Similar Features Research Articles

EMC+GD_C: circle-based enhanced motion consistency and guided diffusion feature matching for 3D reconstruction

AbstractRobust matching, especially the number, precision and distribution of feature point matching, directly affects the effect of 3D reconstruction. However, the existing methods rarely consider these three aspects comprehensively to improve the quality of feature matching, which in turn affects the effect of 3D reconstruction. Therefore, to effectively improve the quality of 3D reconstruction, we propose a circle-based enhanced motion consistency and guided diffusion feature matching algorithm for 3D reconstruction named EMC+GD_C. Firstly, a circle-based neighborhood division method is proposed, which increases the number of initial matching points. Secondly, to improve the precision of feature point matching, on the one hand, we put forward the idea of enhancing motion consistency, reducing the mismatch of high similarity feature points by enhancing the judgment conditions of true and false matching points; on the other hand, we combine the RANSAC optimization method to filter out the outliers and further improve the precision of feature point matching. Finally, a novel guided diffusion idea combining guided matching and motion consistency is proposed, which expands the distribution range of feature point matching and improves the stability of 3D models. Experiments on 8 sets of 908 pairs of images in the public 3D reconstruction datasets demonstrate that our method can achieve better matching performance and show stronger stability in 3D reconstruction. Specifically, EMC+GD_C achieves an average improvement of 24.07% compared to SIFT-based ratio test, 9.18% to GMS and 1.94% to EMC+GD_G in feature matching precision.

Machine learning and multi-omics data reveal driver gene-based molecular subtypes in hepatocellular carcinoma for precision treatment.

The heterogeneity of Hepatocellular Carcinoma (HCC) poses a barrier to effective treatment. Stratifying highly heterogeneous HCC into molecular subtypes with similar features is crucial for personalized anti-tumor therapies. Although driver genes play pivotal roles in cancer progression, their potential in HCC subtyping has been largely overlooked. This study aims to utilize driver genes to construct HCC subtype models and unravel their molecular mechanisms. Utilizing a novel computational framework, we expanded the initially identified 96 driver genes to 1192 based on mutational aspects and an additional 233 considering driver dysregulation. These genes were subsequently employed as stratification markers for further analyses. A novel multi-omics subtype classification algorithm was developed, leveraging mutation and expression data of the identified stratification genes. This algorithm successfully categorized HCC into two distinct subtypes, CLASS A and CLASS B, demonstrating significant differences in survival outcomes. Integrating multi-omics and single-cell data unveiled substantial distinctions between these subtypes regarding transcriptomics, mutations, copy number variations, and epigenomics. Moreover, our prognostic model exhibited excellent predictive performance in training and external validation cohorts. Finally, a 10-gene classification model for these subtypes identified TTK as a promising therapeutic target with robust classification capabilities. This comprehensive study provides a novel perspective on HCC stratification, offering crucial insights for a deeper understanding of its pathogenesis and the development of promising treatment strategies.

Like Father, Like Daughter - Ectodermal Dysplasia-Syndactyly Syndrome: A Case Report.

Ectodermal dysplasia-syndactyly syndrome 1 (EDSS1) is an exceedingly rare condition associated with mutations in the PVL4 gene. It is characterised by sparse, brittle hair, eyebrows and eyelashes, abnormal dentition and nails, along with bilateral cutaneous syndactyly involving the fingers and toes. We report a 2-year-old girl who presented to us with bilateral complete simple syndactyly of the third and fourth web spaces of the hands, along with bilateral syndactyly of both feet involving the second to fourth toes. Upon examination, sparse hair and eyebrows, along with abnormal dentition, were noted. Thorough clinical examination and genetic analysis were conducted on the affected child and her father, who exhibited similar clinical features. Genetic analysis revealed a homozygous nonsense mutation in the PVL4 gene in both individuals. According to the literature, EDSS1 has been reported in only 10 families worldwide, and there are no reported cases from India. Level of Evidence: Level V (Therapeutic).

Tuberculosis Detection and Air Purifier Suggestion

Tuberculosis (TB) is a disease that can be fatal if not promptly treated. Ensemble deep learning methods have shown promise in aiding the early detection of TB. Previous research typically trained ensemble classifiers using images with similar features, but for optimal performance, an ensemble requires a range of errors, achievable through diverse classification techniques or feature sets. This study focuses on the latter approach, presenting TB detection using deep learning alongside contrast-enhanced canny edge detected (CEED-Canny) X-ray images. CEED-Canny was employed to generate edge-detected lung X-ray images. Two sets of features were derived: one from the enhanced X- ray images and the other from the edge-detected images. By introducing this variation in features, the diversity of errors among the base classifiers was increased, resulting in improved TB detection. The proposed ensemble method achieved a comparable accuracy of 93.59%, sensitivity of 92.31%, and specificity of 94.87% compared to prior research.

Differentiation of hemispheric white matter lesions in migraine and multiple sclerosis with similar radiological features using advanced MRI

Background and aimWhite matter hyperintensities (WMHs), presented on T2-weighted or fluid-attenuated inversion recovery magnetic resonance imaging (MRI) sequences, are lesions in the human brain that can be observed in both migraine and multiple sclerosis (MS).MethodsSeventeen migraine patients and 15 patients with relapsing–remitting multiple sclerosis with WMHs, and 17 healthy subjects age-and sex-matched to the migraine group were prospectively enrolled and underwent conventional and advanced MRI studies with diffusion-and perfusion-weighted imaging and single voxel proton magnetic resonance spectroscopy.ResultsIn both disease groups, elevated T2 relaxation time, apparent diffusion coefficient (ADC) values, and decreased N-acetyl-aspartate levels were found in the intralesional white matter compared to the contralateral normal-appearing white matter (NAWM), while there was no difference between the hemispheres of the control subjects. Migraine patients had the lowest intralesional creatine + phosphocreatine and myo-inositol (mI) values among the three groups, while patients with MS showed the highest intralesional T1 and T2 relaxation times, ADC, and mI values. In the contralateral NAWM, the same trend with mI changes was observed in migraineurs and MS patients. No differences in perfusion variables were observed in any groups.ConclusionOur multimodal study showed that tissue damage is detectable in both diseases. Despite the differences in various advanced MRI measures, with more severe injury detected in MS lesions, we could not clearly differentiate the two white matter lesion types.

RANSAC-Based Planar Point Cloud Segmentation Enhanced by Normal Vector and Maximum Principal Curvature Clustering

Abstract. Planar feature segmentation is an essential task for 3D point cloud processing, finding many applications in various fields such as robotics and computer vision. The Random Sample Consensus (RANSAC) is one of the most common algorithms for the segmentation, but its performance, as given by the original form, is usually limited due to the use of a single threshold and interruption of similar planar features presented close to each other. To address these issues, we present a novel point cloud processing workflow which aims at developing an initial segmentation stage before the basic RANSAC is performed. Initially, normal vectors and maximum principal curvatures for each point of a given point cloud are analyzed and integrated. Subsequently, a subset of normal vectors and curvature is utilized to cluster planes with similar geometry based on the region growing algorithm, serving as a coarse but fast segmentation process. The segmentation is therefore refined with the RANSAC algorithm which can be now performed with higher accuracy and speed due to the reduced interference. After the RANSAC process is applied, resultant planar point clouds are built from the sparse ones via a point aggregation process based on geometric constraints. Four datasets (three real and one simulated) were used to verify the method. Compared to the classic segmentation method, our method achieves higher accuracy, with an RMSE from fitting equal to 0.0521 m, along with a higher recall of 93.31% and a higher F1-score of 95.38%.

Driving Domain Classification Based on Kernel Density Estimation of Urban Land Use and Road Network Scaling Models

Current research on automated driving systems focuses on Level 4 automated driving (AD) in specific operational design Domains (ODD). Measurement data from customer fleet operation are commonly used to extract scenarios and ODD features (road infrastructure, etc.) for the testing of AD functions. To ensure data relevance for the vehicle use case, driving domain classification of the data is required. Generally, classification into urban, extra-urban and highway domains provides data with similar ODD features. Highway classification can be implemented using global navigation satellite system coordinates of the driving route, map-matching algorithms, and road classes stored in digital maps. However, the distinction between urban and extra-urban driving domains is more complex, as settlement taxonomies and administrative-level hierarchies are not globally consistent. Therefore, this paper presents a map-based method for driving domain classification. First, potential urban areas (PUA) are identified based on urban land-use density, which is determined based on land-use categories from OpenStreetMap (OSM) and then spatially smoothed by kernel density estimation. Subsequently, two road network scaling models are used to distinguish between urban and extra-urban domains for the PUA. Finally, statistics of ODD feature distribution are analysed for the classified urban and extra-urban areas.

Heterogeneous transfer learning considering feature representation and environmental consistency for landslide spatial prediction

ABSTRACT The difficult, time-consuming, and imbalanced acquisition of landslide inventories in complex and heterogeneous large areas often results in limited predictive performance for most statistical landslide spatial prediction methods. Although partial transfer learning methods have produced reliable predictive results and successfully implemented knowledge transfer between data-rich and data-scarce areas, most of these methods generally only extract inadequate environmental features and lack the ability to interpret when, what, and how to effectively transfer knowledge to other regions with limited data. In this paper, a heterogeneous transfer learning method considering feature representations and environmental consistency, which features robustness, similarity, and transferability, is proposed for landslide spatial prediction. Specifically, we trained a stacked autoencoder (SAE) to extract more nonlinear features among environmental factors, and added an environmental similarity criterion to the transfer adaptation boosting (TrAdaBoost) algorithm to minimize feature dissimilarities and avoid negative transfer in different scenarios. To evaluate the robustness of the proposed method, we first selected two target areas (Lushan County and Luding County, China) and a source area (Wenchuan County, China) as case study areas. Then, we directly combined the source area and target area as an additional dataset without considering transfer learning to validate the significance and necessity of the proposed method. The area under the receiver operating characteristic curve (AUC) of the two target regions for the proposed method were 0.920 and 0.972, respectively, which were greater than those of the traditional TrAdaBoost (0.909 and 0.969, respectively), SAE (0.790 and 0.937, respectively), and random forest (0.915 and 0.966, respectively) methods. Furthermore, the AUC values of the SAE (0.851 and 0.900) and random forest (0.890 and 0.935) models based on the expanded datasets were also lower than those of the proposed method. Therefore, the experimental results show that the proposed method can be generalized well due to its efficient utilization and high adaptability. Moreover, the proposed method can not only be applied to emergency rescue and disaster prevention but can also offer a promising way to improve landslide predictions of models with incomplete landslide inventories.

Exploiting relation of video segments for temporal action detection

With the development of urban intelligence, intelligent identification of human activities has become imminent. Temporal Action Detection (TAD) is designed to identify real human social activities, which is a challenging task in video understanding. The current methods mainly use global features for boundary matching or predefine all possible proposals, while ignoring the interference of background and the causal relevance of the temporal action, resulting in the generation of more redundant proposals and the decline of detection accuracy. To fill this gap, we propose a novel Dilated Convolution Locate and Action Relevant Score model called DCAR. Specifically, DCAR includes a Dilated Location Network (DL-Net) and Action Relevance Calculation (ARC) block. For the DL-Net, we design a Boundary Feature Enhancement (BFE) block, which enhances the boundary feature of actions and fuses the similar features of different channels by pooling and channel squeeze to reduce the interference of the background. Meanwhile, we also design multiple dilated convolutional structures to aggregate long contextual information in time point/interval after boundary enhancement. For the ARC block, we use the hyperbolic space distance and cross attention to calculate the causal correlation of action proposals classification, which removes the misclassification of action proposals. We conduct extensive experiments on Thumos14 and ActivityNet-1.3, which shows our method significantly improves the performance and achieves state-of-the-art results. On Thumos14, at tIoU 0.3, 0.4 and 0.5, it achieves mAP of 63.2%, 57.0% and 48.5%, respectively; On ActivityNet-1.3 it reaches 9.42% at tIoU 0.95.

Rare Middle Triassic coleoids from the Alpine-Carpathian system: new records from Slovakia and their significance

Two stratigraphically well constrained (by ammonites and conodonts) coleoid remains have been recorded from the Triassic (Anisian) dark-grey organodetritic limestones (Ráztoka Limestone) of Western Carpathians (Hronic Nappe). The limestones deposited at the periphery of a former carbonate platform. It yields a highly diverse cephalopod fauna including nautiloids (2 taxa), ammonoids (7 taxa) and indetermined aulacoceratids. Two unusual coleoid specimens are referred to genus Mojsisovicsteuthis (M. boeckhi) and probably to a new taxon (described as Breviconoteuthis aff. breviconus herein) possessing similar morphological features of genus Breviconoteuthis (Phragmoteuthida) and/or Zugmontites. Based on index ammonites and conodonts, both records are of the uppermost Trinodosus through the lowermost Reitzi zones (Anisian—lower Illyrian). While the genus Mojsisovicsteuthis has been widely dispersed (however its records are rare), the occurrence of Breviconoteuthis and Zugmontites is strictly limited to the Alpine-Carpathian region. Comparing with the holotype and additional specimens stored in the Hungarian Natural History Museum, the overal shell of Mojsisovicsteuthis and its size has been reconstructed. Its relationship to aulacoceratids and phragmoteuthids is briefly discussed. Geochemical record (n-alkanes from the bulk rock) provided a relevant signal of the existence of algal meadows.

Paleo-geomorphic features of pre-Jurassic and its oil-controlling effect in Wuqi–Dingbian area

The Fuxian-Yan10 layers are the main oil-producing reservoirs of Jurassic in Wuqi–Dingbian area of Ordos Basin. However, due to the lack of understanding of the pattern and distribution characteristics of oil reservoirs, the benefits of exploration and development are restricted. In order to provide theoretical guidance for the study of similar geological features, based on the analysis of paleogeomorphic features and evolution, the analysis focuses on the influence of palaeo-geomorphology on oil reservoir distribution, and summarizes the main types of reservoir models in the study area. The results show that there are four types of palaeo-geomorphic units in the Wuding area: palaeo-river, slope, highland and interriver hill. In the study area, the Jurassic paleogeomorphology controls the sedimentary development and distribution from Fuxian Formation to Yan 9 Formation. The compacted structure and lithologic barrier provide good trapping conditions for the paleogeomorphic oil. Moreover, the swampy coal measures and mudstone at the top of Yan 9 play a sealing role for oil accumulation, and the bottom water was obviously driven. In addition, the pre-Jurassic deep valley was the main channels for oil migration. On this basis, it is concluded that there are four reservoir-forming models in Wuding area: slope type, river hill type, ancient river type and highland type.

Real-Time Detection Technology of Corn Kernel Breakage and Mildew Based on Improved YOLOv5s

In order to solve low recognition of corn kernel breakage degree and corn kernel mildew degree during corn kernel harvesting, this paper proposes a real-time detection method for corn kernel breakage and mildew based on improved YOlOv5s, which is referred to as the CST-YOLOv5s model algorithm in this paper. The method continuously obtains images through the discrete uniform sampling device of corn kernels and generates whole corn kernels, breakage corn kernels, and mildew corn kernel dataset samples. We aimed at the problems of high similarity of some corn kernel features in the acquired images and the low precision of corn kernel breakage and mildew recognition. Firstly, the CBAM attention mechanism is added to the backbone network of YOLOv5s to finely allocate and process the feature information, highlighting the features of corn breakage and mildew. Secondly, the pyramid pooling structure SPPCPSC, which integrates cross-stage local networks, is adopted to replace the SPPF in YOLOv5s. SPP and CPSC technologies are used to extract and fuse features of different scales, improving the precision of object detection. Finally, the original prediction head is converted into a transformer prediction head to explore the prediction potential with a multi-head attention mechanism. The experimental results show that the CST-YOLOv5s model has a significant improvement in the detection of corn kernel breakage and mildew. Compared with the original YOLOv5s model, the average precision (AP) of corn kernel breakage and mildew recognition increased by 5.2% and 7.1%, respectively, and the mean average precision (mAP) of all kinds of corn kernel recognition is 96.1%, and the frame rate is 36.7 FPS. Compared with YOLOv4-tiny, YOLOv6n, YOLOv7, YOLOv8s, and YOLOv9-E detection model algorithms, the CST-YOLOv5s model has better overall performance in terms of detection accuracy and speed. This study can provide a reference for real-time detection of breakage and mildew kernels during the harvesting process of corn kernels.

Exploring Dissent and Critique in Anti-war Literature: A Comparative Examination of Narrative Styles in Dalton Trumbo's Johnny Got His Gun and Kurt Vonnegut's Slaughterhouse-five

Narrative techniques are poweful tools in conveying dissent and critique in Anti War literature. This paper aims to study and analyse Dalton Trumbo’s Johnny Got His Gun and Kurt Vonnegut’s Slaughterhouse Five with the objective of exploring the narrative techniques employed by each author to express dissent and critique against war. The paper also aims to trace the similar and distinct features of narrative styles in the two works to highlight how two different narratives which are the results two World Wars strive to achieve the same goal, that is, to portray the incomprehensible and utterly unjustified human cost of wars. sheds light on the impact war and the criticisms of war have on literature.

Reversible encephalitis-like episodes in fragile X-associated tremor/ataxia syndrome: a case report

BackgroundFragile X-associated tremor/ataxia syndrome (FXTAS) is a neurodegenerative disorder caused by CGG repeat expansion of FMR1 gene. Both FXTAS and neuronal intranuclear inclusion disease (NIID) belong to polyglycine diseases and present similar clinical, radiological, and pathological features, making it difficult to distinguish these diseases. Reversible encephalitis-like attacks are often observed in NIID. It is unclear whether they are presented in FXTAS and can be used for differential diagnosis of NIID and FXTAS.Case presentationA 63-year-old Chinese male with late-onset gait disturbance, cognitive decline, and reversible attacks of fever, consciousness impairment, dizziness, vomiting, and urinary incontinence underwent neurological assessment and examinations, including laboratory tests, electroencephalogram test, imaging, skin biopsy, and genetic test. Brain MRI showed T2 hyperintensities in middle cerebellar peduncle and cerebrum, in addition to cerebellar atrophy and DWI hyperintensities along the corticomedullary junction. Lesions in the brainstem were observed. Skin biopsy showed p62-positive intranuclear inclusions. The possibilities of hypoglycemia, lactic acidosis, epileptic seizures, and cerebrovascular attacks were excluded. Genetic analysis revealed CGG repeat expansion in FMR1 gene, and the number of repeats was 111. The patient was finally diagnosed as FXTAS. He received supportive treatment as well as symptomatic treatment during hospitalization. His encephalitic symptoms were completely relieved within one week.ConclusionsThis is a detailed report of a case of FXTAS with reversible encephalitis-like episodes. This report provides new information for the possible and rare features of FXTAS, highlighting that encephalitis-like episodes are common in polyglycine diseases and unable to be used for differential diagnosis.

Li-ion storage chemistry of metal layers influences on metal oxides

A suitable current collector is a critical criterion for attaining high areal capacity in lithium batteries (LIBs). Here, we discover that the transition metal foam-type current collectors and their corresponding active materials upon a simple oxidation process, enable us to design a strategy for controlling the thickness of active materials toward achieving high areal capacity LIBs. Taking nickel foam (NiF) as a case study, we demonstrated that by calcining it in the air at different temperatures of 500 ℃, 600 ℃, 700 ℃, and 800 ℃, the electrodes exhibit distinct physical properties such as the thickness of active material layers, mechanical strength, and electrochemical properties including the rate capabilities, pseudocapacitive contribution, and lithium-ion diffusion. Ultimately, the optimized anode calcined at 700 °C (700-NiO@NiF) maintains a high reversible areal capacity of 7.6 mAh cm−2@0.4 mA cm−2, and 2.0 mAh cm−2 after 200 cycles@3.0 mA cm−2, We further verify that both cobalt foam and copper foam exhibited similar features like that of NiF by tuning their annealing temperatures, creating more opportunities for maximizing the areal capacity of LIBs.

Abstract PO2-02-07: Clinicopathological Features and Prognostic Role of HER2 Low in Early Breast Cancer: A real world study

Abstract Background: The definition of human epidermal growth factor receptor-2(HER2) low was proposed with the development of T-DXd, a novel antibody-drug conjugate (ADCs) targeting HER2 in the field of breast cancer (BC) treatments. However, whether HER2 low is a distinct subgroup is still an ongoing debate with mixed results. Current retrospective studies focused on HER2 low concentrated more on metastatic BC, less was known in early BC. The retrospective study was conducted based on HER2 low early BC to comprehensively analyze its clinicopathological features and prognostic roles compared with HER2 IHC0. Methods: This single institution retrospective study enrolled 999 early stage (stage I through III) HER2 negative invasive BC patients diagnosed at Chinese PLA General Hospital from January 2010 to December 2015. Tumors with HER2 immunohistochemical (IHC) score of 1+ or 2+ with negative in situ hybridization assay were defined as HER2 low. Clinicopathological characteristics s and survival outcomes including disease-free survival (DFS) and overall survival (OS) were compared between HER2 IHC0 and HER2 low groups. The correlations between HER2 expression and estrogen receptor (ER) status were determined by Mantel-Haenszel χ2 test. Cohen’s kappa coefficient (k) was adopted to evaluate the concordance between HER2 expression on primary tumors and matched biopsies. Results: 999 patients were eligible for inclusion criteria, among which the percentages of HER2 low in the whole cohort, HR positive group and HR negative group were 83.6%, 86.54%, 72.14% respectively. It was observed the presence of a significant higher proportion of ER level > 10% (p < 0.001), PgR positive (p < 0.001) and more patients who received endocrine therapy (p = 0.002). In HR positive group, clinicopathological features were balanced between HER2 IHC0 and HER2 low. In HR negative group, HER2 low manifested a higher proportion of invasive ductal type (p = 0.016) compared with HER2 IHC0. Survival analysis showed a significantly longer OS in HER2 low BC than HER2 IHC0. (HR: 0.458, 95%CI: 0.262-0.800, p = 0.0005). Nevertheless, the independent role was not reached in DFS. When stratified by HR status, patients with HER2 low in HR positive group demonstrated a longer OS than HER2 IHC0 (HR: 0.55, 95%CI: 0.273-1.105, p = 0.0385). The same trend was observed in HR negative group (HR: 0.385, 95%CI: 0.158-0.939, p = 0.0132). Neither the HR positive group nor the HR negative group achieved significant DFS. Besides, a positive correlation was observed between ER status and the rate of HER2 low (Mantel-Haenszel c2 test, p < 0.001, Pearson’s R = 0.159, P < 0.001). No regularity of survival differences between HER2 IHC0 and HER2 low was found in each ER level in the exploratory study. Finally, 119 patients who developed relapsed or metastatic disease were identified, with the most prevalent metastatic site being bone (31.1%), following by lung (16.80%) and lymph node (13.44%). Among 52 patients with matched biopsy samples, dynamic changes of HER2 IHC status were confirmed in both primary and relapsed statuses, with an discordance rate of 28.84% (K = 0.194, 95%CI: 0.168-0.219). conclusion: HER2 low expression breast cancer may not be regarded as a unique BC group in this real-world population due to similar clinicopathological features and prognostic roles especially in hormone-receptor positive cases. ER level was positively correlated with the rate of HER2 low population. Overall, whether HER2 low is a distinct subgroup remains to be supported and validated by more data both in clinical and molecular levels. Key words: HER2 low; Breast cancer; Clinicopathological Features; Prognosis; Molecular type Clinicopathological Features and Prognostic Role of HER2 Low in Early Breast Cancer: A real world study Citation Format: Ziqing Kong, Weihong Zhao. Clinicopathological Features and Prognostic Role of HER2 Low in Early Breast Cancer: A real world study [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO2-02-07.

Region‐dependent meteorological conditions of thunderstorm based on clustering analysis of stability and precipitable water indices

AbstractThunderstorms contribute to a large number of destructive weather events. However, due to the availability of comprehensive thunderstorm datasets, current researches on thresholds of stability indices during thunderstorms have always been limited within a specific local area during a relatively short time span. This study uses observations of thunderstorms data and ERA5 reanalysis data in summer (June–August) at 0600 and 1200 UTC from 1995 to 2019 to investigate the region‐dependent thresholds of four indices for indicating the occurrence of thunderstorm. The indices include the K index (K), the Showalter index (SI), convective available potential energy (CAPE), and precipitable water (PW). By applying the K‐means clustering method on the composite mean fields of four indices on thunderstorm days, China is objectively divided into three regions based on three stability indices (K, SI, and CAPE) and two regions based on PW index. Region‐dependent thresholds have been established for each index related to thunderstorms. When these thresholds are applied, accuracy rate of detecting thunderstorms exceeds 60% at 85% of the stations over China. The remarkable regional difference in the thresholds of the four indices are further attributed to the climatic backgrounds. Summer climatological means of surface potential temperature and surface relative humidity play a pivotal role in shaping the distinctive thresholds of K, CAPE, and PW across diverse regions, while surface potential temperature and topography are pivotal for the region‐dependent thresholds of SI. Based on the terrain characteristics and relevant summer climatology in each region, the thunderstorm thresholds can be potentially applied to regions with limited thunderstorm records but exhibiting similar climatic features beyond China. These findings would provide a reliable reference for thunderstorm monitoring by meteorological departments and related decision‐making services.

Abstract PO2-25-06: A hybrid breast cancer/mesenchymal stem cell population increases distant metastasis and chemoresistance of metaplastic breast carcinomas

Abstract Background: Metaplastic breast carcinoma is a heterogenous, highly aggressive and chemoresistant subtype of triple negative breast cancer (TNBC) with histological evidence of deregulated differentiation towards non-glandular components including spindle, squamous, and sarcomatoid. The role of the tumor microenvironment in the metastatic behavior and chemotherapy response of metaplastic carcinomas needs investigation to improve patient outcomes. Our lab has generated a model of spindle metaplastic breast carcinomas based on mammary epithelial cell-specific CCN6 knockout (MMTV-Cre;Ccn6fl/fl or CCN6KO). We have recently shown that a subset of TNBC cells engulf mesenchymal stem/stromal cells (MSC) generating a hybrid cell population with metastatic ability, which is detected in breast cancer tissue samples. In this study, we tested the hypothesis that hybrid cells may promote chemoresistant distant metastasis in spindle metaplastic breast carcinomas and that the tumor immune microenvironment may play a role in this function. Methods: Luc-GFP-CCN6KO cells were cultured with DsRED-MSCs for 7 days, yielding >90% of hybrid cells detected by flow cytometry. Hybrids or CCN6KO single cultures (1x10(5) cells) were injected intracardially in FVB/NJ syngeneic immune competent mice (n=20 mice per group). Four days after injection, each group was treated with Paclitaxel (PTX, 10 mg/kg twice a week, i.p.) or PBS control. The metastatic burden was evaluated by bioluminescence imaging twice a week. Mice were euthanized 23 days after heart injections, followed by necropsy and histopathological evaluation of the harvested organs. Results: All mice injected with single cultures of CCN6KO cells or hybrid cells formed distant metastasis. CCN6KO cells were highly invasive and spread to the lungs and lymph nodes in the chest,. We noted that 5 of 20 (25%) mice injected with hybrid cells showed metastasis in the liver (n=3 mice), pancreas (n=1), and abdominal lymph nodes (n=1). PTX treatment significantly reduced metastatic burden in mice injected with CCN6KO single cultures compared to vehicle (p=0.035) but had no effect on the metastatic burden of hybrid cells. Histologically, while the metastasis from CCN6KO cells and hybrids showed similar pathological features, consisting of spindle cancer cells with high grade nuclei and >10 mitoses/10 high power fields, hybrid-derived tumors had abundant infiltration by immune cells, mainly neutrophils, compared to single culture tumors. Conclusion: Our data show that hybrid cancer cell/MSC derived from CCN6KO spindle metaplastic carcinomas have higher ability to disseminate to distant organs, especially in the abdomen, compared to cancer cells. Our data show that hybrid cell-derived metastases have increased immune cell infiltration. We provide evidence that hybrid cells enhance Paclitaxel resistance in a mouse model of metaplastic breast carcinoma. Ongoing studies are aimed at elucidating the role of hybrids in the immune microenvironment of metaplastic breast carcinoma and clinical significance. Citation Format: Ahmad Eido, Maria Gonzalez, Celina Kleer. A hybrid breast cancer/mesenchymal stem cell population increases distant metastasis and chemoresistance of metaplastic breast carcinomas [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO2-25-06.

Improving Graph Collaborative Filtering with Directional Behavior Enhanced Contrastive Learning

Graph Collaborative Filtering is a widely adopted approach for recommendation, which captures similar behavior features through graph neural network. Recently, Contrastive Learning (CL) has been demonstrated as an effective method to enhance the performance of graph collaborative filtering. Typically, CL-based methods first perturb users’ history behavior data (e.g., drop clicked items), then construct a self-discriminating task for behavior representations under different random perturbations. However, for widely existing inactive users, random perturbation makes their sparse behavior information more incomplete, thereby harming the behavior feature extraction. To tackle the above issue, we design a novel directional perturbation-based CL method to improve the graph collaborative filtering performance. The idea is to perturb node representations through directionally enhancing behavior features. To do so, we propose a simple yet effective feedback mechanism, which fuses the representations of nodes based on behavior similarity. Then, to avoid irrelevant behavior preferences introduced by the feedback mechanism, we construct a behavior self-contrast task before and after feedback, to align the node representations between the final output and the first layer of GNN. Different from the widely-adopted self-discriminating task, the behavior self-contrast task avoids complex message propagation on different perturbed graphs, which is more efficient than previous methods. Extensive experiments on three public datasets demonstrate that the proposed method has distinct advantages over other contrastive learning methods on recommendation accuracy.

SCFNet: Lightweight Steel Defect Detection Network Based on Spatial Channel Reorganization and Weighted Jump Fusion

The goal of steel defect detection is to enhance the recognition accuracy and accelerate the detection speed with fewer parameters. However, challenges arise in steel sample detection due to issues such as feature ambiguity, low contrast, and similarity among inter-class features. Moreover, limited computing capability makes it difficult for small and medium-sized enterprises to deploy and utilize networks effectively. Therefore, we propose a novel lightweight steel detection network (SCFNet), which is based on spatial channel reconstruction and deep feature fusion. The network adopts a lightweight and efficient feature extraction module (LEM) for multi-scale feature extraction, enhancing the capability to extract blurry features. Simultaneously, we adopt spatial and channel reconstruction convolution (ScConv) to reconstruct the spatial and channel features of the feature maps, enhancing the spatial localization and semantic representation of defects. Additionally, we adopt the Weighted Bidirectional Feature Pyramid Network (BiFPN) for defect feature fusion, thereby enhancing the capability of the model in detecting low-contrast defects. Finally, we discuss the impact of different data augmentation methods on the model accuracy. Extensive experiments are conducted on the NEU-DET dataset, resulting in a final model achieving an mAP of 81.2%. Remarkably, this model only required 2.01 M parameters and 5.9 GFLOPs of computation. Compared to state-of-the-art object detection algorithms, our approach achieves a higher detection accuracy while requiring fewer computational resources, effectively balancing the model size and detection accuracy.