Lightweight prohibited item detection method based on YOLOV4 for x-ray security inspection.

In the area of public safety and crime prevention, some research based on deep learning has achieved success in the detection of prohibited items for x-ray security inspection. However, the number of parameters and computational consumption of most object detection methods based on deep learning are huge, which makes the hardware requirements of these methods extremely high and limits their applications. In this paper, a lightweight prohibited item detection method based on YOLOV4 is proposed for x-ray security inspection. First, the MobilenetV3 is used to replace the backbone network of YOLOV4, and the depthwise separable convolution is used to optimize the neck and head of YOLOV4 to reduce the number of parameters and computational consumption. Second, an adaptive spatial-and-channel attention block is designed to optimize the neck of YOLOV4 in order to improve the feature extraction capability of our method and maintain the detection accuracy. Third, the focal loss is utilized to avoid the class imbalance problem during the training process. Finally, the method is evaluated on our real x-ray pseudocolor image dataset with YOLOV4 and YOLOV4-tiny. For the overall performance, the mean average precision of our method is 4.98% higher than YOLOV4-tiny and 0.07% lower than YOLOV4. The number of parameters and computational consumption of our method are slightly higher than YOLOV4-tiny and much lower than YOLOV4.

Multi-strategy improved salp swarm algorithm and its application in reliability optimization.

To improve the convergence speed and solution precision of the standard Salp Swarm Algorithm (SSA), a hybrid Salp Swarm Algorithm based on Dimension-by-dimension Centroid Opposition-based learning strategy, Random factor and Particle Swarm Optimization's social learning strategy (DCORSSA-PSO) is proposed. Firstly, a dimension-by-dimension centroid opposition-based learning strategy is added in the food source update stage of SSA to increase the population diversity and reduce the inter-dimensional interference. Secondly, in the followers' position update equation of SSA, constant 1 is replaced by a random number between 0 and 1 to increase the randomness of the search and the ability to jump out of local optima. Finally, the social learning strategy of PSO is also added to the followers' position update equation to accelerate the population convergence. The statistical results on ten classical benchmark functions by the Wilcoxon test and Friedman test show that compared with SSA and other well-known optimization algorithms, the proposed DCORSSA-PSO has significantly improved the precision of the solution and the convergence speed, as well as its robustness. The DCORSSA-PSO is applied to system reliability optimization design based on the T-S fault tree. The simulation results show that the failure probability of the designed system under the cost constraint is less than other algorithms, which illustrates that the application of DCORSSA-PSO can effectively improve the design level of reliability optimization.

A Data Augmentation Method for Prohibited Item X-Ray Pseudocolor Images in X-Ray Security Inspection Based on Wasserstein Generative Adversarial Network and Spatial-and-Channel Attention Block.

For public security and crime prevention, the detection of prohibited items in X-ray security inspection based on deep learning has attracted widespread attention. However, the pseudocolor image dataset is scarce due to security, which brings an enormous challenge to the detection of prohibited items in X-ray security inspection. In this paper, a data augmentation method for prohibited item X-ray pseudocolor images in X-ray security inspection is proposed. Firstly, we design a framework of our method to achieve the dataset augmentation using the datasets with and without prohibited items. Secondly, in the framework, we design a spatial-and-channel attention block and a new base block to compose our X-ray Wasserstein generative adversarial network model with gradient penalty. The model directly generates high-quality dual-energy X-ray data instead of pseudocolor images. Thirdly, we design a composite strategy to composite the generated and real dual-energy X-ray data with background data into a new X-ray pseudocolor image, which can simulate the real overlapping relationship among items. Finally, two object detection models with and without our data augmentation method are applied to verify the effectiveness of our method. The experimental results demonstrate that our method can achieve the data augmentation for prohibited item X-ray pseudocolor images in X-ray security inspection effectively.

A Lightweight Dangerous Liquid Detection Method Based on Depthwise Separable Convolution for X-Ray Security Inspection.

For personal safety and crime prevention, some research studies based on deep learning have achieved success in the object detection of X-ray security inspection. However, the research on dangerous liquid detection is still scarce, and most research studies are focused on the detection of some prohibited and common items. In this paper, a lightweight dangerous liquid detection method based on the Depthwise Separable convolution for X-ray security inspection is proposed. Firstly, a dataset of seven common dangerous liquids with multiple postures in two detection environments is established. Secondly, we propose a novel detection framework using the dual-energy X-ray data instead of pseudocolor images as the objects to be detected, which improves the detection accuracy and realizes the parallel operation of detection and imaging. Thirdly, in order to ensure the detection accuracy and reduce the computational consumption and the number of parameters, based on the Depthwise Separable convolution and the Squeeze-and-Excitation block, a lightweight object location network and a lightweight dangerous liquid classification network are designed as the backbone networks of our method to achieve the location and classification of the dangerous liquids, respectively. Finally, a semiautomatic labeling method is proposed to improve the efficiency of data labeling. Compared with the existing methods, the experimental results demonstrate that our method has better performance and wider applicability.

Closed-loop time-varying continuous-time recursive subspace-based prediction via principle angles rotation.

This paper presents a closed-loop time-varying continuous-time recursive subspace-based prediction method utilizing principle angles rotation. A simple linear mapping can be provided by generalized Poisson moment functionals, which can deal with the time-derivatives problems of input-output Hankel matrices. The parity space employed in fault detection field is adopted instead of using the observable subspace. The system matrices are estimated consistently by the instrumental variable method and principal component analysis, which solves the identification problems of biased results for the system operating in closed-loop with a feedback controller. The system matrices are predicted by the principle angles rotation of the signal subspaces spanned from the extended observability matrices. The effectiveness of the proposed method is illustrated by the numerical simulations and real applications.

Cullin 4B regulates cell survival and apoptosis in clear cell renal cell carcinoma as a target of microRNA-217.

Cullin 4B (CUL4B) was reported to be closely related to the progression of some tumors, but its function in clear cell renal cell carcinoma (ccRCC) has not been reported. Our present study found CUL4B was upregulated in ccRCC, and CUL4B knockdown markedly inhibited ccRCC cell growth and induced apoptosis. In addition, CUL4B knockdown markedly inhibited antiapoptotic proteins' expression in ccRCC cells, including Mcl-1 and Bcl-2, and silenced CUL4B also induced the cleavages of PARP, an important index of apoptosis. We also confirmed microRNA-217 (miR-217) was downregulated in ccRCC tumor tissues, and negatively correlated with CUL4B expression. Further investigations revealed miR-217 targeted CUL4B and markedly inhibited its expression in ccRCC cells. In addition, overexpression of miR-217 by mimics significantly suppressed ccRCC cell growth. In contrast, enforced expression of CUL4B significantly abolished miR-217-induced cell survival inhibition in ccRCC cells. In conclusion, our present results suggested targeting miR-217-CUL4B axis would be a promising strategy for ccRCC treatment.

1,1-Difluoroethyl chloride (CH3CF2Cl), a novel difluoroalkylating reagent for 1,1-difluoroethylation of arylboronic acids.

1,1-Difluoroethylated aromatics are of great importance in medicinal chemistry and related fields. 1,1-Difluoroethyl chloride (CH3CF2Cl), a cheap and abundant industrial raw material, is viewed as an ideal 1,1-difluoroethylating reagent, but the direct introduction of the difluoroethyl (CF2CH3) group onto aromatic rings using CH3CF2Cl has not been successfully accomplished. Herein, we disclose a nickel-catalyzed 1,1-difluoroethylation of arylboronic acids with CH3CF2Cl for the synthesis of (1,1-difluoroethyl)arenes.

Nuclear norm subspace identification for continuous-time stochastic systems based on distribution theory method.

A novel method for nuclear norm subspace identification of continuous-time stochastic systems based on distribution theory is proposed. The time-derivative problem of the system is solved by using random distribution theory, which is the key to obtain the input-output algebraic equation in the time-domain. Due to the fact that the system encounters the stochastic noise, we design a Kalman filter to achieve the state estimation and noise reduction. Nuclear norm minimization is constructed to optimize the system order in the process of subspace identification. Further, the optimization problem is solved by the alternating direction method of multipliers. Simulation results are provided to show the effectiveness of the proposed method.

Diversity of fish sound types in the Pearl River Estuary, China.

BACKGROUND: Repetitive species-specific sound enables the identification of the presence and behavior of soniferous species by acoustic means. Passive acoustic monitoring has been widely applied to monitor the spatial and temporal occurrence and behavior of calling species. METHODS: Underwater biological sounds in the Pearl River Estuary, China, were collected using passive acoustic monitoring, with special attention paid to fish sounds. A total of 1,408 suspected fish calls comprising 18,942 pulses were qualitatively analyzed using a customized acoustic analysis routine. RESULTS: We identified a diversity of 66 types of fish sounds. In addition to single pulse, the sounds tended to have a pulse train structure. The pulses were characterized by an approximate 8 ms duration, with a peak frequency from 500 to 2,600 Hz and a majority of the energy below 4,000 Hz. The median inter-pulsepeak interval (IPPI) of most call types was 9 or 10 ms. Most call types with median IPPIs of 9 ms and 10 ms were observed at times that were exclusive from each other, suggesting that they might be produced by different species. According to the literature, the two section signal types of 1 + 1 and 1 + N10 might belong to big-snout croaker (Johnius macrorhynus), and 1 + N19 might be produced by Belanger's croaker (J. belangerii). DISCUSSION: Categorization of the baseline ambient biological sound is an important first step in mapping the spatial and temporal patterns of soniferous fishes. The next step is the identification of the species producing each sound. The distribution pattern of soniferous fishes will be helpful for the protection and management of local fishery resources and in marine environmental impact assessment. Since the local vulnerable Indo-Pacific humpback dolphin (Sousa chinensis) mainly preys on soniferous fishes, the fine-scale distribution pattern of soniferous fishes can aid in the conservation of this species. Additionally, prey and predator relationships can be observed when a database of species-identified sounds is completed.

Apparent source levels and active communication space of whistles of free-ranging Indo-Pacific humpback dolphins (Sousa chinensis) in the Pearl River Estuary and Beibu Gulf, China.

Background. Knowledge of species-specific vocalization characteristics and their associated active communication space, the effective range over which a communication signal can be detected by a conspecific, is critical for understanding the impacts of underwater acoustic pollution, as well as other threats. Methods. We used a two-dimensional cross-shaped hydrophone array system to record the whistles of free-ranging Indo-Pacific humpback dolphins (Sousa chinensis) in shallow-water environments of the Pearl River Estuary (PRE) and Beibu Gulf (BG), China. Using hyperbolic position fixing, which exploits time differences of arrival of a signal between pairs of hydrophone receivers, we obtained source location estimates for whistles with good signal-to-noise ratio (SNR ≥10 dB) and not polluted by other sounds and back-calculated their apparent source levels (ASL). Combining with the masking levels (including simultaneous noise levels, masking tonal threshold, and the Sousa auditory threshold) and the custom made site-specific sound propagation models, we further estimated their active communication space (ACS). Results. Humpback dolphins produced whistles with average root-mean-square ASL of 138.5 ± 6.8 (mean ± standard deviation) and 137.2 ± 7.0 dB re 1 µPa in PRE (N = 33) and BG (N = 209), respectively. We found statistically significant differences in ASLs among different whistle contour types. The mean and maximum ACS of whistles were estimated to be 14.7 ± 2.6 (median ± quartile deviation) and 17.1± 3.5 m in PRE, and 34.2 ± 9.5 and 43.5 ± 12.2 m in BG. Using just the auditory threshold as the masking level produced the mean and maximum ACSat of 24.3 ± 4.8 and 35.7 ± 4.6 m for PRE, and 60.7 ± 18.1 and 74.3 ± 25.3 m for BG. The small ACSs were due to the high ambient noise level. Significant differences in ACSs were also observed among different whistle contour types. Discussion. Besides shedding some light for evaluating appropriate noise exposure levels and information for the regulation of underwater acoustic pollution, these baseline data can also be used for aiding the passive acoustic monitoring of dolphin populations, defining the boundaries of separate groups in a more biologically meaningful way during field surveys, and guiding the appropriate approach distance for local dolphin-watching boats and research boat during focal group following.

TLR4 Signaling via NANOG Cooperates With STAT3 to Activate Twist1 and Promote Formation of Tumor-Initiating Stem-Like Cells in Livers of Mice.

BACKGROUND & AIMS: Obesity and alcohol consumption contribute to steatohepatitis, which increases the risk for hepatitis C virus (HCV)-associated hepatocellular carcinomas (HCCs). Mouse hepatocytes that express HCV-NS5A in liver up-regulate the expression of Toll-like receptor 4 (TLR4), and develop liver tumors containing tumor-initiating stem-like cells (TICs) that express NANOG. We investigated whether the TLR4 signals to NANOG to promote the development of TICs and tumorigenesis in mice placed on a Western diet high in cholesterol and saturated fat (HCFD). METHODS: We expressed HCV-NS5A from a transgene (NS5A Tg) in Tlr4-/- (C57Bl6/10ScN), and wild-type control mice. Mice were fed a HCFD for 12 months. TICs were identified and isolated based on being CD133+, CD49f+, and CD45-. We obtained 142 paraffin-embedded sections of different stage HCCs and adjacent nontumor areas from the same patients, and performed gene expression, immunofluorescence, and immunohistochemical analyses. RESULTS: A higher proportion of NS5A Tg mice developed liver tumors (39%) than mice that did not express HCV NS5A after the HCFD (6%); only 9% of Tlr4-/- NS5A Tg mice fed HCFD developed liver tumors. Livers from NS5A Tg mice fed the HCFD had increased levels of TLR4, NANOG, phosphorylated signal transducer and activator of transcription (pSTAT3), and TWIST1 proteins, and increases in Tlr4, Nanog, Stat3, and Twist1 messenger RNAs. In TICs from NS5A Tg mice, NANOG and pSTAT3 directly interact to activate expression of Twist1. Levels of TLR4, NANOG, pSTAT3, and TWIST were increased in HCC compared with nontumor tissues from patients. CONCLUSIONS: HCFD and HCV-NS5A together stimulated TLR4-NANOG and the leptin receptor (OB-R)-pSTAT3 signaling pathways, resulting in liver tumorigenesis through an exaggerated mesenchymal phenotype with prominent Twist1-expressing TICs.

Passive Acoustic Monitoring the Diel, Lunar, Seasonal and Tidal Patterns in the Biosonar Activity of the Indo-Pacific Humpback Dolphins (Sousa chinensis) in the Pearl River Estuary, China.

A growing demand for sustainable energy has led to an increase in construction of offshore windfarms. Guishan windmill farm will be constructed in the Pearl River Estuary, China, which sustains the world's largest known population of Indo-Pacific humpback dolphins (Sousa chinensis). Dolphin conservation is an urgent issue in this region. By using passive acoustic monitoring, a baseline distribution of data on this species in the Pearl River Estuary during pre-construction period had been collected. Dolphin biosonar detection and its diel, lunar, seasonal and tidal patterns were examined using a Generalized Linear Model. Significant higher echolocation detections at night than during the day, in winter-spring than in summer-autumn, at high tide than at flood tide were recognized. Significant higher echolocation detections during the new moon were recognized at night time. The diel, lunar and seasonal patterns for the echolocation encounter duration also significantly varied. These patterns could be due to the spatial-temporal variability of dolphin prey and illumination conditions. The baseline information will be useful for driving further effective action on the conservation of this species and in facilitating later assessments of the effects of the offshore windfarm on the dolphins by comparing the baseline to post construction and post mitigation efforts.

Development of a scalable suspension culture for cardiac differentiation from human pluripotent stem cells.

To meet the need of a large quantity of hPSC-derived cardiomyocytes (CM) for pre-clinical and clinical studies, a robust and scalable differentiation system for CM production is essential. With a human pluripotent stem cells (hPSC) aggregate suspension culture system we established previously, we developed a matrix-free, scalable, and GMP-compliant process for directing hPSC differentiation to CM in suspension culture by modulating Wnt pathways with small molecules. By optimizing critical process parameters including: cell aggregate size, small molecule concentrations, induction timing, and agitation rate, we were able to consistently differentiate hPSCs to >90% CM purity with an average yield of 1.5 to 2×10(9) CM/L at scales up to 1L spinner flasks. CM generated from the suspension culture displayed typical genetic, morphological, and electrophysiological cardiac cell characteristics. This suspension culture system allows seamless transition from hPSC expansion to CM differentiation in a continuous suspension culture. It not only provides a cost and labor effective scalable process for large scale CM production, but also provides a bioreactor prototype for automation of cell manufacturing, which will accelerate the advance of hPSC research towards therapeutic applications.

Ultra-fast formation control of high-order discrete-time multi-agent systems based on multi-step predictive mechanism.

This paper deals with the ultra-fast formation control problem of high-order discrete-time multi-agent systems. Using the local neighbor-error knowledge, a novel ultra-fast protocol with multi-step predictive information and self-feedback term is proposed. The asymptotic convergence factor is improved by a power of q+1 compared to the routine protocol. To some extent, the ultra-fast algorithm overcomes the influence of communication topology to the convergence speed. Furthermore, some sufficient conditions are given herein. The ones decouple the design of the synchronizing gains from the detailed graph properties, and explicitly reveal how the agent dynamic and the communication graph jointly affect the ultra-fast formationability. Finally, some simulations are worked out to illustrate the effectiveness of our theoretical results.

Trade-off between water pollution prevention, agriculture profit, and farmer practice--an optimization methodology for discussion on land-use adjustment in China.

Agricultural decision-making to control nonpoint source (NPS) water pollution may not be efficiently implemented, if there is no appropriate cost-benefit analysis on agricultural management practices. This paper presents an interval-fuzzy linear programming (IFLP) model to deal with the trade-off between agricultural revenue, NPS pollution control, and alternative practices through land adjustment for Wuchuan catchment, a typical agricultural area in Jiulong River watershed, Fujian Province of China. From the results, the lower combination of practice 1, practice 2, practice 3, and practice 7 with the land area of 12.6, 5.2, 145.2, and 85.3 hm(2), respectively, could reduce NPS pollution load by 10%. The combination yields an income of 98,580 Chinese Yuan/a. If the pollution reduction is 15%, the higher combination need practice 1, practice 2, practice 3, practice 5, and practice 7 with the land area of 54.4, 23.6, 18.0, 6.3, and 85.3 hm(2), respectively. The income of this combination is 915,170 Chinese Yuan/a. The sensitivity analysis of IFLP indicates that the cost-effective practices are ranked as follows: practice 7 > practice 2 > practice 1 > practice 5 > practice 3 > practice 6 > practice 4. In addition, the uncertainties in the agriculture NPS pollution control system could be effectively quantified by the IFLP model. Furthermore, to accomplish a reasonable and applicable project of land-use adjustment, decision-makers could also integrate above solutions with their own experience and other information.

Assessing the underwater acoustics of the world's largest vibration hammer (OCTA-KONG) and its potential effects on the Indo-Pacific humpbacked dolphin (Sousa chinensis).

Anthropogenic noise in aquatic environments is a worldwide concern due to its potential adverse effects on the environment and aquatic life. The Hongkong-Zhuhai-Macao Bridge is currently under construction in the Pearl River Estuary, a hot spot for the Indo-Pacific humpbacked dolphin (Sousa chinensis) in China. The OCTA-KONG, the world's largest vibration hammer, is being used during this construction project to drive or extract steel shell piles 22 m in diameter. This activity poses a substantial threat to marine mammals, and an environmental assessment is critically needed. The underwater acoustic properties of the OCTA-KONG were analyzed, and the potential impacts of the underwater acoustic energy on Sousa, including auditory masking and physiological impacts, were assessed. The fundamental frequency of the OCTA-KONG vibration ranged from 15 Hz to 16 Hz, and the noise increments were below 20 kHz, with a dominant frequency and energy below 10 kHz. The resulting sounds are most likely detectable by Sousa over distances of up to 3.5 km from the source. Although Sousa clicks do not appear to be adversely affected, Sousa whistles are susceptible to auditory masking, which may negatively impact this species' social life. Therefore, a safety zone with a radius of 500 m is proposed. Although the zero-to-peak source level (SL) of the OCTA-KONG was lower than the physiological damage level, the maximum root-mean-square SL exceeded the cetacean safety exposure level on several occasions. Moreover, the majority of the unweighted cumulative source sound exposure levels (SSELs) and the cetacean auditory weighted cumulative SSELs exceeded the acoustic threshold levels for the onset of temporary threshold shift, a type of potentially recoverable auditory damage resulting from prolonged sound exposure. These findings may aid in the identification and design of appropriate mitigation methods, such as the use of air bubble curtains, "soft start" and "power down" techniques.

A consensus-based multi-agent approach for estimation in robust fault detection.

This paper is devoted to distributed estimation in robust fault detection for sensor networks with networked-induced delays and packet dropouts by using a consensus-based multi-agent approach. Utilizing the information interaction and coordination among the neighboring networks based on multi-agent theory, we design novel and multiple agent-based robust fault detection filters (RFDFs) subject to only partial estimated and measured information. Asymptotically stable sufficient conditions for the innovative constructed filters are derived in the form of linear matrix inequality (LMI) and the threshold fit for each agent-based RFDF is determined. An illustrative example is given to demonstrate the effectiveness of the consensus-based multi-agent approach for the estimation in robust fault detection.

[Superposition impact character of air pollution from decentralization docks in a freshwater port].

Air pollution from freshwater port is mainly caused by dust pollution, including material loading and unloading dust, road dust, and wind erosion dust from stockpile, bare soil. The dust pollution from a single dock characterized in obvious difference with air pollution from multiple scattered docks. Jining Port of Shandong Province was selected as a case study to get superposition impact contribution of air pollution for regional air environment from multiple scattered docks and to provide technical support for system evaluation of port air pollution. The results indicate that (1) the air pollution from freshwater port occupies a low proportion of pollution impact on regional environmental quality because the port is consisted of serveral small scattered docks; (2) however, the geometric center of the region distributed by docks is severely affected with the most superposition of the air pollution; and (3) the ADMS model is helpful to attain an effective and integrated assessment to predict a superposition impact of multiple non-point pollution sources when the differences of high-altitude weather conditions was not considered on a large scale.

Reciprocal regulation by TLR4 and TGF-ß in tumor-initiating stem-like cells.

Tumor-initiating stem-like cells (TICs) are resistant to chemotherapy and associated with hepatocellular carcinoma (HCC) caused by HCV and/or alcohol-related chronic liver injury. Using HCV Tg mouse models and patients with HCC, we isolated CD133(+) TICs and identified the pluripotency marker NANOG as a direct target of TLR4, which drives the tumor-initiating activity of TICs. These TLR4/NANOG-dependent TICs were defective in the TGF-ß tumor suppressor pathway. Functional oncogene screening of a TIC cDNA library identified Yap1 and Igf2bp3 as NANOG-dependent genes that inactivate TGF-ß signaling. Mechanistically, we determined that YAP1 mediates cytoplasmic retention of phosphorylated SMAD3 and suppresses SMAD3 phosphorylation/activation by the IGF2BP3/AKT/mTOR pathway. Silencing of both YAP1 and IGF2BP3 restored TGF-ß signaling, inhibited pluripotency genes and tumorigenesis, and abrogated chemoresistance of TICs. Mice with defective TGF-ß signaling (Spnb2(+/-) mice) exhibited enhanced liver TLR4 expression and developed HCC in a TLR4-dependent manner. Taken together, these results suggest that the activated TLR4/NANOG oncogenic pathway is linked to suppression of cytostatic TGF-ß signaling and could potentially serve as a therapeutic target for HCV-related HCC.

Cancer stem cells generated by alcohol, diabetes, and hepatitis C virus.

Cancer stem cells (tumor-initiating stem-like cells: TISCs) are resistant to chemotherapy and are associated with metastatic hepatocellular carcinoma (HCC), which is commonly observed in hepatitis C virus (HCV)-infected patients with obesity or alcohol abuse. However, it is unknown whether the TLR4-NANOG pathway serves as a universal oncogenic signaling in the genesis of TISCs and HCC. We aimed to determine whether Tlr4 is a putative proto-oncogene for TISCs in liver oncogenesis due to different etiologies and how Tlr4 is regulated at the transcriptional and epigenetic levels. CD133+/CD49f+ TISCs were isolated using FACS from HCC developed in HCV Core Tg mice fed alcohol, diethylnitrosamine-treated mice, and alcoholic patients with or without HCV infection. CD133+/CD49f+ cells isolated from the animal models and patients are tumorigenic both in vitro and in a xenograft model, and Tlr4 or Nanog silencing with shRNA attenuates their tumor initiating property. Functional oncogene screening of a cDNA library identified the organ size control pathway targets Yap1 and AKT activator Igf2bp3 as NANOG-dependent genes that inhibit transforming growth factor-ß signaling in TISCs. Tlr4 expression is higher in TISCs compared with CD133-/CD49f+ cells. Taken together, Tlr4 may be a universal proto-oncogene responsible for the genesis of TLR4-NANOG dependent TISCs, and this pathway serves as a novel therapeutic target for HCC.