has recently demonstrated hardware capable of running 1 million neurons and 256 million synapses for extremely low power (just 70 mW at real-time operation) (6). Bringing these approaches together holds the promise of a new generation of embedded, real-time systems, but first requires reconciling key differences in the structure and operation between contemporary algorithms and hardware. Here, we introduce and demonstrate an approach we call Eedn, energy-efficient deep neuromorphic networks, which creates convolutional networks whose connections, neurons, and weights have been adapted to run inference tasks on neuromorphic hardware.For structure, typical convolutional networks place no constraints on filter sizes, whereas neuromorphic systems can take advantage of blockwise connectivity that limits filter sizes, thereby saving energy because weights can now be stored in local on-chip memory within dedicated neural cores. Here, we present a convolutional network structure that naturally maps to the efficient connection primitives used in contemporary neuromorphic systems. We enforce this connectivity constraint by partitioning filters into multiple groups and yet maintain network integration by interspersing layers whose filter support region is able to cover incoming features from many groups by using a small topographic size (7).For operation, contemporary convolutional networks typically use high precision ( ≥ 32-bit) neurons and synapses to provide continuous derivatives and support small incremental changes to network state, both formally required for backpropagation-based gradient learning. In comparison, neuromorphic designs can use one-bit spikes to provide event-based computation and communication (consuming energy only when necessary) and can use low-precision synapses to colocate memory with computation (keeping data movement local and avoiding off-chip memory bottlenecks). Here, we demonstrate that by introducing two constraints into the learning rule—binary-valued neurons with approximate derivatives and trinary-valued ({−1,0,1}) synapses—it is possible to adapt backpropagation to create networks directly implementable using energy efficient neuromorphic dynamics. This approach draws inspiration from the spiking neurons and low-precision synapses of the brain (8) and builds on work showing that deep learning can create networks with constrained connectivity (9), low-precision synapses (10, 11), low-precision neurons (12–14), or both low-precision synapses and neurons (15, 16). For input data, we use a first layer to transform multivalued, multichannel input into binary channels using convolution filters that are learned via backpropagation (12, 16) and whose output can be sent on chip in the form of spikes. These binary channels, intuitively akin to independent components (17) learned with supervision, provide a parallel distributed representation to carry out high-fidelity computation without the need for high-precision representation.Critically, we demonstrate that bringing the above innovations together allows us to create networks that approach state-of-the-art accuracy performing inference on eight standard datasets, running on a neuromorphic chip at between 1,200 and 2,600 frames/s (FPS), using between 25 and 275 mW. We further explore how our approach scales by simulating multichip configurations. Ease-of-use is achieved using training tools built from existing, optimized deep learning frameworks (18), with learned parameters mapped to hardware using a high-level deployment language (19). Although we choose the IBM TrueNorth chip (6) for our example deployment platform, the essence of our constructions can apply to other emerging neuromorphic approaches (20–23) and may lead to new architectures that incorporate deep learning and efficient hardware primitives from the ground up.     

Roles of a 67-kDa polypeptide in reversal of protein synthesis inhibition in heme-deficient reticulocyte lysate.

During heme deficiency in reticulocyte lysates, the heme-regulated protein synthesis inhibitor, HRI, phosphorylates the alpha subunit of eukaryotic initiation factor 2 (eIF-2) and thus inhibits protein synthesis. Two factors, eIF-2 and a reticulocyte-lysate supernatant factor that we term RF, reverse this inhibition. We now report the following. (i) An active eIF-2 preparation contained, in addition to the three subunits (alpha, beta, and gamma), a 67-kDa polypeptide. Pretreatment of eIF-2 with polyclonal antibodies against either isolated alpha subunit or 67-kDa polypeptide almost completely inhibited the reversal activity. Upon further fractionation, three-subunit eIF-2 and the 67-kDa polypeptide were resolved. Neither the three-subunit eIF-2 nor the 67-kDa polypeptide alone was active in protein synthesis inhibition reversal. The activity was, however, restored by combining both the three-subunit eIF-2 and the 67-kDa polypeptide. (ii) Active RF preparations contained eIF-2 alpha (unphosphorylated) and beta subunits and the 67-kDa polypeptide. As with eIF-2, prior treatment of the RF preparation with antibodies to either the alpha subunit or the 67-kDa polypeptide almost completely inhibited the reversal activity. The RF preparation devoid of eIF-2 gamma subunit did not form ternary complex (Met-tRNA(fMet).eIF-2.GTP). The eIF-2 gamma subunit in the free form was isolated, and addition of this isolated gamma subunit to RF promoted significant ternary-complex formation. (iii) Purified HRI efficiently phosphorylated the alpha subunit in the three subunit eIF-2. However, the extent of such phosphorylation was significantly reduced when eIF-2 containing the 67-kDa polypeptide was used. The 67-kDa polypeptide apparently protected eIF-2 alpha subunit from HRI-catalyzed phosphorylation but did not inhibit HRI activity. Based on these results, we suggest that the protein synthesis inhibition reversal activity in both eIF-2 and RF is due to the same components--namely, eIF-2 alpha subunit and the 67-kDa polypeptide. The 67-kDa polypeptide protects eIF-2 alpha subunit from HRI-catalyzed phosphorylation and may also be a necessary component of the functioning eIF-2 molecule.     

Evaluation of Hybridization Capture Versus Amplicon‐Based Methods for Whole‐Exome Sequencing

Next‐generation sequencing has aided characterization of genomic variation. While whole‐genome sequencing may capture all possible mutations, whole‐exome sequencing remains cost‐effective and captures most phenotype‐altering mutations. Initial strategies for exome enrichment utilized a hybridization‐based capture approach. Recently, amplicon‐based methods were designed to simplify preparation and utilize smaller DNA inputs. We evaluated two hybridization capture‐based and two amplicon‐based whole‐exome sequencing approaches, utilizing both Illumina and Ion Torrent sequencers, comparing on‐target alignment, uniformity, and variant calling. While the amplicon methods had higher on‐target rates, the hybridization capture‐based approaches demonstrated better uniformity. All methods identified many of the same single‐nucleotide variants, but each amplicon‐based method missed variants detected by the other three methods and reported additional variants discordant with all three other technologies. Many of these potential false positives or negatives appear to result from limited coverage, low variant frequency, vicinity to read starts/ends, or the need for platform‐specific variant calling algorithms. All methods demonstrated effective copy‐number variant calling when evaluated against a single‐nucleotide polymorphism array. This study illustrates some differences between whole‐exome sequencing approaches, highlights the need for selecting appropriate variant calling based on capture method, and will aid laboratories in selecting their preferred approach.     

Multidimensional fetal flow imaging with cardiovascular magnetic resonance: a feasibility study

Purpose

To image multidimensional flow in fetuses using golden-angle radial phase contrast cardiovascular magnetic resonance (PC-CMR) with motion correction and retrospective gating.

Methods

A novel PC-CMR method was developed using an ungated golden-angle radial acquisition with continuously incremented velocity encoding. Healthy subjects (n?=?5, 27?±?3 years, males) and pregnant females (n?=?5, 34?±?2 weeks gestation) were imaged at 3 T using the proposed sequence. Real-time reconstructions were first performed for retrospective motion correction and cardiac gating (using metric optimized gating, MOG). CINE reconstructions of multidimensional flow were then performed using the corrected and gated data.

Results

In adults, flows obtained using the proposed method agreed strongly with those obtained using a conventionally gated Cartesian acquisition. Across the five adults, bias and limits of agreement were ??1.0 cm/s and [??5.1, 3.2] cm/s for mean velocities and???1.1 cm/s and [??6.5, 4.3] cm/s for peak velocities. Temporal correlation between corresponding waveforms was also high (R~?0.98). Calculated timing errors between MOG and pulse-gating RR intervals were low (~?20 ms). First insights into multidimensional fetal blood flows were achieved. Inter-subject consistency in fetal descending aortic flows (n =?3) was strong with an average velocity of 27.1?±?0.4 cm/s, peak systolic velocity of 70.0?±?1.8 cm/s and an intra-class correlation coefficient of 0.95 between the velocity waveforms. In one fetal case, high flow waveform reproducibility was demonstrated in the ascending aorta (R =?0.97) and main pulmonary artery (R =?0.99).

Conclusion

Multidimensional PC-CMR of fetal flow was developed and validated, incorporating retrospective motion compensation and cardiac gating. Using this method, the first quantification and visualization of multidimensional fetal blood flow was achieved using CMR.