AI-Assisted Cotton Grading: Active and Semi-Supervised Learning to Reduce the Image-Labelling Burden.

The assessment of food and industrial crops during harvesting is important to determine the quality and downstream processing requirements, which in turn affect their market value. While machine learning models have been developed for this purpose, their deployment is hindered by the high cost of labelling the crop images to provide data for model training. This study examines the capabilities of semi-supervised and active learning to minimise effort when labelling cotton lint samples while maintaining high classification accuracy. Random forest classification models were developed using supervised learning, semi-supervised learning, and active learning to determine Egyptian cotton grade. Compared to supervised learning (80.20-82.66%) and semi-supervised learning (81.39-85.26%), active learning models were able to achieve higher accuracy (82.85-85.33%) with up to 46.4% reduction in the volume of labelled data required. The primary obstacle when using machine learning for Egyptian cotton grading is the time required for labelling cotton lint samples. However, by applying active learning, this study successfully decreased the time needed from 422.5 to 177.5 min. The findings of this study demonstrate that active learning is a promising approach for developing accurate and efficient machine learning models for grading food and industrial crops.

'Do they ever think about people like us?': The experiences of people with learning disabilities in England and Scotland during the COVID-19 pandemic.

People with learning disabilities in England and Scotland have experienced an increased risk of illness and death during the COVID-19 pandemic. Drawing on data of a longitudinal qualitative study with 71 disabled people and 31 disability organisations, this article examines the experiences of 24 people with learning disabilities in England and Scotland during the pandemic, reflecting on what rendered them vulnerable and placed them at risk. Qualitative interviews were conducted with participants and key informants at two timepoints; June-August 2020 and February-April 2021. Findings emerged across four key themes: failure to plan for the needs of people with learning disabilities; the suspension and removal of social care; the impact of the pandemic on people's everyday routines; and lack of vaccine prioritisation. The inequalities experienced by people with learning disabilities in this study are not particular to the pandemic. We explore the findings in the context of theoretical frameworks of vulnerability, including Fineman's conceptualisation of a 'vulnerability paradigm'. We conclude that the structured marginalisation of people with disabilities, entrenched by government action and inaction, have created and exacerbated their vulnerability. Structures, policies and action must change.

Epistemic sabotage: The production and disqualification of evidence in disability benefit claims.

Disability benefits function by demarcating categories of need (the administrative category of disability) and determine eligibility using assessments of functioning. In the UK, these assessments are the Work Capability Assessment and PIP assessment. Inherently technical and abstruse processes, these assessments have been opportune sites for welfare reform in recent years. Disability benefits have also been a central point of contention between disability studies and sociology. Sociology has traditionally favoured an 'incomes approach' and called for more adequate financial support from the state. Early figures in the disabled people's movement rejected this position, and aligned with an oppression paradigm, argued for a more radical economic and social inclusion. We contend that this divide, set out in the Fundamental Principles of Disability, remains relevant for researching welfare reform today. This article treats benefits assessments as epistemic practices-interactional processes wherein claimants, their personal health professionals and commercial assessment providers come together in the production of knowledge about disability. Data include 50 in-depth interviews with benefit claimants and a discourse analysis of official texts directed at claimants, personal health professionals and commercial assessment providers. We outline a phenomenon we term 'epistemic sabotage', whereby the knowledge claims of claimants and their health professionals are systemically disqualified.

Domain Adaptation for In-Line Allergen Classification of Agri-Food Powders Using Near-Infrared Spectroscopy.

The addition of incorrect agri-food powders to a production line due to human error is a large safety concern in food and drink manufacturing, owing to incorporation of allergens in the final product. This work combines near-infrared spectroscopy with machine-learning models for early detection of this problem. Specifically, domain adaptation is used to transfer models from spectra acquired under stationary conditions to moving samples, thereby minimizing the volume of labelled data required to collect on a production line. Two deep-learning domain-adaptation methodologies are used: domain-adversarial neural networks and semisupervised generative adversarial neural networks. Overall, accuracy of up to 96.0% was achieved using no labelled data from the target domain moving spectra, and up to 99.68% was achieved when incorporating a single labelled data instance for each material into model training. Using both domain-adaptation methodologies together achieved the highest prediction accuracies on average, as did combining measurements from two near-infrared spectroscopy sensors with different wavelength ranges. Ensemble methods were used to further increase model accuracy and provide quantification of model uncertainty, and a feature-permutation method was used for global interpretability of the models.

Mechanism of Visible Light-Mediated Alkene Aminoarylation with Arylsulfonylacetamides.

Alkene aminoarylation with arylsulfonylacetamides via a visible-light mediated radical Smiles-Truce rearrangement represents a convenient approach to the privileged arylethylamine pharmacaphore traditionally generated by circuitous, multi-step sequences. Herein, we report detailed synthetic, spectroscopic, kinetic, and computational studies designed to interrogate the proposed mechanism, including the key aryl transfer event. The data are consistent with a rate-limiting 1,4-aryl migration occurring either via a stepwise process involving a radical Meisenheimer-like intermediate or in a concerted fashion dependent on both arene electronics and alkene sterics. Our efforts to probe the mechanism have significantly expanded the substrate scope of the transformation with respect to the migrating aryl group and provide further credence to the synthetic potential of radical aryl migrations.

A review of ultrasonic sensing and machine learning methods to monitor industrial processes.

Supervised machine learning techniques are increasingly being combined with ultrasonic sensor measurements owing to their strong performance. These techniques also offer advantages over calibration procedures of more complex fitting, improved generalisation, reduced development time, ability for continuous retraining, and the correlation of sensor data to important process information. However, their implementation requires expertise to extract and select appropriate features from the sensor measurements as model inputs, select the type of machine learning algorithm to use, and find a suitable set of model hyperparameters. The aim of this article is to facilitate implementation of machine learning techniques in combination with ultrasonic measurements for in-line and on-line monitoring of industrial processes and other similar applications. The article first reviews the use of ultrasonic sensors for monitoring processes, before reviewing the combination of ultrasonic measurements and machine learning. We include literature from other sectors such as structural health monitoring. This review covers feature extraction, feature selection, algorithm choice, hyperparameter selection, data augmentation, domain adaptation, semi-supervised learning and machine learning interpretability. Finally, recommendations for applying machine learning to the reviewed processes are made.

The Best Services Trial (BeST?): a cluster randomised controlled trial comparing the clinical and cost-effectiveness of New Orleans Intervention Model with services as usual (SAU) for infants and young children entering care.

BACKGROUND: Abused and neglected children are at increased risk of health problems throughout life, but negative effects may be ameliorated by nurturing family care. It is not known whether it is better to place these children permanently with substitute (foster or adoptive) families or to attempt to reform their birth families. Previously, we conducted a feasibility randomised controlled trial (RCT) of the New Orleans Intervention Model (NIM) for children aged 0-60 months coming into foster care in Glasgow. NIM is delivered by a multidisciplinary health and social care team and offers families, whose child has been taken into foster care, a structured assessment of family relationships followed by a trial of treatment aiming to improve family functioning. A recommendation is then made for the child to return home or for adoption. In the feasibility RCT, families were willing to be randomised to NIM or optimised social work services as usual and equipoise was maintained. Here we present the protocol of a substantive RCT of NIM including a new London site. METHODS: The study is a multi-site, pragmatic, single-blind, parallel group, cluster randomised controlled superiority trial with an allocation ratio of 1:1. We plan to recruit approximately 390 families across the sites, including those recruited in our feasibility RCT. They will be randomly allocated to NIM or optimised services as usual and followed up to 2.5 years post-randomisation. The principal outcome measure will be child mental health, and secondary outcomes will be child quality of life, the time taken for the child to be placed in permanent care (rehabilitation home or adoption) and the quality of the relationship with the primary caregiver. DISCUSSION: The study is novel in that infant mental health professionals rarely have a role in judicial decisions about children's care placements, and RCTs are rare in the judicial context. The trial will allow us to determine whether NIM is clinically and cost-effective in the UK and findings may have important implications for the use of mental health assessment and treatment as part of the decision-making about children in the care system.

Disabled people in Britain and the impact of the COVID-19 pandemic.

This paper reports on in-depth qualitative interviews conducted with 69 disabled people in England and Scotland, and with 28 key informants from infrastructure organisations in the voluntary and statutory sectors, about the impact of COVID-19, and measures taken to control it. Participants were recruited through voluntary organisations. As with everyone, the Pandemic has had a huge impact: we discuss the dislocations it has caused in everyday life; the failures of social care; the use of new technologies; and participants' view on leadership and communication. We conclude with suggestions for urgent short term and medium term responses, so that the United Kingdom and other countries can respond better to this and other pandemics, and build a more inclusive world.

Can categorised values of maximal oxygen uptake discriminate patterns of exercise dysfunction in pectus excavatum: a prospective cohort study?

Cohort studies of patients with pectus excavatum have inadequately characterised exercise dysfunction experienced. Cardiopulmonary exercise test data were delineated by maximal oxygen uptake values >80%, which was tested to examine whether patterns of exercise physiology were distinguished. METHODS: Seventy-two patients considered for surgical treatment underwent assessment of pulmonary function and exercise physiology with pulmonary function tests and cardiopulmonary exercise test between 2006 and 2019. Seventy who achieved a threshold respiratory gas exchange ratio of >1.1 were delineated by maximal oxygen uptake >80%, (group A, n=33) and <80% (group B, n=37) and comparison of constituent physiological parameters performed. RESULTS: The cohort was 20.8 (±SD 6.6) years of age, 60 men, with a Haller's Index of 4.1 (±SD 1.4). Groups A and B exhibited similar demography, pulmonary function test results and Haller's index values. Exercise test parameters of group B were lower than group A; work 79.2% (±SD 11.3) versus 97.7 (±SD 10.1), anaerobic threshold 38.1% (±SD 7.8) versus 49.7% (±SD 9.1) and O2 pulse 77.4% (±SD 9.8) versus 101.8% (±SD 11.7), but breathing reserve was higher, 54.9% (±SD 13.1) versus 44.2% (±SD 10.8), p<0.001 for each. Both groups exhibited similar incidences of carbon dioxide retention at peak exercise. A total of 65 (93%) exhibited abnormal values of at least one of four exercise test measures. CONCLUSION: This study showed that patients with pectus excavatum exhibited multiple physiological characteristics of compromised exercise function. It is the first study that defines differing patterns of exercise dysfunction and provides evidence that patients with symptomatic pectus excavatum should be considered for surgical treatment.

Transfer learning for process monitoring using reflection-mode ultrasonic sensing.

The fourth industrial revolution is set to integrate entire manufacturing processes using industrial digital technologies such as the Internet of Things, Cloud Computing, and machine learning to improve process productivity, efficiency, and sustainability. Sensors collect the real-time data required to optimise manufacturing processes and are therefore a key technology in this transformation. Ultrasonic sensors have benefits of being low-cost, in-line, non-invasive, and able to operate in opaque systems. Supervised machine learning models can correlate ultrasonic sensor data to useful information about the manufacturing materials and processes. However, this requires a reference measurement of the process material to label each data point for model training. Labelled data is often difficult to obtain in factory environments, and so a method of training models without this is desirable. This work compares two domain adaptation methods to transfer models across processes, so that no labelled data is required to accurately monitor a target process. The two method compared are a Single Feature transfer learning approach and Transfer Component Analysis using three features. Ultrasonic waveforms are unique to the sensor used, attachment procedure, and contact pressure. Therefore, only a small number of transferable features are investigated. Two industrially relevant processes were used as case studies: mixing and cleaning of fouling in pipes. A reflection-mode ultrasonic sensing technique was used, which monitors the sound wave reflected from the interface between the vessel wall and process material. Overall, the Single Feature method produced the highest prediction accuracies: up to 96.0% and 98.4% to classify the completion of mixing and cleaning, respectively; and R2 values of up to 0.947 and 0.999 to predict the time remaining until completion. These results highlight the potential of combining ultrasonic measurements with transfer learning techniques to monitor industrial processes. Although, further work is required to study various effects such as changing sensor location between source and target domains.

Performance of the AusSCORE II and STS Score for Coronary Artery Bypass Grafting in a New Zealand Population.

BACKGROUND: The updated Australian System for Cardiac Operative Risk Evaluation (AusSCORE II) and the Society of Thoracic Surgeons (STS) Score are well-established tools in cardiac surgery for estimating operative mortality risk. No validation analysis of both risk models has been undertaken for a contemporary New Zealand population undergoing isolated coronary bypass surgery. We therefore aimed to assess the efficacy of these models in predicting mortality for New Zealand patients receiving isolated coronary artery bypass grafting (CABG). MATERIAL AND METHODS: A prospective database was maintained of patients undergoing isolated CABG at a major tertiary referral centre in New Zealand between September 2014 and September 2017. This database collected the patients' demographic, clinical, biochemical, operative and mortality data. The primary outcome measure was the correlation between the predicted AusSCORE II and STS Score mortality risks and the observed 30-day mortality events for all patients in the database using discrimination and calibration statistics. Discrimination and calibration were assessed using receiver operating characteristic (ROC) curves and the Hosmer-Lemeshow test respectively. RESULTS: A total of 933 patients underwent isolated CABG during the 3-year study period. There were seven deaths in the study cohort occurring within 30 days of surgery. Discrimination analysis demonstrated the area under the ROC curve (AUC) of the AusSCORE II and STS Score as 88.2% (95% CI: 85.9-90.2, p<0.0001) and 92.1% (95% CI: 90.2-93.7, p<0.0001) respectively. Calibration analysis revealed Hosmer-Lemeshow test p-values for the AusSCORE II and STS Score as 0.696 and 0.294 respectively. DISCUSSION: ROC curve analysis produced very high and statistically significant AUC values for the AusSCORE II and STS Score. Hosmer-Lemeshow test analysis revealed that both risk scoring tools are well calibrated for our study cohort. Therefore, the AusSCORE II and STS Score are both strongly predictive of 30-day mortality for isolated coronary artery bypass grafting surgery in our New Zealand patient population. Both risk models have performed with excellent discrimination and calibration. There is, however, a need to consider the performance of these risk stratification models in other cardiac surgical procedures outside isolated coronary bypass surgery where appropriate.

Hypomorphic mTOR Downregulates CDK6 and Delays Thymic Pre-T LBL Tumorigenesis.

PI3K/AKT/mTOR pathway hyperactivation is frequent in T-cell acute lymphoblastic leukemia/lymphoma (T-ALL/LBL). To model inhibition of mTOR, pre-T-cell lymphoblastic leukemia/lymphoma (pre-T LBL) tumor development was monitored in mice with T lymphocyte-specific, constitutively active AKT (Lck-MyrAkt2) that were either crossed to mTOR knockdown (KD) mice or treated with the mTOR inhibitor everolimus. Lck-MyrAkt2;mTOR KD mice lived significantly longer than Lck-MyrAkt2;mTOR wild-type (WT) mice, although both groups ultimately developed thymic pre-T LBL. An increase in survival was also observed when Lck-MyrAkt2;mTOR WT mice were treated for 8 weeks with everolimus. The transcriptional profiles of WT and KD thymic lymphomas were compared, and Ingenuity Pathway Upstream Regulator Analysis of differentially expressed genes in tumors from mTOR WT versus KD mice identified let-7 and miR-21 as potential regulatory genes. mTOR KD mice had higher levels of let-7a and miR-21 than mTOR WT mice, and rapamycin induced their expression in mTOR WT cells. CDK6 was one of the most downregulated targets of both let-7 and miR21 in mTOR KD tumors. CDK6 overexpression and decreased expression of let-7 in mTOR KD cells rescued a G1 arrest phenotype. Combined mTOR (rapamycin) and CDK4/6 (palbociclib) inhibition decreased tumor size and proliferation in tumor flank transplants, increased survival in an intravenous transplant model of disseminated leukemia compared with single agent treatment, and cooperatively decreased cell viability in human T-ALL/LBL cell lines. Thus, mTOR KD mice provide a model to explore drug combinations synergizing with mTOR inhibitors and can be used to identify downstream targets of inhibition.

Intelligent Industrial Cleaning: A Multi-Sensor Approach Utilising Machine Learning-Based Regression.

Effectively cleaning equipment is essential for the safe production of food but requires a significant amount of time and resources such as water, energy, and chemicals. To optimize the cleaning of food production equipment, there is the need for innovative technologies to monitor the removal of fouling from equipment surfaces. In this work, optical and ultrasonic sensors are used to monitor the fouling removal of food materials with different physicochemical properties from a benchtop rig. Tailored signal and image processing procedures are developed to monitor the cleaning process, and a neural network regression model is developed to predict the amount of fouling remaining on the surface. The results show that the three dissimilar food fouling materials investigated were removed from the test section via different cleaning mechanisms, and the neural network models were able to predict the area and volume of fouling present during cleaning with accuracies as high as 98% and 97%, respectively. This work demonstrates that sensors and machine learning methods can be effectively combined to monitor cleaning processes.

Monitoring Mixing Processes Using Ultrasonic Sensors and Machine Learning.

Mixing is one of the most common processes across food, chemical, and pharmaceutical manufacturing. Real-time, in-line sensors are required for monitoring, and subsequently optimising, essential processes such as mixing. Ultrasonic sensors are low-cost, real-time, in-line, and applicable to characterise opaque systems. In this study, a non-invasive, reflection-mode ultrasonic measurement technique was used to monitor two model mixing systems. The two systems studied were honey-water blending and flour-water batter mixing. Classification machine learning models were developed to predict if materials were mixed or not mixed. Regression machine learning models were developed to predict the time remaining until mixing completion. Artificial neural networks, support vector machines, long short-term memory neural networks, and convolutional neural networks were tested, along with different methods for engineering features from ultrasonic waveforms in both the time and frequency domain. Comparisons between using a single sensor and performing multisensor data fusion between two sensors were made. Classification accuracies of up to 96.3% for honey-water blending and 92.5% for flour-water batter mixing were achieved, along with R2 values for the regression models of up to 0.977 for honey-water blending and 0.968 for flour-water batter mixing. Each prediction task produced optimal performance with different algorithms and feature engineering methods, vindicating the extensive comparison between different machine learning approaches.

Planar Laser Activated Neuronal Scanning (PLANS) System for in vivo Flow Cytometry.

We present a light-sheet flow cytometer for screening of C. elegans. A machine learning approach is utilized to enable real-time analysis of protein aggregation models.

The Effect of Light Intensity, Sensor Height, and Spectral Pre-Processing Methods when using NIR Spectroscopy to Identify Different Allergen-Containing Powdered Foods.

: Food allergens present a significant health risk to the human population, so their presence must be monitored and controlled within food production environments. This is especially important for powdered food, which can contain nearly all known food allergens. Manufacturing is experiencing the fourth industrial revolution (Industry 4.0), which is the use of digital technologies, such as sensors, Internet of Things (IoT), artificial intelligence, and cloud computing, to improve the productivity, efficiency, and safety of manufacturing processes. This work studied the potential of small low-cost sensors and machine learning to identify different powdered foods which naturally contain allergens. The research utilised a near-infrared (NIR) sensor and measurements were performed on over 50 different powdered food materials. This work focussed on several measurement and data processing parameters, which must be determined when using these sensors. These included sensor light intensity, height between sensor and food sample, and the most suitable spectra pre-processing method. It was found that the K-nearest neighbour and linear discriminant analysis machine learning methods had the highest classification prediction accuracy for identifying samples containing allergens of all methods studied. The height between the sensor and the sample had a greater effect than the sensor light intensity and the classification models performed much better when the sensor was positioned closer to the sample with the highest light intensity. The spectra pre-processing methods, which had the largest positive impact on the classification prediction accuracy, were the standard normal variate (SNV) and multiplicative scattering correction (MSC) methods. It was found that with the optimal combination of sensor height, light intensity, and spectra pre-processing, a classification prediction accuracy of 100% could be achieved, making the technique suitable for use within production environments.

Enhanced Clean-In-Place Monitoring Using Ultraviolet Induced Fluorescence and Neural Networks.

Clean-in-place (CIP) processes are extensively used to clean industrial equipment without the need for disassembly. In food manufacturing, cleaning can account for up to 70% of water use and is also a heavy user of energy and chemicals. Due to a current lack of real-time in-process monitoring, the non-optimal control of the cleaning process parameters and durations result in excessive resource consumption and periods of non-productivity. In this paper, an optical monitoring system is designed and realized to assess the amount of fouling material remaining in process tanks, and to predict the required cleaning time. An experimental campaign of CIP tests was carried out utilizing white chocolate as fouling medium. During the experiments, an image acquisition system endowed with a digital camera and ultraviolet light source was employed to collect digital images from the process tank. Diverse image segmentation techniques were considered to develop an image processing procedure with the aim of assessing the area of surface fouling and the fouling volume throughout the cleaning process. An intelligent decision-making support system utilizing nonlinear autoregressive models with exogenous inputs (NARX) Neural Network was configured, trained and tested to predict the cleaning time based on the image processing results. Results are discussed in terms of prediction accuracy and a comparative study on computation time against different image resolutions is reported. The potential benefits of the system for resource and time efficiency in food manufacturing are highlighted.

Clout or doubt? Perspectives on an infant mental health service for young children placed in foster care due to abuse and neglect.

Despite knowledge about the profound effects of child abuse and neglect, we know little about how best to assess whether maltreated children should return home. The effectiveness of the New Orleans Intervention Model (NIM) is being tested in a randomized controlled trial where the comparison is social work 'services as usual.' The future trial results will tell us which approach produces the best outcomes for children; meanwhile qualitative process evaluation is generating intriguing findings about the perceived impact of NIM on decision-making about childrens' futures. Interviews and focus groups were conducted with social workers, foster carers, legal decision-makers and the NIM team (n=63). Data were analysed thematically. Findings suggest that NIM is seen as bringing greater influence ('clout') to decision-making due to its depth of focus, provision of treatment for the family, health professional input and perceived objectivity. Simultaneously, the NIM approach and the detailed information it produces potentially throws judgments into doubt in the legal system. Clout/doubt perceptions permeate opinions about NIM and are inter-related with a historical discourse about 'health versus social' models of information gathering, with implications for assessment of child abuse and neglect that extend beyond the study context. The juxtaposition of 'clout versus doubt' both highlights and is strengthened by an intense focus among social workers and legal professionals on how evidence will be regarded within legal fora when making decisions about children. There is continuing uncertainty in the child welfare system about the best ways of assessing maltreated children, underscoring a continued need for the trial.