Improving biosensor accuracy and speed using dynamic signal change and theory-guided deep learning.

False results and time delay are longstanding challenges in biosensing. While classification models and deep learning may provide new opportunities for improving biosensor performance, such as measurement confidence and speed, it remains a challenge to ensure that predictions are explainable and consistent with domain knowledge. Here, we show that consistency of deep learning classification model predictions with domain knowledge in biosensing can be achieved by cost function supervision and enables rapid and accurate biosensing using the biosensor dynamic response. The impact and utility of the methodology were validated by rapid and accurate quantification of microRNA (let-7a) across the nanomolar (nM) to femtomolar (fM) concentration range using the dynamic response of cantilever biosensors. Data augmentation and cost function supervision based on the consistency of model predictions and experimental observations with the theory of surface-based biosensors improved the F1 score, precision, and recall of a recurrent neural network (RNN) classifier by an average of 13.8%. The theory-guided RNN (TGRNN) classifier enabled quantification of target analyte concentration and false results with an average prediction accuracy, precision, and recall of 98.5% using the initial transient or entire dynamic response, which is indicative of high prediction accuracy and low probability of false-negative and false-positive results. Classification scores were used to establish new relationships among biosensor performance characteristics (e.g., measurement confidence) and design parameters (e.g., inputs and hyperparameters of classification models and data acquisition parameters) that may be used for characterizing biosensor performance.

A novel disassembly process of end-of-life lithium-ion batteries enhanced by online sensing and machine learning techniques.

An effective lithium-ion battery (LIB) recycling infrastructure is of great importance to alleviate the concerns over the disposal of waste LIBs and the sustainability of critical elements for producing LIB components. The End-of-life (EOL) LIBs are in various sizes and shapes, which create significant challenges to automate a few unit operations (e.g., disassembly at the cell level) of the recycling process. Meanwhile, hazardous and flammable materials are contained in LIBs, posing great threats to the human exposure. Therefore, it is difficult to dismantle the LIBs safely and efficiently to recover critical materials. Automation has become a competitive solution in manufacturing world, which allows for mass production at outstanding speeds and with great repeatability or quality. It is imperative to develop automatic disassembly solution to effectively disassemble the LIBs while safeguarding human workers against the hazards environment. In this work, we demonstrate an automatic battery disassembly platform enhanced by online sensing and machine learning technologies. The computer vision is used to classify different types of batteries based on their brands and sizes. The real-time temperature data is captured from a thermal camera. A data-driven model is built to predict the cutting temperature pattern and the temperature spike can be mitigated by the close-loop control system. Furthermore, quality control is conducted using a neural network model to detect and mitigate the cutting defects. The integrated disassembly platform can realize the real-time diagnosis and closed-loop control of the cutting process to optimize the cutting quality and improve the safety. Supplementary Information: The online version contains supplementary material available at 10.1007/s10845-022-01936-x.

In situ melt pool measurements for laser powder bed fusion using multi sensing and correlation analysis.

Laser powder bed fusion is a promising technology for local deposition and microstructure control, but it suffers from defects such as delamination and porosity due to the lack of understanding of melt pool dynamics. To study the fundamental behavior of the melt pool, both geometric and thermal sensing with high spatial and temporal resolutions are necessary. This work applies and integrates three advanced sensing technologies: synchrotron X-ray imaging, high-speed IR camera, and high-spatial-resolution IR camera to characterize the evolution of the melt pool shape, keyhole, vapor plume, and thermal evolution in Ti-6Al-4V and 410 stainless steel spot melt cases. Aside from presenting the sensing capability, this paper develops an effective algorithm for high-speed X-ray imaging data to identify melt pool geometries accurately. Preprocessing methods are also implemented for the IR data to estimate the emissivity value and extrapolate the saturated pixels. Quantifications on boundary velocities, melt pool dimensions, thermal gradients, and cooling rates are performed, enabling future comprehensive melt pool dynamics and microstructure analysis. The study discovers a strong correlation between the thermal and X-ray data, demonstrating the feasibility of using relatively cheap IR cameras to predict features that currently can only be captured using costly synchrotron X-ray imaging. Such correlation can be used for future thermal-based melt pool control and model validation.

Using a smart textile system for classifying occupational manual material handling tasks: evidence from lab-based simulations.

Physical monitoring systems represent potentially powerful assessment devices to detect and describe occupational physical activities. A promising technology for such use is smart textile systems (STSs). Our goal in this exploratory study was to assess the feasibility and accuracy of using two STSs to classify several manual material handling (MMH) tasks. Specifically, commercially-available 'smart' socks and a custom 'smart' shirt were used individually and in combination. Eleven participants simulated nine separate MMH tasks while wearing the STSs, and task classification accuracy was quantified subsequently using several common models. The shirt and socks, both individually and in combination, could classify the simulated tasks with greater than 97% accuracy. Thus, using STSs appears to have potential utility for discriminating occupational physical tasks in the work environment. Practitioner summary: A smart textile system could classify diverse MMH tasks with high accuracy. This technology may help in developing future ergonomic exposure assessment systems, with the goal of preventing occupational injuries.

A machine learning-based approach to prognostic analysis of thoracic transplantations.

OBJECTIVE: The prediction of survival time after organ transplantations and prognosis analysis of different risk groups of transplant patients are not only clinically important but also technically challenging. The current studies, which are mostly linear modeling-based statistical analyses, have focused on small sets of disparate predictive factors where many potentially important variables are neglected in their analyses. Data mining methods, such as machine learning-based approaches, are capable of providing an effective way of overcoming these limitations by utilizing sufficiently large data sets with many predictive factors to identify not only linear associations but also highly complex, non-linear relationships. Therefore, this study is aimed at exploring risk groups of thoracic recipients through machine learning-based methods. METHODS AND MATERIAL: A large, feature-rich, nation-wide thoracic transplantation dataset (obtained from the United Network for Organ Sharing-UNOS) is used to develop predictive models for the survival time estimation. The predictive factors that are most relevant to the survival time identified via, (1) conducting sensitivity analysis on models developed by the machine learning methods, (2) extraction of variables from the published literature, and (3) eliciting variables from the medical experts and other domain specific knowledge bases. A unified set of predictors is then used to develop a Cox regression model and the related prognosis indices. A comparison of clustering algorithm-based and conventional risk grouping techniques is conducted based on the outcome of the Cox regression model in order to identify optimal number of risk groups of thoracic recipients. Finally, the Kaplan-Meier survival analysis is performed to validate the discrimination among the identified various risk groups. RESULTS: The machine learning models performed very effectively in predicting the survival time: the support vector machine model with a radial basis Kernel function produced the best fit with an R(2) value of 0.879, the artificial neural network (multilayer perceptron-MLP-model) came the second with an R(2) value of 0.847, and the M5 algorithm-based regression tree model came last with an R(2) value of 0.785. Following the proposed method, a consolidated set of predictive variables are determined and used to build the Cox survival model. Using the prognosis indices revealed by the Cox survival model along with a k-means clustering algorithm, an optimal number of "three" risk groups is identified. The significance of differences among these risk groups are also validated using the Kaplan-Meier survival analysis. CONCLUSIONS: This study demonstrated that the integrated machine learning method to select the predictor variables is more effective in developing the Cox survival models than the traditional methods commonly found in the literature. The significant distinction among the risk groups of thoracic patients also validates the effectiveness of the methodology proposed herein. We anticipate that this study (and other AI based analytic studies like this one) will lead to more effective analyses of thoracic transplant procedures to better understand the prognosis of thoracic organ recipients. It would potentially lead to new medical and biological advances and more effective allocation policies in the field of organ transplantation.

Predicting the graft survival for heart-lung transplantation patients: an integrated data mining methodology.

BACKGROUND: Predicting the survival of heart-lung transplant patients has the potential to play a critical role in understanding and improving the matching procedure between the recipient and graft. Although voluminous data related to the transplantation procedures is being collected and stored, only a small subset of the predictive factors has been used in modeling heart-lung transplantation outcomes. The previous studies have mainly focused on applying statistical techniques to a small set of factors selected by the domain-experts in order to reveal the simple linear relationships between the factors and survival. The collection of methods known as 'data mining' offers significant advantages over conventional statistical techniques in dealing with the latter's limitations such as normality assumption of observations, independence of observations from each other, and linearity of the relationship between the observations and the output measure(s). There are statistical methods that overcome these limitations. Yet, they are computationally more expensive and do not provide fast and flexible solutions as do data mining techniques in large datasets. PURPOSE: The main objective of this study is to improve the prediction of outcomes following combined heart-lung transplantation by proposing an integrated data-mining methodology. METHODS: A large and feature-rich dataset (16,604 cases with 283 variables) is used to (1) develop machine learning based predictive models and (2) extract the most important predictive factors. Then, using three different variable selection methods, namely, (i) machine learning methods driven variables-using decision trees, neural networks, logistic regression, (ii) the literature review-based expert-defined variables, and (iii) common sense-based interaction variables, a consolidated set of factors is generated and used to develop Cox regression models for heart-lung graft survival. RESULTS: The predictive models' performance in terms of 10-fold cross-validation accuracy rates for two multi-imputed datasets ranged from 79% to 86% for neural networks, from 78% to 86% for logistic regression, and from 71% to 79% for decision trees. The results indicate that the proposed integrated data mining methodology using Cox hazard models better predicted the graft survival with different variables than the conventional approaches commonly used in the literature. This result is validated by the comparison of the corresponding Gains charts for our proposed methodology and the literature review based Cox results, and by the comparison of Akaike information criteria (AIC) values received from each. CONCLUSIONS: Data mining-based methodology proposed in this study reveals that there are undiscovered relationships (i.e. interactions of the existing variables) among the survival-related variables, which helps better predict the survival of the heart-lung transplants. It also brings a different set of variables into the scene to be evaluated by the domain-experts and be considered prior to the organ transplantation.