Darwin3: a large-scale neuromorphic chip with a novel ISA and on-chip learning.

Spiking neural networks (SNNs) are gaining increasing attention for their biological plausibility and potential for improved computational efficiency. To match the high spatial-temporal dynamics in SNNs, neuromorphic chips are highly desired to execute SNNs in hardware-based neuron and synapse circuits directly. This paper presents a large-scale neuromorphic chip named Darwin3 with a novel instruction set architecture, which comprises 10 primary instructions and a few extended instructions. It supports flexible neuron model programming and local learning rule designs. The Darwin3 chip architecture is designed in a mesh of computing nodes with an innovative routing algorithm. We used a compression mechanism to represent synaptic connections, significantly reducing memory usage. The Darwin3 chip supports up to 2.35 million neurons, making it the largest of its kind on the neuron scale. The experimental results showed that the code density was improved by up to 28.3× in Darwin3, and that the neuron core fan-in and fan-out were improved by up to 4096× and 3072× by connection compression compared to the physical memory depth. Our Darwin3 chip also provided memory saving between 6.8× and 200.8× when mapping convolutional spiking neural networks onto the chip, demonstrating state-of-the-art performance in accuracy and latency compared to other neuromorphic chips.

Simplifying Multimodal With Single EOG Modality for Automatic Sleep Staging.

Polysomnography (PSG) recordings have been widely used for sleep staging in clinics, containing multiple modality signals (i.e., EEG and EOG). Recently, many studies have combined EEG and EOG modalities for sleep staging, since they are the most and the second most powerful modality for sleep staging among PSG recordings, respectively. However, EEG is complex to collect and sensitive to environment noise or other body activities, imbedding its use in clinical practice. Comparatively, EOG is much more easily to be obtained. In order to make full use of the powerful ability of EEG and the easy collection of EOG, we propose a novel framework to simplify multimodal sleep staging with a single EOG modality. It still performs well with only EOG modality in the absence of the EEG. Specifically, we first model the correlation between EEG and EOG, and then based on the correlation we generate multimodal features with time and frequency guided generators by adopting the idea of generative adversarial learning. We collected a real-world sleep dataset containing 67 recordings and used other four public datasets for evaluation. Compared with other existing sleep staging methods, our framework performs the best when solely using the EOG modality. Moreover, under our framework, EOG provides a comparable performance to EEG.

Association of coagulation function with the risk of in-hospital mortality in patients with severe acute respiratory distress syndrome.

BACKGROUND: To evaluate the association of coagulation disorder score with the risk of in-hospital mortality in acute respiratory distress syndrome (ARDS) patients. METHODS: In this cohort study, 7,001 adult patients with ARDS were identified from the Medical Information Mart for Intensive Care Database-IV (MIMIC-IV). Univariate and multivariate Logistic stepwise regression models were used to explore the associations of coagulation-associated biomarkers with the risk of in-hospital mortality in patients with ADRS. Restricted cubic spline (RCS) was plotted to explore the association between coagulation disorder score and in-hospital mortality of ARDS patients. RESULTS: The follow-up time for in-hospital death was 7.15 (4.62, 13.88) days. There were 1,187 patients died and 5,814 people survived in hospital. After adjusting for confounding factors, increased risk of in-hospital mortality in ARDS patients was observed in those with median coagulation disorder score [odds ratio (OR) = 1.22, 95% confidence interval (CI): 1.01-1.47) and high coagulation disorder score (OR = 1.38, 95% CI: 1.06-1.80). The results of RCS indicated that when the coagulation disorder score >2, the trend of in-hospital mortality rose gradually, and OR was >1. CONCLUSIONS: Poor coagulation function was associated with increased risk of in-hospital mortality in ARDS patients. The findings implied that clinicians should regularly detect the levels of coagulation-associated biomarkers for the management of ARDS patients.

Development of TSHR-CAR NK-92 cells for Differentiated Thyroid Cancer.

Differentiated thyroid cancer (DTC) is the predominant type of thyroid cancer, with some patients experiencing relapse, distant metastases, or refractoriness, revealing limited treatment options. Chimeric antigen receptor (CAR)-modified Natural Killer (NK) cells are revolutionary therapeutic agents effective against various resistant cancers. Thyroid-stimulating hormone receptor (TSHR) expression in DTC provides a unique tumor-specific target for CAR therapy. Here, we developed an innovative strategy for treating DTC using modified NK-92 cells armed with a TSHR-targeted CAR. The modified cells showed enhanced cytotoxicity against TSHR-positive DTC cell lines and exhibited elevated degranulation and cytokine release. After undergoing irradiation, the cells effectively halted their proliferative capacity while maintaining potent targeted killing ability. Transfer of these irradiation-treated cells into NSG mice with DTC tumors resulted in profound tumor suppression. NK-92 cells modified with TSHR-CAR offer a promising, off-the-shelf option for advancing DTC immunotherapy.

Rice ins(3)P synthase1 (RINO1) participates in embryonic development by regulating inositol-associated changes in auxin synthesis and its distribution.

The breeding of low phytic acid (LPA) crops is widely considered an effective strategy to improve crop nutrition, but the LPA crops usually have inferior seed germination performance. To clarify the reason for the suboptimal seed performance of LPA rice, this study investigated the impact of reduced seed phytic acid (InsP6) content in rice ins(3)P synthase1 (EC 5.5.1.4, RINO1), one of the key targets for engineering LPA rice, knockouton cellular differentiation in seed embryos and its relation to myo-inositol metabolism and auxin signalling during embryogenesis. The results indicated that the homozygotes of RINO1 knockout could initiate differentiation at the early stage of embryogenesis but failed to form normal differentiation of plumule and radicle primordia. The loss of RINO1 function disrupted vesicle trafficking and auxin signalling due to the significantly lowered phosphatidylinositides (PIs) concentration in seed embryos, thereby leading to the defects of seed embryos without the recognizable differentiation of shoot apex meristem (SAM) and radicle apex meristem (RAM) for the homozygotes of RINO1 knockout. The abnormal embryo phenotype of RINO1 homozygotes was partially rescued by exogenous spraying of inositol and indole-3-acetic acid (IAA) in rice panicle. Thus, RINO1 is crucial for both seed InsP6 biosynthesis and embryonic development. The lower phosphatidylinositol (4,5)-bisphosphate (PI (4,5) P2) concentration and the disorder auxin distribution induced by insufficient inositol supply in seed embryos were among the regulatory switch steps leading to aberrant embryogenesis and failure of seed germination in RINO1 knockout.

New insights into changes in phosphorus profile at sediment-water interface by microplastics: Role of benthic bioturbation.

Microplastics (MPs) have attracted increasing attention due to their ubiquitous occurrence in freshwater sediments and the detrimental effects on benthic invertebrates. However, a clear understanding of their downstream impacts on ecosystem services is still lacking. This study examines the effects of bio-based polylactic acid (PLA), fuel-based polyethylene terephthalate (PET), and biofilm-covered PET (BPET) MPs on the bioturbator chironomid larvae (Tanypus chinensis), and the influence on phosphorus (P) profiles in microcosms. The changes in biochemical responses and metabolic pathways indicated that MPs disrupted energy synthesis by causing intestinal blockage and oxidative stress in T. chinensis, leading to energy depletion and impaired bioturbation activity. The impairment further resulted in enhanced sedimentary P immobilization. For larval treatments, the internal-P loadings were respectively 11.4%, 8.6%, and 9.0% higher in the PLA, PET, and BPET groups compared to the non-MP control. Furthermore, the influence of bioturbation on P profiles was MP-type dependent. Both BPET and PLA treatments displayed more obvious impacts on P profiles compared to PET due to the changes in MP bioavailability or sediment microenvironment. This study connects individual physiological responses to broader ecosystem services, showing that MPs alter P biogeochemical processes by disrupting the bioturbation activities of chironomid larvae.

How well do neural signatures of resting-state EEG detect consciousness? A large-scale clinical study.

The assessment of consciousness states, especially distinguishing minimally conscious states (MCS) from unresponsive wakefulness states (UWS), constitutes a pivotal role in clinical therapies. Despite that numerous neural signatures of consciousness have been proposed, the effectiveness and reliability of such signatures for clinical consciousness assessment still remains an intense debate. Through a comprehensive review of the literature, inconsistent findings are observed about the effectiveness of diverse neural signatures. Notably, the majority of existing studies have evaluated neural signatures on a limited number of subjects (usually below 30), which may result in uncertain conclusions due to small data bias. This study presents a systematic evaluation of neural signatures with large-scale clinical resting-state electroencephalography (EEG) signals containing 99 UWS, 129 MCS, 36 emergence from the minimally conscious state, and 32 healthy subjects (296 total) collected over 3 years. A total of 380 EEG-based metrics for consciousness detection, including spectrum features, nonlinear measures, functional connectivity, and graph-based measures, are summarized and evaluated. To further mitigate the effect of data bias, the evaluation is performed with bootstrap sampling so that reliable measures can be obtained. The results of this study suggest that relative power in alpha and delta serve as dependable indicators of consciousness. With the MCS group, there is a notable increase in the phase lag index-related connectivity measures and enhanced functional connectivity between brain regions in comparison to the UWS group. A combination of features enables the development of an automatic detector of conscious states.

Efficient spiking neural network design via neural architecture search.

Spiking neural networks (SNNs) are brain-inspired models that utilize discrete and sparse spikes to transmit information, thus having the property of energy efficiency. Recent advances in learning algorithms have greatly improved SNN performance due to the automation of feature engineering. While the choice of neural architecture plays a significant role in deep learning, the current SNN architectures are mainly designed manually, which is a time-consuming and error-prone process. In this paper, we propose a spiking neural architecture search (NAS) method that can automatically find efficient SNNs. To tackle the challenge of long search time faced by SNNs when utilizing NAS, the proposed NAS encodes candidate architectures in a branchless spiking supernet which significantly reduces the computation requirements in the search process. Considering that real-world tasks prefer efficient networks with optimal accuracy under a limited computational budget, we propose a Synaptic Operation (SynOps)-aware optimization to automatically find the computationally efficient subspace of the supernet. Experimental results show that, in less search time, our proposed NAS can find SNNs with higher accuracy and lower computational cost than state-of-the-art SNNs. We also conduct experiments to validate the search process and the trade-off between accuracy and computational cost.

Trainable Spiking-YOLO for low-latency and high-performance object detection.

Spiking neural networks (SNNs) are considered an attractive option for edge-side applications due to their sparse, asynchronous and event-driven characteristics. However, the application of SNNs to object detection tasks faces challenges in achieving good detection accuracy and high detection speed. To overcome the aforementioned challenges, we propose an end-to-end Trainable Spiking-YOLO (Tr-Spiking-YOLO) for low-latency and high-performance object detection. We evaluate our model on not only frame-based PASCAL VOC dataset but also event-based GEN1 Automotive Detection dataset, and investigate the impacts of different decoding methods on detection performance. The experimental results show that our model achieves competitive/better performance in terms of accuracy, latency and energy consumption compared to similar artificial neural network (ANN) and conversion-based SNN object detection model. Furthermore, when deployed on an edge device, our model achieves a processing speed of approximately from 14 to 39 FPS while maintaining a desirable mean Average Precision (mAP), which is capable of real-time detection on resource-constrained platforms.

Cross-Subject Motor Imagery Decoding by Transfer Learning of Tactile ERD.

For Brain-Computer Interface (BCI) based on motor imagery (MI), the MI task is abstract and spontaneous, presenting challenges in measurement and control and resulting in a lower signal-to-noise ratio. The quality of the collected MI data significantly impacts the cross-subject calibration results. To address this challenge, we introduce a novel cross-subject calibration method based on passive tactile afferent stimulation, in which data induced by tactile stimulation is utilized to calibrate transfer learning models for cross-subject decoding. During the experiments, tactile stimulation was applied to either the left or right hand, with subjects only required to sense tactile stimulation. Data from these tactile tasks were used to train or fine-tune models and subsequently applied to decode pure MI data. We evaluated BCI performance using both the classical Common Spatial Pattern (CSP) combined with the Linear Discriminant Analysis (LDA) algorithm and a state-of-the-art deep transfer learning model. The results demonstrate that the proposed calibration method achieved decoding performance at an equivalent level to traditional MI calibration, with the added benefit of outperforming traditional MI calibration with fewer trials. The simplicity and effectiveness of the proposed cross-subject tactile calibration method make it valuable for practical applications of BCI, especially in clinical settings.

DiffMDD: A Diffusion-Based Deep Learning Framework for MDD Diagnosis Using EEG.

Major Depression Disorder (MDD) is a common yet destructive mental disorder that affects millions of people worldwide. Making early and accurate diagnosis of it is very meaningful. Recently, EEG, a non-invasive technique of recording spontaneous electrical activity of brains, has been widely used for MDD diagnosis. However, there are still some challenges in data quality and data size of EEG: (1) A large amount of noise is inevitable during EEG collection, making it difficult to extract discriminative features from raw EEG; (2) It is difficult to recruit a large number of subjects to collect sufficient and diverse data for model training. Both of the challenges cause the overfitting problem, especially for deep learning methods. In this paper, we propose DiffMDD, a diffusion-based deep learning framework for MDD diagnosis using EEG. Specifically, we extract more noise-irrelevant features to improve the model's robustness by designing the Forward Diffusion Noisy Training Module. Then we increase the size and diversity of data to help the model learn more generalized features by designing the Reverse Diffusion Data Augmentation Module. Finally, we re-train the classifier on the augmented dataset for MDD diagnosis. We conducted comprehensive experiments to test the overall performance and each module's effectiveness. The framework was validated on two public MDD diagnosis datasets, achieving the state-of-the-art performance.

Etiologies and clinical characteristics of acute ataxia in a single national children's medical center.

OBJECTIVE: To analyze etiologic factors of pediatric acute ataxia and to identify the severity of its underlying causes for urgent medical intervention. METHODS: Clinical data of children diagnosed with acute ataxia between December 2015 and December 2021 from one national medical center were analyzed retrospectively. RESULTS: A total of 99 children (59 boys, 40 girls), median age at disease onset 55 (range: 12-168) months, were enrolled. The median follow period was 46 (range 6-78) months. Eighty-six (86.9 %) children were diagnosed with immune-associated acute ataxia, among which acute post-infectious cerebellar ataxia (APCA) was the most common diagnosis (50.5 %), followed by demyelinating diseases of the central nervous system (18.2 %) and Guillain-Barré syndrome (9.1 %). On cerebrospinal fluid (CSF) examination, 35/73 (47.9 %) patients had pleocytosis (>5 cells/mm3), and 18/73 (24.7 %) had elevated protein levels. Thirty-one patients (31.3 %) had an abnormal cerebral MRI. Children with other immune-associated acute cerebellar ataxia had more extracerebellar symptoms, intracranial MRI lesions, abnormal CSF results, longer hospital stay, higher recurrence rates and incidence of neurological sequelae than children with APCA. CONCLUSION: Immune-associated acute ataxia is the main cause of pediatric acute ataxia, among which APCA is the most common phenotype. However, some immune-associated diseases, especially autoantibody-mediated disease, which has a higher recurrence rate and neurological sequelae account for an increasing proportion of pediatric acute ataxia. When children present with extracerebellar symptoms, abnormal cranial MRI or CSF results, and without prodromal infection, prudent differential diagnosis is recommended.

A Self-Adaptive Dual-ILRO Clock-Recovery Technique for Two-Tone Battery-Free Crystal-Free Neural-Recording SoC.

Wireless implantable devices are widely used in medical treatment, which should meet clinical constraints such as longevity, miniaturization, and reliable communication. Wireless power transfer (WPT) can eliminate the battery to reduce system size and prolong device life, while it's challenging to generate a reliable clock without a crystal. In this work, we propose a self-adaptive dual-injection-locked-ring-oscillator (dual-ILRO) clock-recovery technique based on two-tone WPT and integrate it into a battery-free neural-recording SoC. The 2[Formula: see text]-order inter-modulation (IM2) component of the two WPT tones is extracted as a low-frequency reference for battery-free SoC, and the proposed self-adaptive dual-ILRO technique extends the lock range to ensure an anti-interference PVT-robust clock generation. The neural-recording SoC includes a low-noise signal acquisition unit, a power management unit, and a backscatter circuit to perform neural signal recording, wireless power harvesting, and neural data transmission. Benefiting from the 6.4 µW low power of the clock recovery circuit, the overall SoC power is cut down to 49.8 µW. In addition, the proposed clock-recovery technique enables both signal acquisition and uplink communication to perform as well as that synchronized by an ideal clock, i.e., an effective number of 9.6 bits and a bit error rate (BER) less than 4.8 × 10-7 in chip measurement. The SoC takes a die area of 2.05 mm 2, and an animal test is conducted in a Sprague-Dawley rat to validate the wireless neural-recording performance, compared to a crystal-synchronized commercial chip.

SPLiT: Single Portrait Lighting Estimation via a Tetrad of Face Intrinsics.

This paper proposes a novel pipeline to estimate a non-parametric environment map with high dynamic range from a single human face image. Lighting-independent and -dependent intrinsic images of the face are first estimated separately in a cascaded network. The influence of face geometry on the two lighting-dependent intrinsics, diffuse shading and specular reflection, are further eliminated by distributing the intrinsics pixel-wise onto spherical representations using the surface normal as indices. This results in two representations simulating images of a diffuse sphere and a glossy sphere under the input scene lighting. Taking into account the distinctive nature of light sources and ambient terms, we further introduce a two-stage lighting estimator to predict both accurate and realistic lighting from these two representations. Our model is trained supervisedly on a large-scale and high-quality synthetic face image dataset. We demonstrate that our method allows accurate and detailed lighting estimation and intrinsic decomposition, outperforming state-of-the-art methods both qualitatively and quantitatively on real face images.

Risk stratification of papillary thyroid cancers using multidimensional machine learning.

BACKGROUND: Papillary thyroid cancer (PTC) is one of the most common endocrine malignancies with different risk levels. However, preoperative risk assessment of PTC is still a challenge in the worldwide. Here, the authors first report a Preoperative Risk Assessment Classifier for PTC (PRAC-PTC) by multidimensional features including clinical indicators, immune indices, genetic feature, and proteomics. MATERIALS AND METHODS: The 558 patients collected from June 2013 to November 2020 were allocated to three groups: the discovery set [274 patients, 274 formalin-fixed paraffin-embedded (FFPE)], the retrospective test set (166 patients, 166 FFPE), and the prospective test set (118 patients, 118 fine-needle aspiration). Proteomic profiling was conducted by FFPE and fine-needle aspiration tissues from the patients. Preoperative clinical information and blood immunological indices were collected. The BRAFV600E mutation were detected by the amplification refractory mutation system. RESULTS: The authors developed a machine learning model of 17 variables based on the multidimensional features of 274 PTC patients from a retrospective cohort. The PRAC-PTC achieved areas under the curve (AUC) of 0.925 in the discovery set and was validated externally by blinded analyses in a retrospective cohort of 166 PTC patients (0.787 AUC) and a prospective cohort of 118 PTC patients (0.799 AUC) from two independent clinical centres. Meanwhile, the preoperative predictive risk effectiveness of clinicians was improved with the assistance of PRAC-PTC, and the accuracies reached at 84.4% (95% CI: 82.9-84.4) and 83.5% (95% CI: 82.2-84.2) in the retrospective and prospective test sets, respectively. CONCLUSION: This study demonstrated that the PRAC-PTC that integrating clinical data, gene mutation information, immune indices, high-throughput proteomics and machine learning technology in multicentre retrospective and prospective clinical cohorts can effectively stratify the preoperative risk of PTC and may decrease unnecessary surgery or overtreatment.

The application of subcapsular saline injection to preserve the parathyroid gland during thyroidectomy.

OBJECTIVE: To evaluate the clinical effect of subcapsular saline injection (SCASI) after total thyroidectomy. METHODS: A total of 77 patients who underwent total thyroidectomy in our hospital from January 2020 to December 2021 were selected and divided into the SCASI group (n = 43) and the non-SCASI group (n = 34). The general clinical data of the patients were collected, and serum parathyroid hormone (PTH) and serum calcium levels were determined preoperatively, on the 1st postoperative day, and at 1 and 6 months after the operation. These data were compared between groups. RESULTS: There was no significant difference in postoperative complications between the two groups. The PTH and serum calcium levels in the SCASI group were significantly higher than those in the non-SCASI group on the 1st postoperative day (t = 2.340, 5.208, both P < 0.05), and the PTH levels in the SCASI group at 1 month after the operation were higher than those in the non-SCASI group (t = 2.141, P < 0.05). In addition, the proportion of transient and permanent hypoparathyroidism in the SCASI group was significantly decreased (χ2 = 3.920, 3.948, P < 0.05). CONCLUSION: Total thyroidectomy requires high surgical precision, and SCASI can reduce the incidence of temporary and permanent hypoparathyroidism.

Ethanol suppresses rice seed germination through inhibiting ROS signaling.

Ethanol is frequently used not only as priming but also as a solvent to dissolve hardly water-soluble phytohormones gibberellic acid (GA3) and abscisic acid (ABA) in seed germination. However, the molecular and physiological mechanisms of ethanol's impact on seed germination remain elusive. In this report, we investigated how ethanol affected reactive oxygen species (ROS) during rice seed germination. Ethanol at a concentration of 3.5% (v/v) inhibited 90% seed germination, which was almost reversed by H2O2. H2O2 contents in embryos were reduced by ethanol after 18 h imbibition. Antioxidant enzymes assays revealed that only superoxide dismutase (SOD) activities in seed embryos were lowered by ethanol, in line with the suppressed mRNA expression of SOD genes during imbibition. Additionally, compared to the mock condition, ethanol increased ABA contents but decreased GA (GA1 and GA3) in seed embryos, resulting in disharmonizing GA/ABA balance. Conceivably ethanol induced transcription of OsNCEDs, the key genes for ABA biosynthesis, and OsABA8ox3, a key gene for ABA catabolism. Furthermore, ethanol promoted ABA signaling by upregulating ABA receptor genes and ABA-responsive element (ABRE)-binding protein/ABRE-binding factors during imbibition. Overall, our results demonstrate that lowering of H2O2 levels due to suppressed SOD activities in rice germinating seed embryos is the decisive factor for ethanol-induced inhibition of seed germination, and GA/ABA balance and ABA signaling also play important roles in ethanol's inhibitory impact on seed germination.

Open-loop analog programmable electrochemical memory array.

Emerging memories have been developed as new physical infrastructures for hosting neural networks owing to their low-power analog computing characteristics. However, accurately and efficiently programming devices in an analog-valued array is still largely limited by the intrinsic physical non-idealities of the devices, thus hampering their applications in in-situ training of neural networks. Here, we demonstrate a passive electrochemical memory (ECRAM) array with many important characteristics necessary for accurate analog programming. Different image patterns can be open-loop and serially programmed into our ECRAM array, achieving high programming accuracies without any feedback adjustments. The excellent open-loop analog programmability has led us to in-situ train a bilayer neural network and reached software-like classification accuracy of 99.4% to detect poisonous mushrooms. The training capability is further studied in simulation for large-scale neural networks such as VGG-8. Our results present a new solution for implementing learning functions in an artificial intelligence hardware using emerging memories.

Fast-SNN: Fast Spiking Neural Network by Converting Quantized ANN.

Spiking neural networks (SNNs) have shown advantages in computation and energy efficiency over traditional artificial neural networks (ANNs) thanks to their event-driven representations. SNNs also replace weight multiplications in ANNs with additions, which are more energy-efficient and less computationally intensive. However, it remains a challenge to train deep SNNs due to the discrete spiking function. A popular approach to circumvent this challenge is ANN-to-SNN conversion. However, due to the quantization error and accumulating error, it often requires lots of time steps (high inference latency) to achieve high performance, which negates SNN's advantages. To this end, this paper proposes Fast-SNN that achieves high performance with low latency. We demonstrate the equivalent mapping between temporal quantization in SNNs and spatial quantization in ANNs, based on which the minimization of the quantization error is transferred to quantized ANN training. With the minimization of the quantization error, we show that the sequential error is the primary cause of the accumulating error, which is addressed by introducing a signed IF neuron model and a layer-wise fine-tuning mechanism. Our method achieves state-of-the-art performance and low latency on various computer vision tasks, including image classification, object detection, and semantic segmentation. Codes are available at: https://github.com/yangfan-hu/Fast-SNN.

NeuSort: an automatic adaptive spike sorting approach with neuromorphic models.

Objective.Spike sorting, a critical step in neural data processing, aims to classify spiking events from single electrode recordings based on different waveforms. This study aims to develop a novel online spike sorter, NeuSort, using neuromorphic models, with the ability to adaptively adjust to changes in neural signals, including waveform deformations and the appearance of new neurons.Approach.NeuSort leverages a neuromorphic model to emulate template-matching processes. This model incorporates plasticity learning mechanisms inspired by biological neural systems, facilitating real-time adjustments to online parameters.Results.Experimental findings demonstrate NeuSort's ability to track neuron activities amidst waveform deformations and identify new neurons in real-time. NeuSort excels in handling non-stationary neural signals, significantly enhancing its applicability for long-term spike sorting tasks. Moreover, its implementation on neuromorphic chips guarantees ultra-low energy consumption during computation.Significance.NeuSort caters to the demand for real-time spike sorting in brain-machine interfaces through a neuromorphic approach. Its unsupervised, automated spike sorting process makes it a plug-and-play solution for online spike sorting.