In visual reasoning, the achievement of deep learning significantly improved the accuracy of results. Image features are primarily used as input to get answers. However, the image features are too ...redundant to learn accurate characterizations within a limited complexity and time. While in the process of human reasoning, abstract description of an image is usually to avoid irrelevant details. Inspired by this, a higher-level representation named semantic representation is introduced. In this paper, a detailed visual reasoning model is proposed. This new model contains an image understanding model based on semantic representation, feature extraction and process model refined with watershed and u-distance method, a feature vector learning model using pyramidal pooling and residual network, and a question understanding model combining problem embedding coding method and machine translation decoding method. The feature vector could better represent the whole image instead of overly focused on specific characteristics. The model using semantic representation as input verifies that more accurate results can be obtained by introducing a high-level semantic representation. The result also shows that it is feasible and effective to introduce high-level and abstract forms of knowledge representation into deep learning tasks. This study lays a theoretical and experimental foundation for introducing different levels of knowledge representation into deep learning in the future.
In the field of visual reasoning, image features are widely used as the input of neural networks to get answers. However, image features are too redundant to learn accurate characterizations for ...regular networks. While in human reasoning, abstract description is usually constructed to avoid irrelevant details. Inspired by this, a higher-level representation named semantic representation is introduced in this paper to make visual reasoning more efficient. The idea of the Gram matrix used in the neural style transfer research is transferred here to build a relation matrix which enables the related information between objects to be better represented. The model using semantic representation as input outperforms the same model using image features as input which verifies that more accurate results can be obtained through the introduction of high-level semantic representation in the field of visual reasoning.
•A hybrid model is proposed directly utilizing tool images to predict surface roughness on machined parts for tool condition assessment.•Pruning based on the UNet3+ architecture eliminates redundant ...structures and reduces model parameters while integrating attention mechanisms to enhance the model's focus on target regions.•A heterogeneous dataset centered on tool images has been developed to comprehensively predict the surface roughness of workpieces.•Achieved an accuracy exceeding 93.1 % in predicting surface roughness on machined parts.
Worn tools might lead to substantial detrimental implications on the surface integrity of workpieces for precision/ultra-precision machining. Most previous research has heavily relied on singular information, which might not be appropriate enough to ascertain tool conditions and guarantee the accuracy of workpieces. This paper proposes a CNN-LSTM hybrid model directly utilizing tool images to predict surface roughness on machined parts for tool condition assessment. This work first performs pruning based on UNet3+ architecture to eliminate redundant structures while integrating attention mechanisms to enhance the model's focus on the target region. On this basis, tool wear region information is intensely mined and heterogeneous data is optimized using Spearman correlation analysis. Subsequently, we innovatively proposed a hybrid model that integrates CNN and RNN, endowing the model with the ability to process spatial and sequential information. The effectiveness of the proposed methodology is validated using the practical data obtained from cutting experiments. The results indicate that the proposed tool condition assessment methodology significantly improves the segmentation accuracy of the tool wear region to 94.52 % (Dice coefficient) and predicts the surface roughness of machined parts with an accuracy exceeding 93.1 % (R2). It can be observed that the developed methodology may provide an effective solution for accurate tool condition assessment and the implementation of tool health management.
Background
Neuroblastoma is the most common malignant extracranial tumor for children. Molecular mechanisms underpinning the pathogenesis of this disease are yet to be fully clarified. This study ...aimed to identify a novel oncogene that could be used as a biomarker informing the prognosis of neuroblastoma, and to predict its biological functions, using bioinformatics and molecular biology tools.
Methods
Three data sets from the TARGET, GSE62564, and GSE85047 databases were used for analysis. Survivals of patients with high or low expression of
bub1
were compared, using the Kaplan-Meier curve and log-rank test. Immune infiltration was evaluated using ESTIMATE and MCP-counter algorithms. Synthetic small interfering RNAs (siRNAs) were employed to silence
bub1
expression in neuroblastoma cell lines SH-SY5Y and SK-N-SH, in order to characterize its biological functions. Gene enrichment analyses of
bub1
were carried out, using Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses.
Results
Expression of
bub1
was found to significantly affect overall survival and event-free survival of patients with neuroblastoma, positively correlate with the expressions of
tpx2
and the
ASPM
gene, and negatively correlate with host immune infiltration. Expression of
bub1
was elevated in patients with neuroblastoma. Silencing
bub1
expression using siRNAs in SH-SY5Y and SK-N-SH resulted in decreased cell growth (
p
< 0.05), reduced migration (
p
< 0.05), and increased apoptosis (
p
< 0.05). Function analysis of
bub1
revealed cancer-promoting effects, probably
via
regulating several important downstream molecules, including that related to the apoptosis process and epithelial-mesenchymal transition.
Conclusion
We identified a potential tumor-promoting gene
bub1
for neuroblastoma that could also serve as a prognostic biomarker.
Driven by a global trend of applying replace-reduce-refine or 3Rs’ guidance for experimental animals in life sciences, chick embryo and particularly allantois with its chorioallantoic membrane have ...been increasingly utilized to substitute laboratory animals, which call for more extensive and updated knowledge about this novel experimental setup. In this study, being noninvasive, nonionizing, and super-contrasting with high spatiotemporal resolutions, magnetic resonance imaging (MRI) was chosen as an imaging modality for in ovo monitoring morphologic evolution of the chick embryo, allantois, and chorioallantoic membrane longitudinally throughout embryonic day (ED) 1 until ED20. Cooled in 0°C ice bath for 60 min to reduce MRI motion artifacts, 3 chick embryos (n = 60 in total) on each ED were scanned by a clinical 3.0T MRI scanner to demonstrate 3D images of both T2- and T1-weighted imaging (T2WI, T1WI) sequences at axial, sagittal, and coronal slices. The volumes of both the entire chick embryo and allantois were semi-automatically segmented based on intensity-based thresholding and region-growing algorithms. The morphometries or quantified 3D structures were achieved by refined segmentation, and confirmed by histological analyses (one for each ED). After MRI, the rest of chick embryos (n = 40) continued for incubation. The images from ED2 to ED4 could demonstrate the structural changes of latebra, suggesting its transition into a nutrient supplying channel of yolk sac. The allantois could be recognized by MRI, and its relative volumes on each ED revealed an evolving profile peaked on ED12, with a statistically significant difference from those of earlier and later EDs (P < 0.01). The hypointensity of the yolk due to the susceptibility effect of its enriched iron content overshadowed the otherwise hyperintensity of its lipid components. The chick embryos survived prior cooling and MRI till hatching on ED21. The results could be further developed into a 3D MRI atlas of chick embryo. Clinical 3.0T MRI proved effective as a noninvasive approach to study in ovo 3D embryonic development across the full period (ED1–ED20), which can complement the present knowhow for poultry industry and biomedical science.
There are often subtle early symptoms of colorectal cancer, a common malignancy of the intestinal tract. However, it is not yet clear how MYC and NCAPG2 are involved in colorectal cancer.
We obtained ...colorectal cancer datasets GSE32323 and GSE113513 from the Gene Expression Omnibus (GEO). After downloading, we identified differentially expressed genes (DEGs) and performed Weighted Gene Co-expression Network Analysis (WGCNA). We then undertook functional enrichment assay, gene set enrichment assay (GSEA) and immune infiltration assay. Protein-protein interaction (PPI) network construction and analysis were undertaken. Survival analysis and Comparative Toxicogenomics Database (CTD) analysis were conducted. A gene expression heat map was generated. We used TargetScan to identify miRNAs that are regulators of DEGs.
1117 DEGs were identified. Their predominant enrichment in activities like the cellular phase of the cell cycle, in cell proliferation, in nuclear and cytoplasmic localisation and in binding to protein-containing complexes was revealed by Gene Ontology (GO). When the enrichment data from GSE32323 and GSE113513 colon cancer datasets were merged, the primary enriched DEGs were linked to the cell cycle, protein complex, cell cycle control, calcium signalling and P53 signalling pathways. In particular, MYC, MAD2L1, CENPF, UBE2C, NUF2 and NCAPG2 were identified as highly expressed in colorectal cancer samples. Comparative Toxicogenomics Database (CTD) demonstrated that the core genes were implicated in the following processes: colorectal neoplasia, tumour cell transformation, inflammation and necrosis.
High MYC and NCAPG2 expression has been observed in colorectal cancer, and increased MYC and NCAPG2 expression correlates with worse prognosis.
Developing electrolyte membranes with a simple preparation process and high performance is a top priority for the commercialization of fuel cells. Inspired by solar cell texturing to improve its ...conversion efficiency, this study prepares a textured membrane by increasing the roughness of a glass plate. The structures of the textured membrane and the flat membrane are characterized and compared. The membranes are assembled in fuel cells for performance testing. The surface area of the textured membrane is 1.27 times that of the flat membrane, which increases the size of the three-phase boundary in fuel cells. The maximum power density of the fuel cell using the textured membrane is 1.17 times of the cell using the flat membrane at 60 °C. The excellent performance of the cell using the textured membrane profit from the enlargement of the three-phase boundary. This work offers a simple way to develop outstanding-performance membranes by changing their surface roughness.
A textured membrane is simply prepared with a rough glass plate, and the improved cell performance benefits from the increased three-phase boundary.
Objective: The real-world study aimed to investigate the current status of Android artificial pancreas system (APS) use among patients with type 1 diabetes mellitus (T1DM) in China, where ...market-approved commercial APS is unavailable. Methods: The healthcare providers of China’s largest online T1DM community (TTQ) helped build up AndroidAPS using a compatible insulin pump, real-time continuous glucose monitoring (CGM), and a heuristic-based algorithm. Patients with T1DM who used AndroidAPS for at least 3 months between January 2020 and July 2023 were included. The baseline data and main glycemic metrics during the last 3 months after using AndroidAPS were collected. Results: We included 191 patients (56 non-adults and 135 adults) with a median age of 24.80 (15.83, 32.74) yrs, T1DM duration of 6.01(1.63,8.58) yrs, and body mass index (BMI) of 20.74(18.51,22.34) kg/m2. The duration of APS use was 153(114.00,206.00) days. After using AndroidAPS, HbA1c was decreased from 7.1(6.4,8.5) % to 6.3(5.6,6.8) % (p<0.001). The median 24-hour time in range (TIR,70-180mg/dL), time above range (TAR,>180mg/dL), time below range (TBR, <70mg/dL), TBR(<54mg/dL), and glucose management indicator were 81.40(71.99,87.10) %, 13.55(7.20,22.37) %, 4.06(2.63,6.28) %, 0.42(0.20,0.87) %, and 6.45(6.09,6.78) %, respectively. The median overnight TIR, TAR, TBR (<70mg/dL), and TBR(<54mg/dL) were 85.42(77.87,91.87) %, 8.00(3.45,16.52) %, 3.82(2.00,6.44) %, and 0.39(0.11,0.95) %, respectively. A total of 155 patients achieved the recommended TIR of 70%, and 66 achieved TIR, TAR, and TBR targets simultaneously. Multiple linear regression revealed age (β=0.285, p<0.01) and duration of APS use (β=-0.167, p<0.05) as determinants of 24-hour TIR after adjusting gender, duration, BMI, and baseline HbA1c (R2=0.13, p<0.01). Conclusion: Our study confirmed the efficacy and safety of AndroidAPS on glycemic control among Chinese patients with T1DM, especially at nighttime. Disclosure M. Lei: None. P. Ling: None. Y. Ni: None. X. Zeng: None. D. Yang: None. H. Deng: None. X. Yang: None. W. Xu: None. J. Yan: None. Funding the Science and Technology Planning Project of Guangzhou (202102010154) and the Diabetes Mellitus Research Fund Program from Shanghai Medical and Health Development Foundation (DMRFP_II_14 from SHMHDF)
With the miniaturization of crystals‐based photoelectric devices, electrode contact‐geometries may play a critical role in determining the device performance. However, investigation of the role of ...electrode contact geometries in situ faces great challenges due to the electrode contact geometry is typically unmodifiable. To this end, a kind of liquid metal is employed as an adaptive‐deformable electrode to study the carrier transport through perovskite microcrystals, in which the electrode contacts geometries/positions and thus the carrier‐pathways can be adjusted. Under light illumination, a spike feature of photocurrent is observed when carriers transport along the perovskite microcrystal surface upon an edge‐contact geometry, which is absent as the carrier mainly transport through crystal interior upon a top‐contact geometry. Switching, rectifying, and memristor functions are selectively realized just by modifying the contact geometry. The underlying mechanism for the observations is further elucidated. This study provides a platform for studying carrier transport through microscale crystals with adjustable contact geometry and supplies an approach for fabricating diverse functional devices by changing the electrode contact‐geometries.
By employing a deformable liquid metal electrode, it is demonstrated that the carrier‐pathway not only affects the carrier generation/recombination and trapping/detrapping process but also strongly influences the conductive filament formation in perovskite microcrystals, providing an approach for fabricating functional devices via the control of contact geometry.
PDF
