From Algorithm to Hardware: A Survey on Efficient and Safe Deployment of Deep Neural Networks.

Deep neural networks (DNNs) have been widely used in many artificial intelligence (AI) tasks. However, deploying them brings significant challenges due to the huge cost of memory, energy, and computation. To address these challenges, researchers have developed various model compression techniques such as model quantization and model pruning. Recently, there has been a surge in research on compression methods to achieve model efficiency while retaining performance. Furthermore, more and more works focus on customizing the DNN hardware accelerators to better leverage the model compression techniques. In addition to efficiency, preserving security and privacy is critical for deploying DNNs. However, the vast and diverse body of related works can be overwhelming. This inspires us to conduct a comprehensive survey on recent research toward the goal of high-performance, cost-efficient, and safe deployment of DNNs. Our survey first covers the mainstream model compression techniques, such as model quantization, model pruning, knowledge distillation, and optimizations of nonlinear operations. We then introduce recent advances in designing hardware accelerators that can adapt to efficient model compression approaches. In addition, we discuss how homomorphic encryption can be integrated to secure DNN deployment. Finally, we discuss several issues, such as hardware evaluation, generalization, and integration of various compression approaches. Overall, we aim to provide a big picture of efficient DNNs from algorithm to hardware accelerators and security perspectives.

Research on the Algorithm of Position Correction for High-Speed Moving Express Packages Based on Traditional Vision and AI Vision.

The rapid development of the logistics industry poses significant challenges to the sorting work within this sector. The fast and precise identification of moving express parcels holds immense significance for the performance of logistics sorting systems. This paper proposes a motion express parcel positioning algorithm that combines traditional vision and AI-based vision. In the traditional vision aspect, we employ a brightness-based traditional visual parcel detection algorithm. In the AI vision aspect, we introduce a Convolutional Block Attention Module (CBAM) and Focal-EIoU to enhance YOLOv5, improving the model's recall rate and robustness. Additionally, we adopt an Optimal Transport Assignment (OTA) label assignment strategy to provide a training dataset based on global optimality for the model training phase. Our experimental results demonstrate that our modified AI model surpasses traditional algorithms in both parcel recognition accuracy and inference speed. The combined approach of traditional vision and AI vision in the motion express parcel positioning algorithm proves applicable for practical logistics sorting systems.

CoIn: Correlation Induced Clustering for Cognition of High Dimensional Bioinformatics Data.

Analysis of high dimensional biomedical data such as microarray gene expression data and mass spectrometry images, is crucial to provide better medical services including cancer subtyping, protein homology detection, etc. Clustering is a fundamental cognitive task which aims to group unlabeled data into multiple clusters based on their intrinsic similarities. However, for most clustering methods, including the most widely used K-means algorithm, all features of the high dimensional data are considered equally in relevance, which distorts the performance when clustering high-dimensional data where there exist many redundant variables and correlated variables. In this paper, we aim at addressing the problem of the high dimensional bioinformatics data clustering and propose a new correlation induced clustering, CoIn, to capture complex correlations among high dimensional data and guarantee the correlation consistency within each cluster. We evaluate the proposed method on a high dimensional mass spectrometry dataset of liver cancer tumor to explore the metabolic differences on tissues and discover the intra-tumor heterogeneity (ITH). By comparing the results of baselines and ours, it has been found that our method produces more explainable and understandable results for clinical analysis, which demonstrates the proposed clustering paradigm has the potential with application to knowledge discovery in high dimensional bioinformatics data.

LE-UDA: Label-Efficient Unsupervised Domain Adaptation for Medical Image Segmentation.

While deep learning methods hitherto have achieved considerable success in medical image segmentation, they are still hampered by two limitations: (i) reliance on large-scale well-labeled datasets, which are difficult to curate due to the expert-driven and time-consuming nature of pixel-level annotations in clinical practices, and (ii) failure to generalize from one domain to another, especially when the target domain is a different modality with severe domain shifts. Recent unsupervised domain adaptation (UDA) techniques leverage abundant labeled source data together with unlabeled target data to reduce the domain gap, but these methods degrade significantly with limited source annotations. In this study, we address this underexplored UDA problem, investigating a challenging but valuable realistic scenario, where the source domain not only exhibits domain shift w.r.t. the target domain but also suffers from label scarcity. In this regard, we propose a novel and generic framework called "Label-Efficient Unsupervised Domain Adaptation" (LE-UDA). In LE-UDA, we construct self-ensembling consistency for knowledge transfer between both domains, as well as a self-ensembling adversarial learning module to achieve better feature alignment for UDA. To assess the effectiveness of our method, we conduct extensive experiments on two different tasks for cross-modality segmentation between MRI and CT images. Experimental results demonstrate that the proposed LE-UDA can efficiently leverage limited source labels to improve cross-domain segmentation performance, outperforming state-of-the-art UDA approaches in the literature.

Self-supervised Assisted Active Learning for Skin Lesion Segmentation.

Label scarcity has been a long-standing issue for biomedical image segmentation, due to high annotation costs and professional requirements. Recently, active learning (AL) strategies strive to reduce annotation costs by querying a small portion of data for annotation, receiving much traction in the field of medical imaging. However, most of the existing AL methods have to initialize models with some randomly selected samples followed by active selection based on various criteria, such as uncertainty and diversity. Such random-start initialization methods inevitably introduce under-value redundant samples and unnecessary annotation costs. For the purpose of addressing the issue, we propose a novel self-supervised assisted active learning framework in the cold-start setting, in which the segmentation model is first warmed up with self-supervised learning (SSL), and then SSL features are used for sample selection via latent feature clustering without accessing labels. We assess our proposed methodology on skin lesions segmentation task. Extensive experiments demonstrate that our approach is capable of achieving promising performance with substantial improvements over existing baselines. Clinical Relevance- The proposed method can smartly select samples to annotate without requiring labels for model initialization, which can save annotation costs in clinical practice.

ViRBase v3.0: a virus and host ncRNA-associated interaction repository with increased coverage and annotation.

As a means to aid in the investigation of viral infection mechanisms and identification of more effective antivirus targets, the availability of a source which continually collects and updates information on the virus and host ncRNA-associated interaction resources is essential. Here, we update the ViRBase database to version 3.0 (http://www.virbase.org/ or http://www.rna-society.org/virbase/). This update represents a major revision: (i) the total number of interaction entries is now greater than 820,000, an approximately 70-fold increment, involving 116 virus and 36 host organisms, (ii) it supplements and provides more details on RNA annotations (including RNA editing, RNA localization and RNA modification), ncRNA SNP and ncRNA-drug related information and (iii) it provides two additional tools for predicting binding sites (IntaRNA and PRIdictor), a visual plug-in to display interactions and a website which is optimized for more practical and user-friendly operation. Overall, ViRBase v3.0 provides a more comprehensive resource for virus and host ncRNA-associated interactions enabling researchers a more effective means for investigation of viral infections.

Hierarchical Consistency Regularized Mean Teacher for Semi-supervised 3D Left Atrium Segmentation.

Deep learning has achieved promising segmentation performance on 3D left atrium MR images. However, annotations for segmentation tasks are expensive, costly and difficult to obtain. In this paper, we introduce a novel hierarchical consistency regularized mean teacher framework for 3D left atrium segmentation. In each iteration, the student model is optimized by multi-scale deep supervision and hierarchical consistency regularization, concurrently. Extensive experiments have shown that our method achieves competitive performance as compared with full annotation, outperforming other state-of-the-art semi-supervised segmentation methods.

DSAL: Deeply Supervised Active Learning From Strong and Weak Labelers for Biomedical Image Segmentation.

Image segmentation is one of the most essential biomedical image processing problems for different imaging modalities, including microscopy and X-ray in the Internet-of-Medical-Things (IoMT) domain. However, annotating biomedical images is knowledge-driven, time-consuming, and labor-intensive, making it difficult to obtain abundant labels with limited costs. Active learning strategies come into ease the burden of human annotation, which queries only a subset of training data for annotation. Despite receiving attention, most of active learning methods still require huge computational costs and utilize unlabeled data inefficiently. They also tend to ignore the intermediate knowledge within networks. In this work, we propose a deep active semi-supervised learning framework, DSAL, combining active learning and semi-supervised learning strategies. In DSAL, a new criterion based on deep supervision mechanism is proposed to select informative samples with high uncertainties and low uncertainties for strong labelers and weak labelers respectively. The internal criterion leverages the disagreement of intermediate features within the deep learning network for active sample selection, which subsequently reduces the computational costs. We use the proposed criteria to select samples for strong and weak labelers to produce oracle labels and pseudo labels simultaneously at each active learning iteration in an ensemble learning manner, which can be examined with IoMT Platform. Extensive experiments on multiple medical image datasets demonstrate the superiority of the proposed method over state-of-the-art active learning methods.

Multi-Instance Multi-Label Learning for Gene Mutation Prediction in Hepatocellular Carcinoma.

Gene mutation prediction in hepatocellular carcinoma (HCC) is of great diagnostic and prognostic value for personalized treatments and precision medicine. In this paper, we tackle this problem with multi-instance multi-label learning to address the difficulties on label correlations, label representations, etc. Furthermore, an effective oversampling strategy is applied for data imbalance. Experimental results have shown the superiority of the proposed approach.

A ratiometric fluorescent hydrogen peroxide chemosensor manipulated by an ICT-activated FRET mechanism and its bioimaging application in living cells and zebrafish.

A H2O2-responsive fluorescent chemosensor (CNBE) with a ratiometric emission signal was elaborately designed and synthesized. The ratio signal of the chemosensor was manipulated by an interplaying ICT-activated FRET mechanism. The ratiometric fluorescence imaging was successfully applied to detect H2O2 using CNBE in living cells and zebrafish.

A multifunctional logic gate by means of a triple-chromophore fluorescent biothiol probe with diverse fluorescence signal patterns.

A novel multifunctional logic gate based on a triple-chromophore (coumarin-NBD-flavylium, CNF) fluorescent biothiol probe with diverse fluorescence signal patterns was rationally designed and synthetized. On the new triad CNF, diverse logic operations such as OR, TRANSFER, INH, NOT, and YES logic gates were achieved by using biothiols and fluorescence signal patterns as the multiple inputs and outputs, respectively.

A multi-signal fluorescent probe for simultaneously distinguishing and sequentially sensing cysteine/homocysteine, glutathione, and hydrogen sulfide in living cells.

Biothiols, which have a close network of generation and metabolic pathways among them, are essential reactive sulfur species (RSS) in the cells and play vital roles in human physiology. However, biothiols possess highly similar chemical structures and properties, resulting in it being an enormous challenge to simultaneously discriminate them from each other. Herein, we develop a unique fluorescent probe (HMN) for not only simultaneously distinguishing Cys/Hcy, GSH, and H2S from each other, but also sequentially sensing Cys/Hcy/GSH and H2S using a multi-channel fluorescence mode for the first time. When responding to the respective biothiols, the robust probe exhibits multiple sets of fluorescence signals at three distinct emission bands (blue-green-red). The new probe can also sense H2S at different concentration levels with changes of fluorescence at the blue and red emission bands. In addition, the novel probe HMN is able to discriminate and sequentially sense biothiols in biological environments via three-color fluorescence imaging. We expect that the development of the robust probe HMN will provide a powerful strategy to design fluorescent probes for the discrimination and sequential detection of biothiols, and offer a promising tool for exploring the interrelated roles of biothiols in various physiological and pathological conditions.

Improved Aromatic Substitution-Rearrangement-Based Ratiometric Fluorescent Cysteine-Specific Probe and Its Application of Real-Time Imaging under Oxidative Stress in Living Zebrafish.

Biothiols, including cysteine (Cys), homocysteine (Hcy), and glutathione (GSH), play a crucial role in many physiological processes. Cys production and metabolism is closely connected with Hcy and GSH; meanwhile, the dynamic antioxidant defenses network by Cys is independent of the GSH system, and Cys can serve as a more effective biomarker of oxidative stress. Hence, it is significant and urgent to develop an efficient method for specific detection of Cys over other biothiols (Hcy/GSH). However, most of the present Cys-specific fluorescent probes distinguished Cys from Hcy through response time, which would suffer from an unavoidable interference from Hcy in long-time detection. In this work, in order to improve the selectivity, we employed an improved aromatic substitution-rearrangement strategy to develop a ratiometric Cys-specific fluorescent probe (Cou-SBD-Cl) based on a new fluorescence resonance energy transfer (FRET) coumarin-sulfonyl benzoxadiazole (Cou-SBD) platform for discrimination of Hcy and GSH. Response of Cou-SBD-Cl to Cys would switch FRET on and generate a new yellow fluorescence emission with a 56.1-fold enhancement of ratio signal and a 99 nm emission shift. The desirable dual-color ratiometric imaging was achieved in living cells and normal zebrafish. In addition, probe Cou-SBD-Cl was also applied to real-time monitor Cys fluctuation in lipopolysaccharide-mediated oxidative stress in zebrafish.

A fluorescent dyad with large emission shift for discrimination of cysteine/homocysteine from glutathione and hydrogen sulfide and the application of bioimaging.

Biothiols, as reactive sulfur species (RSS), play important roles in human physiology, and they have a close connection of generation and metabolism pathways among of them. It is challenging to discriminate biothiols from each other due to the similar chemical structures and properties of them. Herein, we develop a fluorescent hybrid dyad (CS-NBD) for efficiently discriminating cysteine (Cys)/homocysteine (Hcy) from glutathione (GSH) and hydrogen sulfide (H2S) by a dual-channel detection method. CS-NBD performs inherently no fluorescence in ranging from visible to near infrared region. However, upon addition of Cys (2-150 µM)/Hcy (2-200 µM), CS-NBD generates significant fluorescence enhancement in two distinct emission bands (Green-Red), while encounter of GSH (2-100 µM) or H2S (2-70 µM) induces the fluorescence increase only in the red channel. The detection limit was determined to be 0.021 µM for Cys, 0.037 µM for Hcy, 0.028 µM for GSH, and 0.015 µM for H2S, respectively (S/N = 3). The interval distance between two emission bands is up to 163 nm, which is favourable to acquire the accurate data in measurement due to the reducing of crosstalk signals. CS-NBD is also successfully applied to distinguish Cys/Hcy in cellular context by dual-color fluorescence imaging.

A mitochondria-targeted fluorescent probe for imaging endogenous malondialdehyde in HeLa cells and onion tissues.

The mitochondria-targeted turn-on fluorescent probe (Mito-FMP) based on a benzoxadiazole platform was developed for detection of malondialdehyde (MDA). Mito-FMP performed with large enhancement of the optical signal (774-fold) in response to MDA in an aqueous system and has the capability of monitoring endogenous MDA in HeLa cells and onion tissues.