Search

Stem cell-derived organoids are a promising tool to model native human tissues as they resemble human organs functionally and structurally compared to traditional monolayer cell-based assays. For instance, colon organoids can spontaneously develop crypt-like structures similar to those found in the native colon. While analyzing the structural development of organoids can be a valuable readout, using traditional image analysis tools makes it challenging because of the heterogeneities and the abstract nature of organoid morphologies. To address this limitation, we developed and validated a deep learning-based image analysis tool, named D-CryptO, for the classification of organoid morphology. D-CryptO can automatically assess the crypt formation and opacity of colorectal organoids from brightfield images to determine the extent of organoid structural maturity. To validate this tool, changes in organoid morphology were analyzed during organoid passaging and short-term forskolin stimulation. To further demonstrate the potential of D-CryptO for drug testing, organoid structures were analyzed following treatments with a panel of chemotherapeutic drugs. With D-CryptO, subtle variations in how colon organoids responded to the different chemotherapeutic drugs were detected, which suggest potentially distinct mechanisms of action. This tool could be expanded to other organoid types, like intestinal organoids, to facilitate 3D tissue morphological analysis.

Three-dimensional (3D) tissue models such as epithelial spheroids or organoids have become popular for pre-clinical drug studies. In contrast to 2D monolayer culture, the characterization of 3D tissue models from non-invasive brightfield images is a significant challenge. To address this issue, here we report a deep-learning uncovered measurement of epithelial networks (Deep-LUMEN) assay. Deep-LUMEN is an object detection algorithm that has been fine-tuned to automatically uncover subtle differences in epithelial spheroid morphology from brightfield images. This algorithm can track changes in the luminal structure of tissue spheroids and distinguish between polarized and non-polarized lung epithelial spheroids. The Deep-LUMEN assay was validated by screening for changes in spheroid epithelial architecture in response to different extracellular matrices and drug treatments. Specifically, we found the dose-dependent toxicity of cyclosporin can be underestimated if the effect of the drug on tissue morphology is not considered. Hence, Deep-LUMEN could be used to assess drug effects and capture morphological changes in 3D spheroid models in a non-invasive manner.

Stress field analysis is an essential part of umbilical component layout design. The stress field analysis of an umbilical, via numerical simulation, has commonly been applied in practical engineering. The high economic and time cost associated with numerical simulation and analysis of the stress field in an umbilical has been replaced by data-driven, deep-learning-based, real-time computational methods. In this study, a novel Pyramidal Efficient U-Net (PyEf-U-Net) network is proposed to predict the stress field distribution of the umbilical. The input dataset is obtained via the Differential Evolution-Generalized Lagrange Multiplier (DE-GLM) method, which is entered into the network for training, with a detailed discussion of the effects of hyperparameters such as optimizer, learning rate, and loss function on the performance of the network. The experimental research demonstrates that the proposed PyEf-U-Net can accurately predict the stress field of the umbilical in real time with a prediction accuracy of 94.2%, which is superior to other deep learning networks. The proposed method can provide an effective way for rapid mechanical analysis and design of the umbilical in practical engineering, while the method can be extended to the mechanical analysis and design of other similar marine engineering equipment structures. [ABSTRACT FROM AUTHOR]

Outside the explosive successful applications of deep learning (DL) in natural language processing, computer vision, and information retrieval, there have been numerous Deep Neural Networks (DNNs) based alternatives for common security-related scenarios with malware detection among more popular. Recently, adversarial learning has gained much focus. However, unlike computer vision applications, malware adversarial attack is expected to guarantee malwares' original maliciousness semantics. This paper proposes a novel adversarial instruction learning technique, DeepMal, based on an adversarial instruction learning approach for static malware detection. So far as we know, DeepMal is the first practical and systematical adversarial learning method, which could directly produce adversarial samples and effectively bypass static malware detectors powered by DL and machine learning (ML) models while preserving attack functionality in the real world. Moreover, our method conducts small-scale attacks, which could evade typical malware variants analysis (e.g., duplication check). We evaluate DeepMal on two real-world datasets, six typical DL models, and three typical ML models. Experimental results demonstrate that, on both datasets, DeepMal can attack typical malware detectors with the mean F1-score and F1-score decreasing maximal 93.94% and 82.86% respectively. Besides, three typical types of malware samples (Trojan horses, Backdoors, Ransomware) prove to preserve original attack functionality, and the mean duplication check ratio of malware adversarial samples is below 2.0%. Besides, DeepMal can evade dynamic detectors and be easily enhanced by learning more dynamic features with specific constraints. [ABSTRACT FROM AUTHOR]

Correction for 'Deep-LUMEN assay – human lung epithelial spheroid classification from brightfield images using deep learning' by Lyan Abdul et al., Lab Chip, 2020, DOI: 10.1039/d0lc01010c. [ABSTRACT FROM AUTHOR]

The moving morphable component (MMC) method, an important engineering structural optimization algorithm, achieves a boundary evolution through the migration and superposition of a series of moving morphable display components, resulting in structural optimization. Instead of the original linear skeleton thickness quadratic variation component, an elliptical initial component is used for topology optimization, which reduces the number of design variables of the initial component itself to achieve the purpose of reducing the computational time cost. As the number of initial components increases, the intermediate iterative computation becomes increasingly time-consuming; the initial components of the different parameters generate significant differences in the topological configuration. To obtain an optimal topology in an accurate and real-time manner, a Pyramid Attention U-Net (PA-U-Net) deep learning model is proposed to improve the optimization design and to avoid intermediate iterative computations. The dataset is generated by an improved MMC method for training. According to the results, the method not only obtains the optimal topology of the initial components under various parameter conditions in an accurate and timely manner and in a negligible amount of computing time, but it also achieves an accuracy of 90.89%, which is superior to other deep learning models. The organic combination of deep learning and topology optimization has many prospects in the design optimization of large engineering structures. • A novel architecture is proposed to improve the efficiency of topology optimization, i. e., a deep learning network PA-U-Net is proposed to replace the traditional topology optimization method to shorten the iterative optimization time, which provides a profitable way to obtain accurate and optimal topology configurations in real-time. • Content-aware pyramid feature extraction module and spatial attention mechanism are introduced, respectively, which can adaptively extract feature information of the initial components and capture abundant contextual features. [ABSTRACT FROM AUTHOR]

The global environmental changes have led to frequent occurrences of extreme rainfall, causing urban waterlogging. It is essential to quickly and accurately identify flooding as well as the depth levels. Therefore, Convolutional block attention module (CBAM)-improved ResNet50 is proposed for training to obtain flood depth levels (FDLs) recognition AI models. In this model, CBAM helps the model focus on important features for classifying FDLs, improving accuracy. It also streamlines the fully connected layers, reducing parameters, cutting computational costs, and preventing overfitting. A total of 6294 images related to urban waterlogging were collected for deep learning. The CBAM-improved ResNet50 algorithm achieved the highest test set accuracy at 92.45%, outperforming other algorithms including AlexNet (62.25%), MobileNet-V2 (64.67%), GoogleNet (75.5%), and ResNet50 (79.2%). It also demonstrates good predictive performance and generalization capabilities for urban waterlogging situations against various backgrounds. The computer vision recognition of the FDLs was explained through Gradient-weighted Class Activation Mapping in visualized images. Finally, through the established WeChat mini-program and AI model, the public can access the current FDLs at any time and place. This study provides information on the FDLs, offering a comprehensive overview of the severity of floods in the early stages of emergency response. • The CBAM-improved ResNet model is proposed for classifying flood depth levels. • AI visual model captures the contact features between flooding and reference objects. • WeChat mini-program + AI model: Public access to urban flooding features. [ABSTRACT FROM AUTHOR]

Monocular image-based 3D reconstruction is widely used in virtual reality, augmented reality, and autonomous driving, which benefits from the rapid development of deep learning approaches. Most of the available methods focused on reconstructing the overall shape of the object while ignoring some fine-grained details. Moreover, these methods make it hard to exactly reconstruct complex topological structures. In this paper, we propose a multi-granularity relationship reasoning network (MGRRNet), which aims to recover 3D shapes with high fidelity and rich details via the relationship reasoning between different granularity information. Specifically, our model captures the discriminative and detailed features at different granularities for extracting attentional regions. Then we perform the relationship reasoning between different granularities to reinforce the multi-granularity consistency and inter-granularity correlation. By doing this, our network is able to achieve robust feature representation and fine reconstruction. During the learning process, we jointly optimize procedures of different granularity feature representations via a sequence of inter-granularity cycle loss iterations. Extensive experimental results on two publicly available datasets justify that our approach achieves competitive performance compared to the state-of-the-art methods. Codes and all resources will be publicly available at https://github.com/Ray-tju/MGRRNet. • The MGRRNet achieves robust feature representation and fine reconstruction. • The MGRRNet captures key features and ensures consistency across granularities. • The proposed cycle loss optimizes feature representation across granularities. • The MGRRNet outperforms other methods in tests on two widely-evaluated datasets. [ABSTRACT FROM AUTHOR]