Search

Virtual try-on is a technology used to predict and visualize how clothing will look on a given body input. Traditional virtual try-on methods rely on expensive 3D body scanning devices and simulations to simulate how clothing appears on the human body. While these methods offer high-quality results, the cost of 3D scanning devices can be a barrier. In contrast, using 2D images provides a more convenient and cost-effective alternative. Users only need to input 2D images of the human body and clothing, and the try-on result can be visualized. This study builds upon previous 2D virtual try-on methods and extends them from image-image input to text-image input. This means that users can now input text descriptions of clothing instead of specific images. By utilizing text descriptions, the system can generate corresponding clothing that matches the provided text, expanding the range of use cases for virtual try-on. To generate accurate try-on results based on the text-image input, this study introduces a comprehensive framework comprising six modules: the text-image encoder, pose extractor, image segmentation, GAN-encoder, GAN-generator, and a mapping module. The overall framework follows a GAN-inversion editing pipeline. First, the GAN encoder encodes the input body image, producing a latent vector that captures the essential characteristics of the image（such as the image style and the body shape）. Second, the obtained latent vector is edited and then the GAN-generator is fed to the edited vector to generate the desired result. Specifically, the obtained latent vector is edited by using the mapping module, which shares the same network structure as the GAN generator. The mapping module generates an additional offset latent vector of the same dimension as the one obtained from the GAN encoder. This offset vector is used to edit the latent vector, ensuring that the generated image fulfills the desired pose and text description requirements. The offset vector also helps constrain the latent vector within the GAN latent space, facilitating the generation of high-quality images by using the GAN generator. To maintain consistency in poses and appearances, the pose extractor and image segmentation modules are utilized to construct loss functions. These loss functions guide the optimization process of the latent vectors, enabling the generator to produce a final generated image that remains consistent with the input human body. To generate accurate clothing images based on the input text descriptions, the pre-training text-image model CLIP is employed. CLIP encodes both the text descriptions and the final output image and constructs a loss function that regulates the optimization process during training. In experimental evaluations, the proposed method successfully generates correct images corresponding to the input body image and text descriptions. Compared to existing methods, quantitatively, the proposed method outperforms existing methods, achieving improvements of 15% in IoU, 8% in semantics, and 77. 1% in image quality evaluation. Compared to traditional physical fitting methods, virtual try-on provides consumers with an economical and convenient way to try on clothes. With the rapid advancements in machine learning and computer vision, virtual try-on has achieved impressive results. Furthermore, as consumers increasingly seek personalized experiences, the proposed virtual try-on method can generate the desired clothing based on text descriptions and present the final fitting results. This further enhances the flexibility and application scope of virtual try-on to meet the diverse needs of consumers. [ABSTRACT FROM AUTHOR]

The automatic generation of stylized co-speech gestures has recently received increasing attention. Previous systems typically allow style control via predefined text labels or example motion clips, which are often not flexible enough to convey user intent accurately. In this work, we present GestureDiffuCLIP, a neural network framework for synthesizing realistic, stylized co-speech gestures with flexible style control. We leverage the power of the large-scale Contrastive-Language-Image-Pre-training (CLIP) model and present a novel CLIP-guided mechanism that extracts efficient style representations from multiple input modalities, such as a piece of text, an example motion clip, or a video. Our system learns a latent diffusion model to generate high-quality gestures and infuses the CLIP representations of style into the generator via an adaptive instance normalization (AdaIN) layer. We further devise a gesture-transcript alignment mechanism that ensures a semantically correct gesture generation based on contrastive learning. Our system can also be extended to allow fine-grained style control of individual body parts. We demonstrate an extensive set of examples showing the flexibility and generalizability of our model to a variety of style descriptions. In a user study, we show that our system outperforms the state-of-the-art approaches regarding human likeness, appropriateness, and style correctness. [ABSTRACT FROM AUTHOR]

Research on video anomaly detection has mainly been based on video data. However, many real-world cases involve users who can conceive potential normal and abnormal situations within the anomaly detection domain. This domain knowledge can be conveniently expressed as text descriptions, such as "walking" or "people fighting", which can be easily obtained, customized for specific applications, and applied to unseen abnormal videos not included in the training dataset. We explore the potential of using these text descriptions with unlabeled video datasets. We use large language models to obtain text descriptions and leverage them to detect abnormal frames by calculating the cosine similarity between the input frame and text descriptions using the CLIP visual language model. To enhance the performance, we refined the CLIP-derived cosine similarity using an unlabeled dataset and the proposed text-conditional similarity, which is a similarity measure between two vectors based on additional learnable parameters and a triplet loss. The proposed method has a simple training and inference process that avoids the computationally intensive analyses of optical flow or multiple frames. The experimental results demonstrate that the proposed method outperforms unsupervised methods by showing 8% and 13% better AUC scores for the ShanghaiTech and UCFcrime datasets, respectively. Although the proposed method shows −6% and −5% than weakly supervised methods for those datasets, in abnormal videos, the proposed method shows 17% and 5% better AUC scores, which means that the proposed method shows comparable results with weakly supervised methods that require resource-intensive dataset labeling. These outcomes validate the potential of using text descriptions in unsupervised video anomaly detection. [ABSTRACT FROM AUTHOR]

Although randomized controlled trials have compared surgery versus endovascular treatment for intracranial aneurysms, the literature is sparse in terms of subgroup analysis for anterior communicating artery (ACoA) aneurysm management. This systematic review and meta-analysis sought to compare surgical versus endovascular treatment for ACoA aneurysms. Medline, PubMed, and Embase were searched from inception to December 12, 2022. Primary outcomes were post-treatment modified Rankin Scale (mRS) >2 and mortality. Secondary outcomes were aneurysm obliteration, retreatment and recurrence, rebleeding, technical failure, vessel rupture, aneurysmal subarachnoid hemorrhage-related hydrocephalus, symptomatic vasospasm, and stroke. Eighteen studies yielded 2368 patients, from which 1196 (50.5%) and 1172 (49.4%) patients underwent surgery and endovascular treatment, respectively. The odds ratio (OR) of mortality was similar in total (OR = 0.92 [0.63–1.37], P = 0.69), ruptured (OR = 0.92 [0.62–1.36], P = 0.66), and unruptured cohorts (OR = 1.58 [0.06–39.60], P = 0.78). The OR of mRS > 2 was similar in total (OR = 0.75 [0.50–1.13], P = 0.17), ruptured (OR = 0.77 [0.49–1.20], P = 0.25), and unruptured cohorts (OR = 0.64 [0.21–1.96], P = 0.44). The OR of obliteration was higher with surgery in the total (OR = 2.52 [1.49–4.27], P = 0.0008) and ruptured cohorts (OR = 2.61 [1.33–5.10], P = 0.005) and unruptured group (OR = 3.46 [1.30–9.20], P = 0.01). The OR of retreatment was lower with surgery in the total (OR = 0.37 [0.17–0.76], P = 0.007) and ruptured cohorts (OR = 0.31 [0.11–0.89], P = 0.03), thought it was similar in the unruptured group (OR = 0.51 [0.08–3.03], P = 0.46). The OR of recurrence was lower with surgery in the total (OR = 0.22 [0.10, 0.47], P = 0.0001), ruptured (OR = 0.16 [0.03, 0.90], P = 0.04), and mixed (un) ruptured cohorts (OR = 0.22 [0.09–0.53], P = 0.0009). The OR of rebleeding in ruptured group was similar (OR = 0.66 [0.29–1.52], P = 0.33). The ORs of other outcomes were similar. ACoA aneurysms may be safely treated with either surgery or endovascular treatment, although microsurgical clipping demonstrates higher obliteration rates and lower rates of retreatment and recurrence. [ABSTRACT FROM AUTHOR]

Assessing severe scoliosis requires the analysis of posturographic X-ray images. One way to analyse these images may involve the use of open-source artificial intelligence models (OSAIMs), such as the contrastive language–image pretraining (CLIP) system, which was designed to combine images with text. This study aims to determine whether the CLIP model can recognise visible severe scoliosis in posturographic X-ray images. This study used 23 posturographic images of patients diagnosed with severe scoliosis that were evaluated by two independent neurosurgery specialists. Subsequently, the X-ray images were input into the CLIP system, where they were subjected to a series of questions with varying levels of difficulty and comprehension. The predictions obtained using the CLIP models in the form of probabilities ranging from 0 to 1 were compared with the actual data. To evaluate the quality of image recognition, true positives, false negatives, and sensitivity were determined. The results of this study show that the CLIP system can perform a basic assessment of X-ray images showing visible severe scoliosis with a high level of sensitivity. It can be assumed that, in the future, OSAIMs dedicated to image analysis may become commonly used to assess X-ray images, including those of scoliosis. [ABSTRACT FROM AUTHOR]

Embeddings, i.e., vector representations of objects, such as texts, images, or graphs, play a key role in deep learning methodologies nowadays. Prior research has shown the importance of analyzing the isotropy of textual embeddings for transformer-based text encoders, such as the BERT model. Anisotropic word embeddings do not use the entire space, instead concentrating on a narrow cone in such a pretrained vector space, negatively affecting the performance of applications, such as textual semantic similarity. Transforming a vector space to optimize isotropy has been shown to be beneficial for improving performance in text processing tasks. This paper is the first comprehensive investigation of the distribution of multimodal embeddings using the example of OpenAI's CLIP pretrained model. We aimed to deepen the understanding of the embedding space of multimodal embeddings, which has previously been unexplored in this respect, and study the impact on various end tasks. Our initial efforts were focused on measuring the alignment of image and text embedding distributions, with an emphasis on their isotropic properties. In addition, we evaluated several gradient-free approaches to enhance these properties, establishing their efficiency in improving the isotropy/alignment of the embeddings and, in certain cases, the zero-shot classification accuracy. Significantly, our analysis revealed that both CLIP and BERT models yielded embeddings situated within a cone immediately after initialization and preceding training. However, they were mostly isotropic in the local sense. We further extended our investigation to the structure of multilingual CLIP text embeddings, confirming that the observed characteristics were language-independent. By computing the few-shot classification accuracy and point-cloud metrics, we provide evidence of a strong correlation among multilingual embeddings. Embeddings transformation using the methods described in this article makes it easier to visualize embeddings. At the same time, multiple experiments that we conducted showed that, in regard to the transformed embeddings, the downstream tasks performance does not drop substantially (and sometimes is even improved). This means that one could obtain an easily visualizable embedding space, without substantially losing the quality of downstream tasks. [ABSTRACT FROM AUTHOR]

Over the last twenty years, both in the interdisciplinary field of game studies as well as in the video game industry more broadly, much attention has been paid to the construction of virtual spaces, not only to set the limits and possibilities of game dynamics but also to contextualise the narrative. Therefore, various researchers and developers have focused on environmental storytelling: a design process with which conspicuous fragments of a narrative are present in virtual environments in the form of objects, characters, or entire areas that should stimulate the user to find them by moving the digital prosthesis. This article focuses on Immortality (Half Mermaid 2022), a videogame that recontextualise environmental storytelling since it does not link space exploration to kinetic movement. However, it bases this design principle on a space-time manipulation process that hybridises the work of an archivist with that of a movie editor inside a hypertext structure formed by hundreds of clips discoverable in a different order. [ABSTRACT FROM AUTHOR]

RNA helicases catalyze the ATP-dependent destabilization of RNA duplexes. DEAD-box helicases share a helicase core that mediates ATP binding and hydrolysis, RNA binding and unwinding. Most members of this family contain domains flanking the core that can confer RNA substrate specificity and guide the helicase to a specific RNA. However, the in vivo RNA substrates of most helicases are currently not defined. The DEAD-box helicase Hera from Thermus thermophilus contains a helicase core, followed by a dimerization domain and an RNA binding domain that folds into an RNA recognition motif (RRM). The RRM mediates high affinity binding to an RNA hairpin, and an adjacent duplex is then unwound by the helicase core. Hera is a cold-shock protein, and has been suggested to act as an RNA chaperone under cold-shock conditions. Using crosslinking immunoprecipitation of Hera/RNA complexes and sequencing, we show that Hera binds to a large fraction of T. thermophilus RNAs under normal-growth and cold-shock conditions without a strong sequence preference, in agreement with a structure-specific recognition of RNAs and a general function in RNA metabolism. Under cold-shock conditions, Hera is recruited to RNAs with high propensities to form stable secondary structures. We show that selected RNAs identified, including a set of tRNAs, bind to Hera in vitro, and activate the Hera helicase core. Gene ontology analysis reveals an enrichment of genes related to translation, including mRNAs of ribosomal proteins, tRNAs, tRNA ligases, and tRNA-modifying enzymes, consistent with a key role of Hera in ribosome and tRNA metabolism.

Cross-modal representation learning aims to learn a shared representation space where data from multiple modalities can be effectively compared, fused, and understood. This paper investigates the role of increased diversity in the similarity score matrix in enhancing the performance of the CLIP (Contrastive Language-Image Pretraining), a multi-modal learning model that establishes a connection between images and text within a joint embedding space. Two transforming approaches, sine and sigmoid (including two versions), are incorporated into the CLIP model to amplify larger values and diminish smaller values within the similarity matrix (logits). Hardware limitations are addressed using a more compact text encoder (DistilBERT) and a pre-trained ResNet50 image encoder. The proposed adaptations are evaluated on various benchmarks, including image classification and image/text retrieval tasks, using 10 benchmark datasets such as Food101, Flickr30k, and COCO. The performance of the adapted models is compared to the base CLIP model using Accuracy, mean per class, and Recall@k metrics. The results demonstrate improvements in Accuracy (up to 5.32% enhancement for the PatchCamelyon dataset), mean per class (up to 14.48% enhancement for the FGVCAircraft dataset), and retrieval precision (with an increase of up to 45.20% in Recall@1 for the COCO dataset), compared to the baseline algorithm (CLIP).

This paper presents CLIP Driven Few-shot Panoptic Segmentation (CLIP-FPS), a novel few-shot panoptic segmentation model that leverages the knowledge of Contrastive Language-Image Pre-training (CLIP) model. The proposed method builds upon a center indexing attention mechanism to facilitate knowledge transfer, which entails representing objects in an image as centers along with their pixel offsets. The model comprises a decoder responsible for generating object center-offset groups and a self-attention module tasked with producing a feature attention map. Subsequently, the object centers index the map to acquire the corresponding embeddings, paving the way for matrix multiplication and SoftMax operation to facilitate text embedding matching and the computation of the final panoptic segmentation masks. Quantitative evaluation on datasets such as COCO and Cityscapes shows that our method outperforms existing panoptic segmentation techniques in terms of Panoptic Quality (PQ) metrics.

The overall purpose of this study is to propose a novel fine-tuning method for the CLIP architecture that enables the retention of pre-existing knowledge from large datasets and the creation of a domain-agnostic image encoder for universal image embedding, addressing the challenge of transferring knowledge from source to target tasks using deep learning models. The basic design of the study involves applying the proposed method directly (without fine-tuning) to a wide range of instance retrieval and recognition tasks to evaluate its effectiveness. The study’s major findings indicate that the proposed method significantly enhances performance on unseen domains without requiring separate fine-tuning for each domain. The authors’ success in the Google Universal Image Embedding competition, where they were awarded a Gold medal out of 1200 teams, inspired their proposed method. These results have significant implications for real-life applications where multiple domains are common. In conclusion, the study offers a practical solution for transfer learning that addresses the challenges of dealing with multiple domains and advances deep learning, potentially inspiring further research in this area and driving progress in the field.

It is commonly understood that RNA-binding proteins crucially determine the fate of their target RNAs. Vice versa, RNAs are becoming increasingly recognized for their functions in protein regulation and the dynamics of RNA-protein complexes. Long non-coding RNAs are emerging as potent regulators of proteins that exert unknown RNA-binding properties and moonlighting functions. A vast array of RNA- and protein-centric techniques have been developed for the identification of protein and RNA targets, respectively, including unbiased protein mass spectrometry and next-generation RNA sequencing as readout. Determining true physiological RNA and protein targets is challenging as RNA–protein interaction is highly dynamic, tissue- and cell-type-specific, and changes with the environment. Here I review current techniques for the analysis of RNA–protein interactions in living cells and in vitro. RNA-centric techniques are presented on the basis of cross-linking or the use of alternative approaches. Protein-centric approaches are discussed in combination with high-throughput sequencing. Finally, the impact of mutations in RNA–protein complexes on human disease is highlighted.

This manuscript presents a new benchmark for assessing the quality of visual summaries without the need for human annotators. It is based on the Signature Transform, specifically focusing on the RMSE and the MAE Signature and Log-Signature metrics, and builds upon the assumption that uniform random sampling can offer accurate summarization capabilities. We provide a new dataset comprising videos from Youtube and their corresponding automatic audio transcriptions. Firstly, we introduce a preliminary baseline for automatic video summarization, which has at its core a Vision Transformer, an image–text model pre-trained with Contrastive Language–Image Pre-training (CLIP), as well as a module of object detection. Following that, we propose an accurate technique grounded in the harmonic components captured by the Signature Transform, which delivers compelling accuracy. The analytical measures are extensively evaluated, and we conclude that they strongly correlate with the notion of a good summary. [ABSTRACT FROM AUTHOR]

Alternative polyadenylation (APA) generates mRNA isoforms and diversifies gene expression. Here we report the discovery that the mTORC1 signaling pathway balances the expression of two Trim9/TRIM9 isoforms through APA regulation in human and mouse. We showed that CFIm components, CPSF6 and NUDT21, promote the short Trim9/TRIM9 isoform (Trim9-S/TRIM9-S) expression. In addition, we identified an evolutionarily conserved twin UGUA motif, UGUAYUGUA, in TRIM9-S polyadenylation site (PAS) that is critical for its regulation by CPSF6. We found additional CPSF6-regulated PASs with similar twin UGUA motifs in human and experimentally validated the twin UGUA motif functionality in BMPR1B, MOB4, and BRD4-L. Importantly, we showed that inserting a twin UGUA motif into a heterologous PAS was sufficient to confer regulation by CPSF6 and mTORC1. Our study reveals an evolutionarily conserved mechanism to regulate gene isoform expression by mTORC1 and implicates possible gene isoform imbalance in cancer and neurological disorders with mTORC1 pathway dysregulation.

Crosslinking and immunoprecipitation (CLIP) technologies have become a central component of the molecular biologists’ toolkit to study protein-RNA interactions and thus to uncover core principles of RNA biology. There has been a proliferation of CLIP-based experimental protocols, as well as computational tools, especially for peak-calling. Consequently, there is an urgent need for a well-documented bioinformatic pipeline that enshrines the principles of robustness, reproducibility, scalability, portability and flexibility while embracing the diversity of experimental and computational CLIP tools. To address this, we present nf-core/clipseq - a robust Nextflow pipeline for quality control and analysis of CLIP sequencing data. It is part of the international nf-core community effort to develop and curate a best-practice, gold-standard set of pipelines for data analysis. The standards enabled by Nextflow and nf-core, including workflow management, version control, continuous integration and containerisation ensure that these key needs are met. Furthermore, multiple tools are implemented (e.g. for peak-calling), alongside visualisation of quality control metrics to empower the user to make their own informed decisions based on their data. nf-core/clipseq remains under active development, with plans to incorporate newly released tools to ensure that pipeline remains up-to-date and relevant for the community. Engagement with users and developers is encouraged through the nf-core GitHub repository and Slack channel to promote collaboration. It is available at https://nf-co.re/clipseq.

Background: RNA-binding proteins (RBPs) play a critical role in regulating gene expression by binding to specific sites on RNA molecules. Identifying these binding sites is crucial for understanding the many functions of RBPs in cellular function, development and disease. Current experimental methods for identifying RBP binding sites, such as ultra-violet (UV) crosslinking and immunoprecipitation (CLIP), and especially the enhanced CLIP (eCLIP) protocol, were developed to identify authentic RBP binding sites experimentally. Methods: To make this data more accessible to the scientific community, we have developed RBP-Tar ( https://ncbr.muni.cz/RBP-Tar), a web server and database that utilises eCLIP data for 167 RBPs mapped on the human genome. The web server allows researchers to easily search and retrieve binding site information by genomic location and RBP name. Use case: Researchers can produce lists of all known RBP binding sites on a gene of interest, or produce lists of binding sites for one RBP on different genomic loci. Conclusions: Our future goal is to continue to populate the web server with additional experimental datasets from CLIP experiments as they become available and processed, making it an increasingly valuable resource for the scientific community.

Abstract Background Crosslinking and immunoprecipitation (CLIP) is a method used to identify in vivo RNA–protein binding sites on a transcriptome-wide scale. With the increasing amounts of available data for RNA-binding proteins (RBPs), it is important to understand to what degree the enriched motifs specify the RNA-binding profiles of RBPs in cells. Results We develop positionally enriched k-mer analysis (PEKA), a computational tool for efficient analysis of enriched motifs from individual CLIP datasets, which minimizes the impact of technical and regional genomic biases by internal data normalization. We cross-validate PEKA with mCross and show that the use of input control for background correction is not required to yield high specificity of enriched motifs. We identify motif classes with common enrichment patterns across eCLIP datasets and across RNA regions, while also observing variations in the specificity and the extent of motif enrichment across eCLIP datasets, between variant CLIP protocols, and between CLIP and in vitro binding data. Thereby, we gain insights into the contributions of technical and regional genomic biases to the enriched motifs, and find how motif enrichment features relate to the domain composition and low-complexity regions of the studied proteins. Conclusions Our study provides insights into the overall contributions of regional binding preferences, protein domains, and low-complexity regions to the specificity of protein-RNA interactions, and shows the value of cross-motif and cross-RBP comparison for data interpretation. Our results are presented for exploratory analysis via an online platform in an RBP-centric and motif-centric manner ( https://imaps.goodwright.com/apps/peka/ ).

[Display omitted] [ABSTRACT FROM AUTHOR]

Intracranial aneurysms (IAs) are dilatations of the cerebral arteries, whose treatment is commonly based on the implant of a metallic clip on the aneurysm neck. Despite the dissection and understanding of the surgical anatomy of the IA when often only parts of it are visible, the choice of the ideal clip to be used is one of the surgical difficulties. Although current imaging tests guarantee IA visualization, currently there is no planning method that allows for a real three-dimensional (3D) visualization for optimal choice of clip prior to surgery. The aim of this study is to evaluate whether IA biomodels generated by additive manufacturing methods are useful for surgical clip selection in microsurgeries for IA. Three-dimensional (3D) IA biomodels of 10 patients with IA were evaluated using computerized tomography, surgical microscope, and 3D printer. The research was divided into 4 phases as follows: development of the 3D biomodels, evaluation of the biomodel dimensional characteristics, surgical planning evaluation with the biomodel and its clipping effectiveness, and evaluation of the actual surgical simulation process within the models. Ten 3D biomodels were obtained, made of a malleable and hollow part, formed by the IA and related arteries, and another rigid part, mimicking the skull and other arteries of the skull base. Based on these 3D models, 10 clips were chosen during the surgical planning, and all exactly matched the clip characteristics used during the actual surgeries. The surgical simulation with the biomodels performed by 2 neurosurgeons still in training obtained 100% accuracy in the identification of the clips that were eventually used during the actual surgeries. 3D biomodels generated by additive manufacturing methods were effective for surgical clip selection in microsurgeries for IA, reducing surgical time, increasing cerebral angioarchitecture understanding, and providing more safety in this type of surgery. [ABSTRACT FROM AUTHOR]

It is commonly understood that RNA-binding proteins crucially determine the fate of their target RNAs. Vice versa, RNAs are becoming increasingly recognized for their functions in protein regulation and the dynamics of RNA-protein complexes. Long non-coding RNAs are emerging as potent regulators of proteins that exert unknown RNA-binding properties and moonlighting functions. A vast array of RNA- and protein-centric techniques have been developed for the identification of protein and RNA targets, respectively, including unbiased protein mass spectrometry and next-generation RNA sequencing as readout. Determining true physiological RNA and protein targets is challenging as RNA–protein interaction is highly dynamic, tissue- and cell-type-specific, and changes with the environment. Here I review current techniques for the analysis of RNA–protein interactions in living cells and in vitro. RNA-centric techniques are presented on the basis of cross-linking or the use of alternative approaches. Protein-centric approaches are discussed in combination with high-throughput sequencing. Finally, the impact of mutations in RNA–protein complexes on human disease is highlighted. [ABSTRACT FROM AUTHOR]

Purpose: An anterior communicating artery is a common location for both ruptured and unruptured intracranial aneurysms, and microsurgery is sometimes necessary for their successful treatment. However, postoperative infarction should be considered during clipping due to the complex surrounding structures of anterior communicating artery aneurysms. This study aimed to evaluate the risk factors of postoperative infarction after surgical clipping of unruptured anterior communicating artery aneurysms and its clinical outcomes. Methods: The data of patients who underwent microsurgical clipping of an unruptured anterior communicating artery aneurysm in our hospital between January 2008 and December 2020 were retrospectively analyzed. The patients' demographic data, anatomical features of the anterior communicating artery complex and aneurysm, surgical technique, characteristics of postoperative infarction, and its clinical course were evaluated. Results: Notably, among 848 patients, 66 (7.8%) and 34 (4%) patients had radiologic and symptomatic infarctions, respectively. Univariate and multivariate logistic regression analyses showed that hypertension (odds ratio (OR), 1.99; p = 0.022 ), previous stroke (OR, 3.89; p = 0.009 ), posterior projection (OR, 5.58; p < 0.001 ), aneurysm size (OR, 1.17; optimal cut-off value, 6.14 mm; p = 0.002 ), and skull base-to-aneurysm distance (OR, 1.15; optimal cut-off value, 11.09 mm; p < 0.001 ) were associated with postoperative infarction. In the pterional approach, a closed A2 plane was an additional risk factor (OR, 1.88; p = 0.041 ). Infarction of the subcallosal and hypothalamic branches was significantly associated with symptomatic infarction ( p = 0.001 ). Conclusion: Hypertension, previous stroke, posteriorly projecting aneurysms, aneurysm size, and highly positioned aneurysms are independent risk factors for postoperative infarction during surgical clipping of an unruptured anterior communicating artery aneurysm. Additionally, a closed A2 plane is an additional risk factor of postoperative infarction in patients undergoing clipping via the pterional approach. [ABSTRACT FROM AUTHOR]

Endoscopic submucosal dissection (ESD), which enables curative en bloc resection of early gastrointestinal neoplasms, has been an attractive minimally invasive surgery during the past two decades. Large post‐ESD defects must be carefully managed to prevent adverse events (AEs). The major AEs comprise delayed bleeding (DB) and delayed perforation (DP), and overall AEs comprise DB plus DP. This review aimed to clarify the clinical efficacy and technical outcomes of endoscopic prophylactic closure for post‐ESD defects. We identified studies involving ≥10 patients up to March 2022 in which endoscopic closure was applied for gastric, duodenal, and colorectal post‐ESD defects. In the stomach, total rates of overall AEs and DB were significantly lower in the closure than non‐closure group. In the duodenum, total rates of overall AEs, DB, and DP were significantly lower in the closure group. In the colorectum, total rates of overall AEs and DB were significantly lower in the closure group. Closure techniques, categorized into three groups (clip‐based techniques, mechanical clipping, and surgical stitch‐based techniques), were illustrated. Endoscopic closure demonstrated a certain ability to reduce DB after gastric, duodenal, and colorectal ESD as well as DP after duodenal ESD. Considering closure‐associated costs, the indications and limitations of closure techniques should be further investigated. [ABSTRACT FROM AUTHOR]

The development of unmanned aerial vehicle (UAV) oblique photogrammetric technology provides a good foundation for the rapid construction of large-scale and high-definition real-scene 3D models. However, due to the limitations of the modeling process, irrelevant feature data cannot be eliminated in the modeling stage. The built models contain irrelevant features and model distortions caused by errors. At present, most existing clipping algorithms cannot effectively clip real-scene 3D models that are organized as a whole or with levels of detail (LODs). Therefore, this paper proposes a novel algorithm for clipping real-scene 3D models from any perspective based on clipping boundary lines that fit the surfaces of the models. The results of the clipping experiments for 3D models constructed with oblique UAV images show that this algorithm can effectively clip any part of the 3D models, that the clipping results of each level model closely fit the corresponding clipping boundary lines, and that the accuracy of the clipping results is very high. Additionally, the time complexity of the algorithm is O(n2). In conclusion, the algorithm proposed in this paper provides correct and effective clipping results for real-scene 3D models with LODs that are constructed with photogrammetric or 3D laser scanning data. [ABSTRACT FROM AUTHOR]

Automatic video understanding is a crucial piece of technology which promotes urban sustainability. Video grounding is a fundamental component of video understanding that has been evolving quickly in recent years, but its use is restricted due to the high labeling costs and typical performance limitations imposed by the pre-defined training dataset. In this paper, a novel atom-based zero-shot video grounding (AZVG) method is proposed to retrieve the segments in the video that correspond to a given input sentence. Although it is training-free, the performance of AZVG is competitive to the weakly supervised methods and better than unsupervised SOTA methods on the Charades-STA dataset. The method can support flexible queries as well as different video content. It can play an important role in a wider range of urban living applications. [ABSTRACT FROM AUTHOR]

Eklemeli imalat teknolojilerinden biri olan havuz fotopolimerizasyonu (HFP) sahip olduğu birçok avantaj nedeniyle hızla yaygınlaşmaktadır. İlk keşfedilen eklemeli imalat yöntemi olan SL (Stereolitography) keşfedildiği yıllarda prototip imalatı ve ürün geliştirme amaçlı kullanılmaktaydı. Sonraki yıllarda DLP (Digital Light Processing) ve CLIP (Continuous Light Interphase Printing) teknolojilerinin geliştirilmesi ile az adetli uygulamalara yönelik plastik parça imalatında da kullanılmaya başlamıştır. HFP teknolojileri diğer plastik eklemeli imalat teknolojilerine göre daha düşük yüzey pürüzlülüğü ve daha yüksek boyutsal hassasiyet sunmaktadır. HFP teknolojilerinin kompleks parçaların direkt imalatına izin vermesi, havacılık, otomotiv, sağlık, dişçilik, kuyumculuk gibi alanlarda kullanımlarını yaygınlaştırmaktadır. Uygulamaya özel geliştirilmiş malzemelerin kullanımı ile HFP teknolojileri hassas döküm, üretime yardımcı ekipmanlar, ortodonti diş kalıpları ve implant delme aparatları gibi uygulamalarda kullanılmaktadır. SL teknolojisinin ticari patentinin geçerliliğini yitirmesi ile açık kaynak kodlu ucuz masa üstü yazıcılar piyasaya sürülmüştür. Bu sayede HFP teknolojisi evlerimize kadar girmiştir. Bu çalışmada HFP teknolojileri ve kullanım alanları hakkında detaylı bilgiler verilmiştir.

Traditional supervised image captioning methods usually rely on a large number of images and paired captions for training. However, the creation of such datasets necessitates considerable temporal and human resources. Therefore, we propose a new semi-supervised image captioning algorithm to solve this problem. The proposed method uses a generative adversarial network to generate images that match captions, and uses these generated images and captions as new training data. This avoids the error accumulation problem when generating pseudo captions with autoregressive method and the network can directly perform backpropagation. At the same time, in order to ensure the correlation between the generated images and captions, we introduced the CLIP model for constraints. The CLIP model has been pre-trained on a large amount of image–text data, so it shows excellent performance in semantic alignment of images and text. To verify the effectiveness of our method, we validate on MSCOCO offline "Karpathy" test split. Experiment results show that our method can significantly improve the performance of the model when using 1% paired data, with the CIDEr score increasing from 69.5% to 77.7%. This shows that our method can effectively utilize unlabeled data for image caption tasks. • The proposed method is to generate images for captions instead of captions for images. • CLIP is used to constraint generator to associate images with captions. • Parameter updates can be completed through backpropagation. [ABSTRACT FROM AUTHOR]

In recent years, the field of few-shot action recognition (FSAR) has garnered significant attention. Although many methods primarily rely on mono-modal data, there is a growing trend towards utilizing multi-modal data. However, existing FSAR methods often employ simplistic fusion techniques, such as concatenation or comparison, which may not fully leverage the potential of multi-modal information. We aim to explore multi-modal information from different views and subsequently perform multi-view fusion. Based on the textual and visual modality information extracted by the CLIP backbone, we propose an MDMF method that comprehensively utilizes these modalities at two levels. At the first level, we extract visual information from two views: Local Temporal Context and Global Temporal Context. Within each view, we fuse visual features with textual information through concatenation and Cross-Transformer operations. We then employ metric comparison to derive probability distributions for classification under the meta-learning paradigm for each view. At the second level, we fuse the probability distributions from both views to make the final decision. Concurrently, during training, for each query, we calculate the posterior distributions of the textual and visual modalities within each view using text information distance and visual information distance. Based on these distributions, we group query samples with higher view reliability. Subsequently, we enhance the representation of the less reliable view of specific samples through mutual distillation. By delving deep into multi-modal data through a multi-view approach, our few-shot action recognition model demonstrates the potential for achieving higher accuracy and enhanced robustness. Our code is available at the URL: https://github.com/cofly2014/CLIP-MDMF. • This paper proposes Multi-view Distillation framework based on Multi-modal Fusion. • This paper introduces Probability Prompt Selector to address information inconsistency. • This paper uses two View Context Extractors with Cross-Transformer for fusion. • This paper employs distance fusion and mutual distillation to enhance multi-view performance. [ABSTRACT FROM AUTHOR]

Keywords: Swin-Transformer; CLIP; multimodal robotic; swimming action recognition; transfer learningDue to a production error, the email address of the corresponding author, Xiaoyu Yue, was mistakenly written as "yuexiayu@njtech.edu.cn". The correct email address is "yuexiaoyu@njtech.edu.cn". A correction has been made to the Correspondence section:*Correspondence:Xiaoyu Yueyuexiaoyu@njtech.edu.cnThe publisher apologizes for this mistake. The original version of this article has been updated. [Extracted from the article]

Intestinal perforation following endoscopic clip placement is an uncommon complication that has been reported in adult patients but, as far as we now, not in pediatric patients. An 8-month-old female infant with a history of metabolic encephalopathy was admitted with a convulsive disorder requiring mechanical ventilation. On the seventh day of hospitalization in the intensive care unit, the patient experienced three bowel movements with dark blood clots. Despite being hemodynamically stable upon physical examination, her hemoglobin level decreased to 8.6 g/dL. Upper endoscopy findings were normal, and exploratory laparoscopy ruled out Meckel's diverticulum or other visible parietal causes of bleeding. Lower gastrointestinal endoscopy revealed ulcers in the rectum, sigmoid, and terminal ileum, where hemostatic clips were placed. No repeated bleeding events occurred. However, 20 days later she presented with fever, abdominal pain, and distension. Abdominal radiography revealed pneumoperitoneum, leading to an emergency exploratory laparotomy. A perforation of the sigmoid colon from one of the clips was found. We resected the short segment of sigmoid colon that contained the perforation and did an end-to-end anastomosis. She recovered well and had no complications. Hemostatic clips placed endoscopically in the colon can lead to perforation even several weeks after placement. [ABSTRACT FROM AUTHOR]

Gastric intestinal metaplasia (GIM) is a frequent lesion of the gastrointestinal tract, and its clinical grading can significantly reduce the risk of gastric cancer. Pre-trained vision–language models have demonstrated impressive generalizability in many domains. However, they suffer from the following limitations when applied to GIM grading: 1) Small GIM lesions of endoscopic images and their colour features are difficult to represent with global representations alone, 2) medical concepts are highly specialized, and their dependencies and semantics cannot be precisely represented and 3) Because of the scarcity of medical prior knowledge, multi-modal prompts are over-reliant on random initialization and fail to maintain semantic consistency. To address the above challenges, we propose a self-supervised visual–textual prompt (SVTP) learning strategy for few-shot GIM grading. In particular, image and colour prompts are simultaneously configured in the visual branch to use the colour information in lesion areas. Meanwhile, a coupling function is introduced to implement the exchange of high-dimensional information. Then, to enhance the encoding of complex information and model the dependencies between medical concepts, text prompts and an adapter are integrated into the textual branch. Finally, a three-dimensional contrastive learning (3DCL) strategy is proposed to enrich medical prior knowledge of prompts and maintain their semantic consistency, where three-modal prompts are projected into a unified 3D space for contrastive learning. We verified SVTP on a public endoscopic image dataset and a private GIM grading dataset. Experimental results demonstrate that SVTP has achieved SOTA grading performance on all datasets. [ABSTRACT FROM AUTHOR]

Contrastive Language-Image Pretraining (CLIP) models have achieved significant success and have markedly improved the performance of various downstream tasks, including action recognition. However, how to effectively introduce knowledge into the field of action recognition remains an open question. In this work, an External and Priori Knowledge CLIP (EPK-CLIP) is proposed to introduce external knowledge into the model. To capture external knowledge, an external knowledge embedding module is proposed, which can generate and utilize human-object interaction relations as external knowledge, enabling the model to learn better features. Furthermore, the sparse regularization is introduced in the loss function, endowing the model with the ability to exploit the sparse priori knowledge inherent in the classification task. Finally, multiple inference module is proposed to obtain classification results from both direct and indirect perspectives. Specifically, the final classification result is obtained by fusing the outputs of different reasoning modules. Moreover, four external knowledge datasets: Kinetics-400-VC, Jester-VC, HMDB-51-VC, and UCF-101-VC are built and released for public usage, which is a multimodal extension of corresponding action datasets respectively. Under fully-supervised settings, our model achieves the top-1 accuracy of 84.3%, 97.1%, 82.9%, and 98.2% on Kinetics-400, Jester, HMDB-51, and UCF-101, respectively. In zero-shot experiments, our model also achieves state-of-the-art results, with top-1 accuracy of 51.6% and 77.7% on HMDB-51 and UCF-101, respectively. All related datasets and code can be found at https://github.com/geek12138/EPK-CLIP. • Build video descriptions as external knowledge and inject it into the model. • Embedding sparse priori knowledge reduces the search range of optimal solutions. • Reasoning from different angles to make the model stronger and more stable. • Build and publish four external knowledge datasets. [ABSTRACT FROM AUTHOR]

Objective: This retrospective study was performed to evaluate the feasibility and safety of surgically clipping intracranial aneurysms using a transcranial neuroendoscopic approach. Methods: A total of 229 patients with cerebral aneurysms were included in our study, all of whom were treated with clamping surgery at Wuhan University People’s Hospital. They were divided into neuroendoscopic and microscopic groups, according to whether or not neuroendoscopy was used for the clamping surgery. We statistically analyzed the patients’ baseline data, surgical outcomes, and complications, which were then evaluated to assess the treatment effect. Results: The baseline characteristics were not statistically significant, except for gender, for which the proportions of female patients in the two groups were 69 (56.1%) and 46 (43.4%). There were no patients with incomplete aneurysm clamping or parent vessel occlusion in the neuroendoscopic group, and there were 4 (3.8%) and 2 (1.9%) in the microscopic group, respectively; however, there was no statistically significant difference in the comparison of the two groups. The mean operative times of the two groups were 181 min and 154 min, respectively, and were statistically different. However, the mRS scores of the two groups showed no significant difference in patient prognosis. The differences in complications (including limb hemiplegia, hydrocephalus, vision loss, and intracranial infection) were not statistically significant, except for cerebral ischemia, for which the proportions of patients in the two groups were 8 (6.5%) and 16 (15.1%). Conclusions: Neuroendoscopy can provide clear visualization and multi-angle views during aneurysm clipping, which is helpful for ensuring adequate clipping and preventing complications.

In this study, we investigate the application of generative models to assist artificial agents, such as delivery drones or service robots, in visualising unfamiliar destinations solely based on textual descriptions. We explore the use of generative models, such as Stable Diffusion, and embedding representations, such as CLIP and VisualBERT, to compare generated images obtained from textual descriptions of target scenes with images of those scenes. Our research encompasses three key strategies: image generation, text generation, and text enhancement, the latter involving tools such as ChatGPT to create concise textual descriptions for evaluation. The findings of this study contribute to an understanding of the impact of combining generative tools with multi-modal embedding representations to enhance the artificial agent’s ability to recognise unknown scenes. Consequently, we assert that this research holds broad applications, particularly in drone parcel delivery, where an aerial robot can employ text descriptions to identify a destination. Furthermore, this concept can also be applied to other service robots tasked with delivering to unfamiliar locations, relying exclusively on user-provided textual descriptions.

We present a rare case of the clip migration after laparoscopic partial nephrectomy in which the clips were misdiagnosed as urinary stones. A 57-years-old female presented with right flank pain 3 months after the surgery. A computed tomography scan showed a 4 mm stone in the left upper ureter and hydronephrosis. Another 3 mm stone was also detected in the left upper calyx. Rigid ureteroscopy and flexible ureterorenoscopy revealed two LAPRA-TY clips, which were embedded in the upper ureter and calyx. The clips were fragmented by holmium laser lithotripsy and removed with stone basket. Pain and hydronephrosis are resolved at follow-up.

Transcriptome-wide analysis of RNA-binding partners is commonly achieved using UV crosslinking and immunoprecipitation (CLIP). Individual-nucleotide-resolution CLIP (iCLIP)enables identification of the specific position of the protein–RNA interaction. In addition to RNA-binding proteins (RBPs), microRNA (miRNA)–mRNA interactions also play a crucial role in the regulation of gene expression. Argonaute-2 (Ago2) mediates miRNA binding to a multitude of mRNA target sites, enabling the identification of miRNA–mRNA interactions by employing modified Ago2-CLIP protocols. Here, we describe an Ago2-specific CLIP protocol optimized for the use of small quantities of cell material, targeting endogenous Ago2 while avoiding possible methodological biases such as metabolic labeling or Ago2 overexpression and applying the latest advances in CLIP library preparation, the iCLIP2 protocol. In particular, we focus on the optimization of lysis conditions and improved radioactive labeling of the 5′ end of the miRNA. [ABSTRACT FROM AUTHOR]

Video clip retrieval and captioning tasks play an essential role in multimodal research and are the fundamental research problem for multimodal understanding and generation. The CLIP (Contrastive Language-Image Pre-training) model has demonstrated the power of visual concepts learning from web collected image-text datasets. In this paper, we propose a CLIP4Clip model to transfer the knowledge of the image-text pretrained CLIP model to video-text tasks in an end-to-end manner. Furthermore, we conduct several empirical studies including 1) Whether image feature is enough for video-text retrieval and captioning? 2) How a post-pretraining on a large-scale video-text dataset based on the CLIP affect the performance? 3) What is the practical mechanism to model temporal dependency between video frames? And 4) The Hyper-parameters sensitivity of the model. Extensive experimental results present that the CLIP4Clip model transferred from the CLIP can achieve SOTA results on various video-text datasets, including MSR-VTT, MSVD, LSMDC, and DiDeMo for multimodal understanding and generation tasks. [ABSTRACT FROM AUTHOR]

Protein fusions are frequently used for fluorescence imaging of individual molecules, both in vivo and in vitro. The SNAP, CLIP, HALO (aka HaloTag7), and DHFR protein tags can be linked to small molecule dyes that provide brightness and photo-stability superior to fluorescent proteins. To facilitate fluorescent dye tagging of proteins in the yeast Saccharomyces cerevisiae, we constructed a modular set of vectors with various combinations of labeling protein tags and selectable markers. These vectors can be used in combination to create strains where multiple proteins labeled with different colored dyes can be simultaneously observed. [ABSTRACT FROM AUTHOR]

Purpose: This meta-analysis aims to investigate the role of prophylactic clipping after endoscopic colorectal polypectomy or endoscopic mucosal resection (EMR) in prevention of delayed bleeding (DB) following polypectomy. Methods: We searched the PubMed, Embase, and Cochrane Library databases for randomized controlled trials comparing the effect of prophylactic clipping versus no clipping on DB since inception to 22nd April 2022. We then performed a meta-analysis using a random-effects model. Results: We included 8 studies with 5648 patients and 10,436 lesions. Prophylactic clipping did not reduce the overall risk of DB compared with no clipping (1.54% vs 2.05%; Log RR, −0.29; 95% confidence interval [CI], −0.59, 0.01; P = 0.06). In subgroup analyses, clipping significantly reduced DB rate in polyps ≥ 2 cm (Log RR, −0.63; 95% CI, −1.08, −0.18; P = 0.01), in non-pedunculated polyps (Log RR, −0.63; 95% CI, −1.01, −0.24; P = 0.00), and in large (≥ 2 cm) proximal polyps (Log RR, −0.81; 95% CI, −1.56, 0.05; P = 0.04), but not in polyps < 2 cm (Log RR, 0.01; 95% CI, −.40, 0.42; P = 0.95). Conclusion: Prophylactic clipping does not prevent post-polypectomy bleeding after all EMR and should not be performed as a routine practice. Although prophylactic clipping may reduce DB rate following resection of large proximal polyps and non-pedunculated polyps, more high-quality studies are needed to determine the effects of factors such as polyp location, polyp morphology, antithrombotic drug use and complete or partial closure on the effectiveness of prophylactic clipping. [ABSTRACT FROM AUTHOR]