Deep Learning for Flow Sculpting: Insights into Efficient Learning using Scientific Simulation Data

Machine Learning-Driven Innovations in Microfluidics

Microfluidic devices have revolutionized biosensing by enabling precise manipulation of minute fluid volumes across diverse applications. This review investigates the incorporation of machine learning (ML) into the design, fabrication, and application of microfluidic biosensors, emphasizing how ML algorithms enhance performance by improving design accuracy, operational efficiency, and the management of complex diagnostic datasets. Integrating microfluidics with ML has fostered intelligent systems capable of automating experimental workflows, enabling real-time data analysis, and supporting informed decision-making. Recent advances in health diagnostics, environmental monitoring, and synthetic biology driven by ML are critically examined. This review highlights the transformative potential of ML-enhanced microfluidic systems, offering insights into the future trajectory of this rapidly evolving field.

AI for biofabrication

Biofabrication is an advanced technology that holds great promise for constructing highly biomimeticin vitrothree-dimensional human organs. Such technology would help address the issues of immune rejection and organ donor shortage in organ transplantation, aiding doctors in formulating personalized treatments for clinical patients and replacing animal experiments. Biofabrication typically involves the interdisciplinary application of biology, materials science, mechanical engineering, and medicine to generate large amounts of data and correlations that require processing and analysis. Artificial intelligence (AI), with its excellent capabilities in big data processing and analysis, can play a crucial role in handling and processing interdisciplinary data and relationships and in better integrating and applying them in biofabrication. In recent years, the development of the semiconductor and integrated circuit industries has propelled the rapid advancement of computer processing power. An AI program can learn and iterate multiple times within a short period, thereby gaining strong automation capabilities for a specific research content or issue. To date, numerous AI programs have been applied to various processes around biofabrication, such as extracting biological information, designing and optimizing structures, intelligent cell sorting, optimizing biomaterials and processes, real-time monitoring and evaluation of models, accelerating the transformation and development of these technologies, and even changing traditional research patterns. This article reviews and summarizes the significant changes and advancements brought about by AI in biofabrication, and discusses its future application value and direction.

Integrating machine learning and biosensors in microfluidic devices: A review

Microfluidic devices are increasingly widespread in the literature, being applied to numerous exciting applications, from chemical research to Point-of-Care devices, passing through drug development and clinical scenarios. Setting up these microenvironments, however, introduces the necessity of locally controlling the variables involved in the phenomena under investigation. For this reason, the literature has deeply explored the possibility of introducing sensing elements to investigate the physical quantities and the biochemical concentration inside microfluidic devices. Biosensors, particularly, are well known for their high accuracy, selectivity, and responsiveness. However, their signals could be challenging to interpret and must be carefully analysed to carry out the correct information. In addition, proper data analysis has been demonstrated even to increase biosensors' mentioned qualities. To this regard, machine learning algorithms are undoubtedly among the most suitable approaches to undertake this job, automatically learning from data and highlighting biosensor signals' characteristics at best. Interestingly, it was also demonstrated to benefit microfluidic devices themselves, in a new paradigm that the literature is starting to name "intelligent microfluidics", ideally closing this benefic interaction among these disciplines. This review aims to demonstrate the advantages of the triad paradigm microfluidics-biosensors-machine learning, which is still little used but has a great perspective. After briefly describing the single entities, the different sections will demonstrate the benefits of the dual interactions, highlighting the applications where the reviewed triad paradigm was employed.

Advancing Intelligent Organ-on-a-Chip Systems with Comprehensive In Situ Bioanalysis

In vitro models are essential to a broad range of biomedical research, such as pathological studies, drug development, and personalized medicine. As a potentially transformative paradigm for 3D in vitro models, organ-on-a-chip (OOC) technology has been extensively developed to recapitulate sophisticated architectures and dynamic microenvironments of human organs by applying the principles of life sciences and leveraging micro- and nanoscale engineering capabilities. A pivotal function of OOC devices is to support multifaceted and timely characterization of cultured cells and their microenvironments. However, in-depth analysis of OOC models typically requires biomedical assay procedures that are labor-intensive and interruptive. Herein, the latest advances toward intelligent OOC (iOOC) systems, where sensors integrated with OOC devices continuously report cellular and microenvironmental information for comprehensive in situ bioanalysis, are examined. It is proposed that the multimodal data in iOOC systems can support closed-loop control of the in vitro models and offer holistic biomedical insights for diverse applications. Essential techniques for establishing iOOC systems are surveyed, encompassing in situ sensing, data processing, and dynamic modulation. Eventually, the future development of iOOC systems featuring cross-disciplinary strategies is discussed.

High-throughput microfluidic systems accelerated by artificial intelligence for biomedical applications

High-throughput microfluidic systems are widely used in biomedical fields for tasks like disease detection, drug testing, and material discovery. Despite the great advances in automation and throughput, the large amounts of data generated by the high-throughput microfluidic systems generally outpace the abilities of manual analysis. Recently, the convergence of microfluidic systems and artificial intelligence (AI) has been promising in solving the issue by significantly accelerating the process of data analysis as well as improving the capability of intelligent decision. This review offers a comprehensive introduction on AI methods and outlines the current advances of high-throughput microfluidic systems accelerated by AI, covering biomedical detection, drug screening, and automated system control and design. Furthermore, the challenges and opportunities in this field are critically discussed as well.

Generation of Multiple Concentration Gradients Using a Two-Dimensional Pyramid Array

Concentration heterogeneity of diffusible reactants is a prevalent phenomenon in biochemical processes, requiring the generation of concentration gradients for the relevant experiments. In this study, we present a high-density pyramid array microfluidic network for the effective and precise generation of multiple concentration gradients. The complex gradient distribution in the 2D array can be adaptively adjusted by modulating the reactant velocities and concentrations at the inlets. In addition, the unique design of each reaction chamber and mixing block in the array ensures uniform concentrations within each chamber during dynamic changes, enabling large-scale reactions with low reactant volumes. Through detailed numerical simulation of mass transport within the complex microchannel networks, the proposed method allows researchers to determine the desired number of reaction chambers within a given concentration range based on experimental requirements and to quickly obtain the operating conditions with the help of machine learning-based prediction. The effectiveness in generating a multiple concentration gradient environment was further demonstrated by concentration-dependent calcium carbonate crystallization experiments. This device provides a highly efficient mixing and adaptable concentration platform that is well suited for high-throughput and multiplexed reactions.

Artificial neural network modelling hydrodenticity for optimal design by microfluidics of polymer nanoparticles to apply in magnetic resonance imaging

The engineering of nanoparticles impacts the control of their nano-bio interactions at each level of the delivery pathway. Therefore, optimal nanoparticle physicochemical properties should be identified to favour on-target interactions and deliver efficiently active compounds to a specific target. To date, traditional batch processes do not guarantee the reproducibility of results and low polydispersity index of the nanostructures, while microfluidics has emerged as cost effectiveness, short-production time approach to control the nanoparticle size and size distribution. Several thermodynamic processes have been implemented in microfluidics, such as nanoprecipitation, ionotropic gelation, self-assembly, etc., to produce nanoparticles in a continuous mode and high throughput way.   In this work, we show how the Artificial Neural Network (ANN) can be adopted to model the impact of microfluidic parameters (namely, flow rates and polymer concentrations) on the size of the nanoparticles. Promising results have been obtained, with the highest model accuracy reaching 98.9 %, thus confirming the proposed approach's potential applicability for an ANN-guided biopolymer nanoparticle design for biomedical applications. Nanostructures with different degrees of complexity are analysed, and a proof-of-concept machine learning approach is proposed to evaluate Hydrodenticity in biopolymer matrices. STATEMENT OF SIGNIFICANCE: Size, shape and surface charge determine nano-bio interactions of nanoparticles and their ability to target diseases. The ideal nanoparticle design avoids off-target interactions and favours on-target interactions. So, tools enabling the identification of the optimal nanoparticle physicochemical properties for delivery to a specific target are required. In this work, we evaluate the use of Artificial Neural Network (ANN) to analyse the role of microfluidic parameters in predicting the optimal size of the different hydrogel nanoparticles and their ability to trigger Hydrodenticity.

In situbiosensing technologies for an organ-on-a-chip

Thein vitrosimulation of organs resolves the accuracy, ethical, and cost challenges accompanyingin vivoexperiments. Organoids and organs-on-chips have been developed to model thein vitro, real-time biological and physiological features of organs. Numerous studies have deployed these systems to assess thein vitro, real-time responses of an organ to external stimuli. Particularly, organs-on-chips can be most efficiently employed in pharmaceutical drug development to predict the responses of organs before approving such drugs. Furthermore, multi-organ-on-a-chip systems facilitate the close representations of thein vivoenvironment. In this review, we discuss the biosensing technology that facilitates thein situ, real-time measurements of organ responses as readouts on organ-on-a-chip systems, including multi-organ models. Notably, a human-on-a-chip system integrated with automated multi-sensing will be established by further advancing the development of chips, as well as their assessment techniques.

Deep learning with microfluidics for on-chip droplet generation, control, and analysis

Droplet microfluidics has gained widespread attention in recent years due to its advantages of high throughput, high integration, high sensitivity and low power consumption in droplet-based micro-reaction. Meanwhile, with the rapid development of computer technology over the past decade, deep learning architectures have been able to process vast amounts of data from various research fields. Nowadays, interdisciplinarity plays an increasingly important role in modern research, and deep learning has contributed greatly to the advancement of many professions. Consequently, intelligent microfluidics has emerged as the times require, and possesses broad prospects in the development of automated and intelligent devices for integrating the merits of microfluidic technology and artificial intelligence. In this article, we provide a general review of the evolution of intelligent microfluidics and some applications related to deep learning, mainly in droplet generation, control, and analysis. We also present the challenges and emerging opportunities in this field.

Microsystem Advances through Integration with Artificial Intelligence

Microfluidics is a rapidly growing discipline that involves studying and manipulating fluids at reduced length scale and volume, typically on the scale of micro- or nanoliters. Under the reduced length scale and larger surface-to-volume ratio, advantages of low reagent consumption, faster reaction kinetics, and more compact systems are evident in microfluidics. However, miniaturization of microfluidic chips and systems introduces challenges of stricter tolerances in designing and controlling them for interdisciplinary applications. Recent advances in artificial intelligence (AI) have brought innovation to microfluidics from design, simulation, automation, and optimization to bioanalysis and data analytics. In microfluidics, the Navier-Stokes equations, which are partial differential equations describing viscous fluid motion that in complete form are known to not have a general analytical solution, can be simplified and have fair performance through numerical approximation due to low inertia and laminar flow. Approximation using neural networks trained by rules of physical knowledge introduces a new possibility to predict the physicochemical nature. The combination of microfluidics and automation can produce large amounts of data, where features and patterns that are difficult to discern by a human can be extracted by machine learning. Therefore, integration with AI introduces the potential to revolutionize the microfluidic workflow by enabling the precision control and automation of data analysis. Deployment of smart microfluidics may be tremendously beneficial in various applications in the future, including high-throughput drug discovery, rapid point-of-care-testing (POCT), and personalized medicine. In this review, we summarize key microfluidic advances integrated with AI and discuss the outlook and possibilities of combining AI and microfluidics.

The Synergy between Deep Learning and Organs-on-Chips for High-Throughput Drug Screening: A Review

Organs-on-chips (OoCs) are miniature microfluidic systems that have arguably become a class of advanced in vitro models. Deep learning, as an emerging topic in machine learning, has the ability to extract a hidden statistical relationship from the input data. Recently, these two areas have become integrated to achieve synergy for accelerating drug screening. This review provides a brief description of the basic concepts of deep learning used in OoCs and exemplifies the successful use cases for different types of OoCs. These microfluidic chips are of potential to be assembled as highly potent human-on-chips with complex physiological or pathological functions. Finally, we discuss the future supply with perspectives and potential challenges in terms of combining OoCs and deep learning for image processing and automation designs.

Liver-on-a-chip: Considerations, advances, and beyond

The liver is the largest internal organ in the human body with largest mass of glandular tissue. Modeling the liver has been challenging due to its variety of major functions, including processing nutrients and vitamins, detoxification, and regulating body metabolism. The intrinsic shortfalls of conventional two-dimensional (2D) cell culture methods for studying pharmacokinetics in parenchymal cells (hepatocytes) have contributed to suboptimal outcomes in clinical trials and drug development. This prompts the development of highly automated, biomimetic liver-on-a-chip (LOC) devices to simulate native liver structure and function, with the aid of recent progress in microfluidics. LOC offers a cost-effective and accurate model for pharmacokinetics, pharmacodynamics, and toxicity studies. This review provides a critical update on recent developments in designing LOCs and fabrication strategies. We highlight biomimetic design approaches for LOCs, including mimicking liver structure and function, and their diverse applications in areas such as drug screening, toxicity assessment, and real-time biosensing. We capture the newest ideas in the field to advance the field of LOCs and address current challenges.

Machine learning for microfluidic design and control

Microfluidics has developed into a mature field with applications across science and engineering, having particular commercial success in molecular diagnostics, next-generation sequencing, and bench-top analysis. Despite its ubiquity, the complexity of designing and controlling custom microfluidic devices present major barriers to adoption, requiring intuitive knowledge gained from years of experience. If these barriers were overcome, microfluidics could miniaturize biological and chemical research for non-experts through fully-automated platform development and operation. The intuition of microfluidic experts can be captured through machine learning, where complex statistical models are trained for pattern recognition and subsequently used for event prediction. Integration of machine learning with microfluidics could significantly expand its adoption and impact. Here, we present the current state of machine learning for the design and control of microfluidic devices, its possible applications, and current limitations.

Integration of reinforcement learning to realize functional variability of microfluidic systems

In this article, we present a proof-of-concept for microfluidic systems with high functional variability using reinforcement learning. By mathematically defining the objective of tasks, we demonstrate that the system can autonomously learn to behave according to its objectives. We applied Q-learning to a peristaltic micropump and showed that two different tasks can be performed on the same platform: adjusting the flow rate of the pump and manipulating the position of the particles. First, we performed typical micropumping with flow rate control. In this task, the system is rewarded according to the deviation between the average flow rate generated by the micropump and the target value. Therefore, the objective of the system is to maintain the target flow rate via an operation of the pump. Next, we demonstrate the micromanipulation of a small object (microbead) on the same platform. The objective was to manipulate the microbead position to the target area, and the system was rewarded for the success of the task. These results confirmed that the system learned to control the flow rate and manipulate the microbead to any randomly chosen target position. In particular, the manipulation technique is a new technology that does not require the use of structures such as wells or weirs. Therefore, this concept not only adds flexibility to the system but also contributes to the development of novel control methods to realize highly versatile microfluidic systems.

An Overview of Organs-on-Chips Based on Deep Learning

Microfluidic-based organs-on-chips (OoCs) are a rapidly developing technology in biomedical and chemical research and have emerged as one of the most advanced and promising in vitro models. The miniaturization, stimulated tissue mechanical forces, and microenvironment of OoCs offer unique properties for biomedical applications. However, the large amount of data generated by the high parallelization of OoC systems has grown far beyond the scope of manual analysis by researchers with biomedical backgrounds. Deep learning, an emerging area of research in the field of machine learning, can automatically mine the inherent characteristics and laws of "big data" and has achieved remarkable applications in computer vision, speech recognition, and natural language processing. The integration of deep learning in OoCs is an emerging field that holds enormous potential for drug development, disease modeling, and personalized medicine. This review briefly describes the basic concepts and mechanisms of microfluidics and deep learning and summarizes their successful integration. We then analyze the combination of OoCs and deep learning for image digitization, data analysis, and automation. Finally, the problems faced in current applications are discussed, and future perspectives and suggestions are provided to further strengthen this integration.

Exploiting machine learning for bestowing intelligence to microfluidics

Intelligent microfluidics is an emerging cross-discipline research area formed by combining microfluidics with machine learning. It uses the advantages of microfluidics, such as high throughput and controllability, and the powerful data processing capabilities of machine learning, resulting in improved systems in biotechnology and chemistry. Compared to traditional microfluidics using manual analysis methods, intelligent microfluidics needs less human intervention, and results in a more user-friendly experience with faster processing. There is a paucity of literature reviewing this burgeoning and highly promising cross-discipline. Therefore, we herein comprehensively and systematically summarize several aspects of microfluidic applications enabled by machine learning. We list the types of microfluidics used in intelligent microfluidic applications over the last five years, as well as the machine learning algorithms and the hardware used for training. We also present the most recent advances in key technologies, developments, challenges, and the emerging opportunities created by intelligent microfluidics.

Application of U-Net with Global Convolution Network Module in Computer-Aided Tongue Diagnosis

The rapid development of intelligent manufacturing provides strong support for the intelligent medical service ecosystem. Researchers are committed to building Wise Information Technology of 120 (WIT 120) for residents and medical personnel with the concept of simple smart medical care and through core technologies such as Internet of Things, Big Data Analytics, Artificial Intelligence, and microservice framework, to improve patient safety, medical quality, clinical efficiency, and operational benefits. Among them, how to use computers and deep learning technology to assist in the diagnosis of tongue images and realize intelligent tongue diagnosis has become a major trend. Tongue crack is an important feature of tongue states. Not only does change of tongue crack states reflect objectively and accurately changed circumstances of some typical diseases and TCM syndrome but also semantic segmentation of fissured tongue can combine the other features of tongue states to further improve tongue diagnosis systems' identification accuracy. Although computer tongue diagnosis technology has made great progress, there are few studies on the fissured tongue, and most of them focus on the analysis of tongue coating and body. In this paper, we do systematic and in-depth researches and propose an improved U-Net network for image semantic segmentation of fissured tongue. By introducing the Global Convolution Network module into the encoder part of U-Net, it solves the problem that the encoder part is relatively simple and cannot extract relatively abstract high-level semantic features. Finally, the method is verified by experiments. The improved U-Net network has a better segmentation effect and higher segmentation accuracy for fissured tongue image dataset. It can be used to design a computer-aided tongue diagnosis system.

Intelligent acoustofluidics enabled mini-bioreactors for human brain organoids

Acoustofluidics, by combining acoustics and microfluidics, provides a unique means to manipulate cells and liquids for broad applications in biomedical sciences and translational medicine. However, it is challenging to standardize and maintain excellent performance of current acoustofluidic devices and systems due to a multiplicity of factors including device-to-device variation, manual operation, environmental factors, sample variability, etc. Herein, to address these challenges, we propose "intelligent acoustofluidics" - an automated system that involves acoustofluidic device design, sensor fusion, and intelligent controller integration. As a proof-of-concept, we developed intelligent acoustofluidics based mini-bioreactors for human brain organoid culture. Our mini-bioreactors consist of three components: (1) rotors for contact-free rotation via an acoustic spiral phase vortex approach, (2) a camera for real-time tracking of rotational actions, and (3) a reinforcement learning-based controller for closed-loop regulation of rotational manipulation. After training the reinforcement learning-based controller in simulation and experimental environments, our mini-bioreactors can achieve the automated rotation of rotors in well-plates. Importantly, our mini-bioreactors can enable excellent control over rotational mode, direction, and speed of rotors, regardless of fluctuations of rotor weight, liquid volume, and operating temperature. Moreover, we demonstrated our mini-bioreactors can stably maintain the rotational speed of organoids during long-term culture, and enhance neural differentiation and uniformity of organoids. Comparing with current acoustofluidics, our intelligent system has a superior performance in terms of automation, robustness, and accuracy, highlighting the potential of novel intelligent systems in microfluidic experimentation.

Integrating Engineering, Automation, and Intelligence to Catalyze the Biomedical Translation of Organoids

Organoid technology has developed at an impressive speed during the past decade. Still, organoids are not widely used in practical applications as expected. It is believed that this translation can be greatly accelerated with the integration of engineering and artificial intelligence into current research practices. It is proposed that this approach is the missing link to realize key milestones in organoid technology, namely, high-throughput, homogeneous, and standardized production, automated manipulation, and intelligent monitoring, evaluation, and control via integrated on-chip instrumentation and artificial intelligence. It is suggested that organoids-on-a-chip are the ideal platform to achieve these feats. Once these techniques are established and adopted by the scientific community, the rapid translation of organoids may be seen from laboratories to the clinics and pharmaceutical industry.

Adoption of reinforcement learning for the intelligent control of a microfluidic peristaltic pump

We herein report a study on the intelligent control of microfluidic systems using reinforcement learning. Integrated microvalves are utilized to realize a variety of microfluidic functional modules, such as switching of flow pass, micropumping, and micromixing. The application of artificial intelligence to control microvalves can potentially contribute to the expansion of the versatility of microfluidic systems. As a preliminary attempt toward this motivation, we investigated the application of a reinforcement learning algorithm to microperistaltic pumps. First, we assumed a Markov property for the operation of diaphragms in the microperistaltic pump. Thereafter, components of the Markov decision process were defined for adaptation to the micropump. To acquire the pumping sequence, which maximizes the flow rate, the reward was defined as the obtained flow rate in a state transition of the microvalves. The present system successfully empirically determines the optimal sequence, which considers the physical characteristics of the components of the system that the authors did not recognize. Therefore, it was proved that reinforcement learning could be applied to microperistaltic pumps and is promising for the operation of larger and more complex microsystems.

Deep learning-based automated and universal bubble detection and mask extraction in complex two-phase flows

While investigating multiphase flows experimentally, the spatiotemporal variation in the interfacial shape between different phases must be measured to analyze the transport phenomena. For this, numerous image processing techniques have been proposed, showing good performance. However, they require trial-and-error optimization of thresholding parameters, which are not universal for all experimental conditions; thus, their accuracy is highly dependent on human experience, and the overall processing cost is high. Motivated by the remarkable improvements in deep learning-based image processing, we trained the Mask R-CNN to develop an automated bubble detection and mask extraction tool that works universally in gas–liquid two-phase flows. The training dataset was rigorously optimized to improve the model performance and delay overfitting with a finite amount of data. The range of detectable bubble size (particularly smaller bubbles) could be extended using a customized weighted loss function. Validation with different bubbly flows yields promising results, with AP₅₀ reaching 98%. Even while testing with bubble-swarm flows not included in the training set, the model detects more than 95% of the bubbles, which is equivalent or superior to conventional image processing methods. The pure processing speed for mask extraction is more than twice as fast as conventional approaches, even without counting the time required for tedious threshold parameter tuning. The present bubble detection and mask extraction tool is available online (https://github.com/ywflow/BubMask).

Automated optimization of double heater convective polymerase chain reaction devices based on CFD simulation database and artificial neural network model

This paper presents a framework for automated optimization of double-heater convective PCR (DH-cPCR) devices by developing a computational fluid dynamics (CFD) simulation database and artificial neural network (ANN) model. The optimization parameter space that includes the capillary tube geometries and the heater sizes of DH-cPCR is established, and a database consisting of nearly 10,000 CFD simulations is constructed. The database is then used to train a two-stage ANN models that select practically relevant data for modeling and predict PCR device performance. The trained ANN model is then combined with the gradient-based and the heuristics optimization approaches to search for optimal device configuration that possesses the shortest DNA doubling time. The entire design process including model meshing and configuration, parallel CFD computation, database organization, and ANN training and utilization is fully automated. Case studies confirm that the proposed framework can successfully find the optimal device configuration with an error of less than 0.3 s, and hence, representing a cost-effective and rapid solution of DH-cPCR device design.

Fast inverse design of microstructures via generative invariance networks

The problem of the efficient design of material microstructures exhibiting desired properties spans a variety of engineering and science applications. The ability to rapidly generate microstructures that exhibit user-specified property distributions can transform the iterative process of traditional microstructure-sensitive design. We reformulate the microstructure design process using a constrained generative adversarial network (GAN) model. This approach explicitly encodes invariance constraints within GANs to generate two-phase morphologies for photovoltaic applications obeying design specifications: specifically, user-defined short-circuit current density and fill factor combinations. Such invariance constraints can be represented by differentiable, deep learning-based surrogates of full physics models mapping microstructures to photovoltaic properties. Furthermore, we propose a multi-fidelity surrogate that reduces expensive label requirements by a factor of five. Our framework enables the incorporation of expensive or non-differentiable constraints for the fast generation of microstructures (in 190 ms) with user-defined properties. Such proposed physics-aware data-driven methods for inverse design problems can be used to considerably accelerate the field of microstructure-sensitive design.

A tongue features fusion approach to predicting prediabetes and diabetes with machine learning

BACKGROUND

Diabetics has become a serious public health burden in China. Multiple complications appear with the progression of diabetics pose a serious threat to the quality of human life and health. We can prevent the progression of prediabetics to diabetics and delay the progression to diabetics by early identification of diabetics and prediabetics and timely intervention, which have positive significance for improving public health.

OBJECTIVE

Using machine learning techniques, we establish the noninvasive diabetics risk prediction model based on tongue features fusion and predict the risk of prediabetics and diabetics.

METHODS

Applying the type TFDA-1 Tongue Diagnosis Instrument, we collect tongue images, extract tongue features including color and texture features using TDAS, and extract the advanced tongue features with ResNet-50, achieve the fusion of the two features with GA_XGBT, finally establish the noninvasive diabetics risk prediction model and evaluate the performance of testing effectiveness.

RESULTS

Cross-validation suggests the best performance of GA_XGBT model with fusion features, whose average CA is 0.821, the average AUROC is 0.924, the average AUPRC is 0.856, the average Precision is 0.834, the average Recall is 0.822, the average F1-score is 0.813. Test set suggests the best testing performance of GA_XGBT model, whose average CA is 0.81, the average AUROC is 0.918, the average AUPRC is 0.839, the average Precision is 0.821, the average Recall is 0.81, the average F1-score is 0.796. When we test prediabetics with GA_XGBT model, we find that the AUROC is 0.914, the Precision is 0.69, the Recall is 0.952, the F1-score is 0.8. When we test diabetics with GA_XGBT model, we find that the AUROC is 0.984, the Precision is 0.929, the Recall is 0.951, the F1-score is 0.94.

CONCLUSIONS

Based on tongue features, the study uses classical machine learning algorithm and deep learning algorithm to maximum the respective advantages. We combine the prior knowledge and potential features together, establish the noninvasive diabetics risk prediction model with features fusion algorithm, and detect prediabetics and diabetics noninvasively. Our study presents a feasible method for establishing the association between diabetics and the tongue image information and prove that tongue image information is a potential marker which facilitates effective early diagnosis of prediabetics and diabetics.

Microfluidics in Biotechnology: Quo Vadis

The emerging technique of microfluidics offers new approaches for precisely controlling fluidic conditions on a small scale, while simultaneously facilitating data collection in both high-throughput and quantitative manners. As such, the so-called lab-on-a-chip (LOC) systems have the potential to revolutionize the field of biotechnology. But what needs to happen in order to truly integrate them into routine biotechnological applications? In this chapter, some of the most promising applications of microfluidic technology within the field of biotechnology are surveyed, and a few strategies for overcoming current challenges posed by microfluidic LOC systems are examined. In addition, we also discuss the intensifying trend (across all biotechnology fields) of using point-of-use applications which is being facilitated by new technological achievements.

Learning from droplet flows in microfluidic channels using deep neural networks

A non-intrusive method is presented for measuring different fluidic properties in a microfluidic chip by optically monitoring the flow of droplets. A neural network is used to extract the desired information from the images of the droplets. We demonstrate the method in two applications: measurement of the concentration of each component of a water/alcohol mixture, and measurement of the flow rate of the same mixture. A large number of droplet images are recorded and used to train deep neural networks (DNN) to predict the flow rate or the concentration. It is shown that this method can be used to quantify the concentrations of each component with a 0.5% accuracy and the flow rate with a resolution of 0.05 ml/h. The proposed method can in principle be used to measure other properties of the fluid such as surface tension and viscosity.

Small asymmetric Brownian objects self-align in nanofluidic channels

Although the self-alignment of asymmetric macro-sized objects of a few tens of microns in size have been studied extensively in experiments and theory, access to much smaller length scales is still hindered by technical challenges. We combine molecular dynamics and stochastic rotation dynamics techniques to investigate the self-orientation phenomenon at different length scales, ranging from the micron to the nano scale by progressively increasing the relative strength of diffusion over convection. To this end, we model an asymmetric dumbbell particle in Hele-Shaw flow and explore a wide range of Péclet numbers (Pe) and different particle shapes, as characterized by the size ratio of the two dumbbell spheres (R[combining tilde]). By independently varying these two parameters we analyse the process of self-orientation and characterize the alignment of the dumbbell with the direction of the fluid flow. We identify three different regimes of strong, weak and no alignment and we map out a state diagram in Pe versus R[combining tilde] plane. Based on these results, we estimate dimensional length scales and flow rates for which these findings would be applicable in experiments. Finally, we find that the characteristic reorientation time of the dumbbell is a monotonically decreasing function of the dumbbell anisotropy.

A deep learning framework to discern and count microscopic nematode eggs

In order to identify and control the menace of destructive pests via microscopic image-based identification state-of-the art deep learning architecture is demonstrated on the parasitic worm, the soybean cyst nematode (SCN), Heterodera glycines. Soybean yield loss is negatively correlated with the density of SCN eggs that are present in the soil. While there has been progress in automating extraction of egg-filled cysts and eggs from soil samples counting SCN eggs obtained from soil samples using computer vision techniques has proven to be an extremely difficult challenge. Here we show that a deep learning architecture developed for rare object identification in clutter-filled images can identify and count the SCN eggs. The architecture is trained with expert-labeled data to effectively build a machine learning model for quantifying SCN eggs via microscopic image analysis. We show dramatic improvements in the quantification time of eggs while maintaining human-level accuracy and avoiding inter-rater and intra-rater variabilities. The nematode eggs are correctly identified even in complex, debris-filled images that are often difficult for experts to identify quickly. Our results illustrate the remarkable promise of applying deep learning approaches to phenotyping for pest assessment and management.

An explainable deep machine vision framework for plant stress phenotyping

Current approaches for accurate identification, classification, and quantification of biotic and abiotic stresses in crop research and production are predominantly visual and require specialized training. However, such techniques are hindered by subjectivity resulting from inter- and intrarater cognitive variability. This translates to erroneous decisions and a significant waste of resources. Here, we demonstrate a machine learning framework's ability to identify and classify a diverse set of foliar stresses in soybean [Glycine max (L.) Merr.] with remarkable accuracy. We also present an explanation mechanism, using the top-K high-resolution feature maps that isolate the visual symptoms used to make predictions. This unsupervised identification of visual symptoms provides a quantitative measure of stress severity, allowing for identification (type of foliar stress), classification (low, medium, or high stress), and quantification (stress severity) in a single framework without detailed symptom annotation by experts. We reliably identified and classified several biotic (bacterial and fungal diseases) and abiotic (chemical injury and nutrient deficiency) stresses by learning from over 25,000 images. The learned model is robust to input image perturbations, demonstrating viability for high-throughput deployment. We also noticed that the learned model appears to be agnostic to species, seemingly demonstrating an ability of transfer learning. The availability of an explainable model that can consistently, rapidly, and accurately identify and quantify foliar stresses would have significant implications in scientific research, plant breeding, and crop production. The trained model could be deployed in mobile platforms (e.g., unmanned air vehicles and automated ground scouts) for rapid, large-scale scouting or as a mobile application for real-time detection of stress by farmers and researchers.

A deep learning framework for causal shape transformation

Recurrent neural network (RNN) and Long Short-term Memory (LSTM) networks are the common go-to architecture for exploiting sequential information where the output is dependent on a sequence of inputs. However, in most considered problems, the dependencies typically lie in the latent domain which may not be suitable for applications involving the prediction of a step-wise transformation sequence that is dependent on the previous states only in the visible domain with a known terminal state. We propose a hybrid architecture of convolution neural networks (CNN) and stacked autoencoders (SAE) to learn a sequence of causal actions that nonlinearly transform an input visual pattern or distribution into a target visual pattern or distribution with the same support and demonstrated its practicality in a real-world engineering problem involving the physics of fluids. We solved a high-dimensional one-to-many inverse mapping problem concerning microfluidic flow sculpting, where the use of deep learning methods as an inverse map is very seldom explored. This work serves as a fruitful use-case to applied scientists and engineers in how deep learning can be beneficial as a solution for high-dimensional physical problems, and potentially opening doors to impactful advance in fields such as material sciences and medical biology where multistep topological transformations is a key element.