Human Activity Recognition (HAR) is widely applied on wearable devices in our daily lives. However, acquiring high-quality wearable sensor data set with ground-truths is challenging due to the high cost in collecting data and necessity of domain experts. In order to achieve generalization from limited data, we study augmentation-based Self-Supervised Learning (SSL) for data from wearable devices. However, there is an issue in one of the most popular SSL approaches, contrastive learning: it is sensitive to the choice of data augmentations. To resolve this, we first propose to combine contrastive learning with generative learning, which is robust to augmentations. Second, we propose an automatic augmentation policy search method to discover the most promising augmentation policy. We empirically verify our approaches on three public HAR datasets. Experimental results show that our proposed SSL approach is robust to augmentations, and delivers higher accuracy than contrastive learning. Additionally, with the searched augmentation policy we are able to further improve the accuracy of HAR task.

As one of the successful Transformer-based models in computer vision tasks, SegFormer demonstrates superior performance in semantic segmentation. Nevertheless, the high computational cost greatly challenges the deployment of SegFormer on edge devices. In this paper, we seek to design a lightweight SegFormer for efficient semantic segmentation. Based on the observation that neurons in SegFormer layers exhibit large variances across different images, we propose a dynamic gated linear layer, which prunes the most uninformative set of neurons based on the input instance. To improve the dynamically pruned SegFormer, we also introduce two-stage knowledge distillation to transfer the knowledge within the original teacher to the pruned student network. Experimental results show that our method can significantly reduce the computation overhead of SegFormer without an apparent performance drop. For instance, we can achieve 36.9% mIoU with only 3.3G FLOPs on ADE20K, saving more than 60% computation with the drop of only 0.5% in mIoU.

Most state-of-the-art Convolutional Neural Networks (CNNs) are bulky and cannot be deployed on resourceconstrained edge devices. In order to leverage the exceptional generalizability of CNNs on edge-devices, they need to be made efficient in terms of memory usage, model size, and power consumption, while maintaining acceptable performance. Neural architecture search (NAS) is a recent approach for developing efficient, edgedeployable CNNs. On the other hand, CNNs used for classification, albeit developed using NAS, often contain large fully-connected (FC) layers with thousands of parameters, contributing to the bulkiness of CNNs. Recent works have shown that FC layers can be compressed, with minimal loss in performance, if any, using tensor processing methods. In this work, for the first time in literature, we leverage tensor methods in the NAS framework to discover efficient CNNs. Specifically, we employ tensor contraction layers (TCLs) to compress fully connected layers in the NAS framework and control the trade-off between compressibility and classification performance by handcrafting the ranks of TCLs. Additionally, we modify the NAS procedure to incorporate automatic TCL rank search in an end-to-end fashion, without human intervention. Our numerical studies on a wide variety of datasets including CIFAR-10, CIFAR-100, and Imagenette (a subset of ImageNet) demonstrate the superior performance of the proposed method in the automatic discovery of CNNs, whose model sizes are manyfold smaller than other cutting-edge mobile CNNs, while maintaining similar classification performance.

BACKGROUND:Anosmia has an estimated prevalence of 5% of the general population. Outside of inflammatory causes, therapeutic options are limited despite research advances. Bypassing peripheral neuronal damage through central stimulation is a potential therapeutic option that has shown success in other sensory systems, most notably with hearing. We performed a pilot study to determine the feasibility of inducing smell through artificial electrical stimulation of the olfactory bulbs in humans.METHODS:Subjects with a history of sinus surgery, including total ethmoidectomy, with intact ability to smell were enrolled. The ability to smell was confirmed with a 40-item smell identification test. Awake subjects underwent nasal endoscopy and either a monopolar or bipolar electrode was positioned at 3 areas along the lateral lamella of the cribriform plate within the ethmoid sinus cavity. A graded stimulation current of 1-20 mA at 3.17 Hz was administered while cortical evoked potential (CEP) recordings were collected. Subjective responses of perceived smell along with reports of discomfort were recorded. Subjects with artificially induced smell underwent repeat stimulation after medically induced anosmia.RESULTS:Five subjects (age, 43-72 years) were enrolled. Three subjects reported smell perception smell with electrical stimulation. This was reproducible after inducing anosmia, but CEP recordings could not provide objective support. All subjects tolerated the study with minimal discomfort.CONCLUSION:This is the first report of induced smell through transethmoid electrical stimulation of the olfactory bulb. These results provide a proof of concept for efforts in development of an olfactory implant system.

Keywords: carotid artery; dissection; hypoglossal nerve; magnetic resonance imaging; tongue

Objective: To describe a patient with novel mutations in the FGD4/FRABIN gene responsible for a predominantly axonal form of Charcot- Marie- Tooth (CMT) type 4H. Background: CMT4H is a rare, autosomal recessive disorder caused by mutation in the gene (FGD4) encoding FRABIN on chromosome 12. To date, 13 different mutations have been reported, predominantly causing demyelinating pathology. Design/Methods: An 18-year old woman from Azerbaijan presented with unsteady gait and repeated falls. Born to non consanguineous parents, she reached all motor milestones on time noted to toe-walk in childhood, and began to fall in her early teenage years. At age 16, testing for a mitofusin mutation was negative. Thereafter, CIDP was suspected but there was no response to corticosteroids or IVIG. On examination, she had a high stepping, waddling gait, bilateral pes cavus, pronounced distal lower limb weakness and preserved reflexes. Electrophysiological studies revealed a predominantly axonal sensory and motor polyneuropathy (absent sural; normal median and radial sensory; absent peroneal (EDB) and tibial motor; normal median and ulnar motor responses; mild slowing of nerve conduction velocities; and active and chronic denervation in the distal leg muscles). DNA testing using next-generation sequencing technology was performed at Athena Diagnostics. Results: There was a heterozygous frameshift mutation (c. 1887-1891: 5bp deletion; Codon: 629-631) in one allele of FGD4 gene which was deemed pathogenic and a heterozygous missense mutation (c.2149G>A; p.Val717Met) of unknown clinical significance in the other. Additional testing on asymptomatic family members revealed that her mother carried the pathogenic mutation, father carried the mutation of unknown significance and brother, neither. Conclusions: This study has identified a compound heterozygote with two novel mutations in the FGD4/FRABIN gene, which resulted in a disease phenotype. This study suggests that the spectrum of neuropathies associated with CMT4H can be broadened to include a predominantly axonal disorder. Disclosure: Dr. Rajan has nothing to disclose. Dr. Chad has nothing to disclose. Dr. Nair has nothing to disclose.

Purpose: The burst suppression (BS) EEG patterns induced by general anesthesia can react to somatosensory stimuli. We investigated this reactivity by studying the effect of peripheral nerve stimulation used for routine intraoperative spinal cord monitoring by somatosensory evoked potentials on BS patterns. Methods: The relative time spent in suppression expressed as BS ratio (BSR) and mean burst duration were measured before (BSRPre), during (BSRStim), and after (BSRPost) a 60-second repetitive electrical ulnar nerve stimulation in nine patients under total intravenous general anesthesia with propofol. The BS reactivity was measured as BSRPre-BSRStim. Results: Overall, 27 trials were included with BSRPre up to 77%, indistinguishable from BSRPost. During stimulation, the mean BSR transiently decreased from 42% to 35%. For each 1% increase in BSRPre, the BS reactivity increased with 0.6%, whereas the burst duration remained approximately 3 seconds. For BSRPre below 30%, the BS reactivity was negligible. Conclusions: Data from this study show that somatosensory input can evoke bursts, altering the “spontaneous” deep BS patterns (BSRPre >30%). Further studies are necessary to objectively assess the clinical relevance of stimulus-induced BS reactivity during deep general anesthesia.

The use of intraoperative mapping of neural structures has come to be an indispensable technique to prevent or minimize postoperative morbidity. In this article, I briefly mention its use in mapping nerve roots, plexuses, and peripheral nerves. The reader may find some mention of monitoring techniques, too, as monitoring and mapping are seldom done separately. I include the technical details of different mapping modalities, relevant anatomy, and clinical applications, as appropriate.

PURPOSE:It is thought that following a stroke the contralesional motor region exerts an undue inhibitory influence on the lesional motor region which might limit recovery. Pilot studies have shown that suppressing the contralesional motor region with cathodal transcranial Direct Current Stimulation (tDCS) can induce a short lasting functional benefit; greater and longer lasting effects might be achieved with combining tDCS with simultaneous occupational therapy (OT) and applying this intervention for multiple sessions.METHODS:We carried out a randomized, double blind, sham controlled study of chronic stroke patients receiving either 5 consecutive days of cathodal tDCS (for 30 minutes) applied to the contralesional motor region and simultaneous OT, or sham tDCS+OT.RESULTS:we showed that cathodal tDCS+OT resulted in significantly more improvement in Range-Of-Motion in multiple joints of the paretic upper extremity and in the Upper-Extremity Fugl-Meyer scores than sham tDCS+OT, and that the effects lasted at least one week post-stimulation. Improvement in motor outcome scores was correlated with decrease in fMRI activation in the contralesional motor region exposed to cathodal stimulation.CONCLUSIONS:This suggests that cathodal tDCS combined with OT leads to significant motor improvement after stroke due to a decrease in the inhibitory effect that the contralesional hemisphere exerts onto the lesional hemisphere.

Objective: Motor recovery after stroke depends on the integrity of ipsilesional motor circuits and interactions between the ipsilesional and contralesional hemispheres. In this sham-controlled randomized trial, we investigated whether noninvasive modulation of regional excitability of bilateral motor cortices in combination with physical and occupational therapy improves motor outcome after stroke. Methods: Twenty chronic stroke patients were randomly assigned to receive 5 consecutive sessions of either 1) bihemispheric transcranial direct current stimulation (tDCS) (anodal tDCS to upregulate excitability of ipsilesional motor cortex and cathodal tDCS to downregulate excitability of contralesional motor cortex) with simultaneous physical/occupational therapy or 2) sham stimulation with simultaneous physical/occupational therapy. Changes in motor impairment (Upper Extremity Fugl-Meyer) and motor activity (Wolf Motor Function Test) assessments were outcome measures while functional imaging parameters were used to identify neural correlates of motor improvement. Results: The improvement of motor function was significantly greater in the real stimulation group (20.7% in Fugl-Meyer and 19.1% in Wolf Motor Function Test scores) when compared to the sham group (3.2% in Fugl-Meyer and 6.0% in Wolf Motor Function Test scores). The effects outlasted the stimulation by at least 1 week. In the real-stimulation group, stronger activation of intact ipsilesional motor regions during paced movements of the affected limb were found postintervention whereas no significant activation changes were seen in the control group. Conclusions: The combination of bihemispheric tDCS and peripheral sensorimotor activities improved motor functions in chronic stroke patients that outlasted the intervention period. This novel approach may potentiate cerebral adaptive processes that facilitate motor recovery after stroke. Classification of evidence: This study provides Class I evidence that for adult patients with ischemic stroke treated at least 5 months after their first and only stroke, bihemispheric tDCS and simultaneous physical/occupational therapy given over 5 consecutive sessions significantly improves motor function as measured by the Upper Extremity Fugl-Meyer assessment (raw change treated 6.1 ± 3.4, sham 1.2 ± 1.0).

Publisher Summary This chapter explains the popular courseware for image processing education known as SIVA—the signal, image, and video audiovisualization—gallery. The image and video processing section of the SIVA gallery consists of a suite of special purpose LabVIEW-based programs known as Virtual Instruments (VIs). Equipped with informative visualization and a user-friendly interface, these VIs were carefully designed to facilitate a gentle introduction to the fascinating concepts in image and video processing. To make things easier, the demos are accompanied by a comprehensive set of help files that describe the various controls, and that highlight some illustrative examples and instructive parameter settings. A demo can be activated by clicking the rightward pointing arrow in the top menu bar. Help for the demo can be activated by clicking the ‘?’ button and moving the cursor over the icon that is located immediately to the right of the ‘?’ button. In addition, when the cursor is placed over any other button/control, the help window automatically updates to describe the function of that button/control. The user will find this visual, hands-on, interactive introduction to image processing to be a fun, enjoyable, and illuminating experience.