Large Amount Of Data Research Articles

Recognition of target domain Japanese speech using language model replacement

End-to-end (E2E) automatic speech recognition (ASR) models, which consist of deep learning models, are able to perform ASR tasks using a single neural network. These models should be trained using a large amount of data; however, collecting speech data which matches the targeted speech domain can be difficult, so speech data is often used that is not an exact match to the target domain, resulting in lower performance. In comparison to speech data, in-domain text data is much easier to obtain. Thus, traditional ASR systems use separately trained language models and HMM-based acoustic models. However, it is difficult to separate language information from an E2E ASR model because the model learns both acoustic and language information in an integrated manner, making it very difficult to create E2E ASR models for specialized target domain which are able to achieve sufficient recognition performance at a reasonable cost. In this paper, we propose a method of replacing the language information within pre-trained E2E ASR models in order to achieve adaptation to a target domain. This is achieved by deleting the “implicit” language information contained within the ASR model by subtracting the source-domain language model trained with a transcription of the ASR’s training data in a logarithmic domain. We then integrate a target domain language model through addition in the logarithmic domain. This subtraction and addition to replace of the language model is based on Bayes’ theorem. In our experiment, we first used two datasets of the Corpus of Spontaneous Japanese (CSJ) to evaluate the effectiveness of our method. We then we evaluated our method using the Japanese Newspaper Article Speech (JNAS) and CSJ corpora, which contain audio data from the read speech and spontaneous speech domain, respectively, to test the effectiveness of our proposed method at bridging the gap between these two language domains. Our results show that our proposed language model replacement method achieved better ASR performance than both non-adapted (baseline) ASR models and ASR models adapted using the conventional Shallow Fusion method.

Microservice-Based Vehicular Network for Seamless and Ultra-Reliable Communications of Connected Vehicles

The fifth-generation (5G) cellular infrastructure is expected to bring about the widespread use of connected vehicles. This technological progress marks the beginning of a new era in vehicular networks, which includes a range of different types and services of self-driving cars and the smooth sharing of information between vehicles. Connected vehicles have also been announced as a main use case of the sixth-generation (6G) cellular, with ultimate requirements beyond the 5G (B5G) and 6G eras. These networks require full coverage, extremely high reliability and availability, very low latency, and significant system adaptability. The significant specifications set for vehicular networks pose considerable design and development challenges. The goals of establishing a latency of 1 millisecond, effectively handling large amounts of data traffic, and facilitating high-speed mobility are of utmost importance. To address these difficulties and meet the demands of upcoming networks, e.g., 6G, it is necessary to improve the performance of vehicle networks by incorporating innovative technology into existing network structures. This work presents significant enhancements to vehicular networks to fulfill the demanding specifications by utilizing state-of-the-art technologies, including distributed edge computing, e.g., mobile edge computing (MEC) and fog computing, software-defined networking (SDN), and microservice. The work provides a novel vehicular network structure based on micro-services architecture that meets the requirements of 6G networks. The required offloading scheme is introduced, and a handover algorithm is presented to provide seamless communication over the network. Moreover, a migration scheme for migrating data between edge servers was developed. The work was evaluated in terms of latency, availability, and reliability. The results outperformed existing traditional approaches, demonstrating the potential of our approach to meet the demanding requirements of next-generation vehicular networks.

Explainable drug repurposing via path based knowledge graph completion

Drug repurposing aims to find new therapeutic applications for existing drugs in the pharmaceutical market, leading to significant savings in time and cost. The use of artificial intelligence and knowledge graphs to propose repurposing candidates facilitates the process, as large amounts of data can be processed. However, it is important to pay attention to the explainability needed to validate the predictions. We propose a general architecture to understand several explainable methods for graph completion based on knowledge graphs and design our own architecture for drug repurposing. We present XG4Repo (eXplainable Graphs for Repurposing), a framework that takes advantage of the connectivity of any biomedical knowledge graph to link compounds to the diseases they can treat. Our method allows methapaths of different types and lengths, which are automatically generated and optimised based on data. XG4Repo focuses on providing meaningful explanations to the predictions, which are based on paths from compounds to diseases. These paths include nodes such as genes, pathways, side effects, or anatomies, so they provide information about the targets and other characteristics of the biomedical mechanism that link compounds and diseases. Paths make predictions interpretable for experts who can validate them and use them in further research on drug repurposing. We also describe three use cases where we analyse new uses for Epirubicin, Paclitaxel, and Predinisone and present the paths that support the predictions.

Building Trust in AI Farming Tools

AbstractPrecision agriculture tools like decision support systems increasingly use machine‐learning algorithms and other types of artificial intelligence (AI) to analyze large quantities of agricultural data and provide recommendations to producers and crop advisers. However, several barriers threaten adoption of these tools. Three papers in the recent Agronomy Journal special section, “Machine Learning in Agriculture,” explore this phenomenon and offer solutions and opportunities for building trust in these technologies. Earn 0.5 CEUs in Crop Management by reading this article and taking the quiz at https://web.sciencesocieties.org/Learning‐Center/Courses.

Automatic quality assessment of knee radiographs using knowledge graphs and convolutional neural networks.

X-ray radiography is a widely used imaging technique worldwide, and its image quality directly affects diagnostic accuracy. Therefore, X-ray image quality control (QC) is essential. However, subjectively assessing image quality is inefficient and inconsistent, especially when large amounts of image data are being evaluated. Thus, subjective assessment cannot meet current QC needs. To meet current QC needs and improve the efficiency of image quality assessment, a complete set of quality assessment criteria must be established and implemented using artificial intelligence (AI) technology. Therefore, we proposed a multi-criteria AI system for automatically assessing the image quality of knee radiographs. A knee radiograph QC knowledge graph containing 16 "acquisition technique" labels representing 16 image quality defects and five "clarity" labels representing five grades of clarity were developed. Ten radiographic technologists conducted three rounds of QC based on this graph. The single-person QC results were denoted as QC1 and QC2, and the multi-person QC results were denoted as QC3. Each technologist labeled each image only once. The ResNet model structure was then used to simultaneously perform classification (detection of image quality defects) and regression (output of a clarity score) tasks to construct an image QC system. The QC3 results, comprising 4324 anteroposterior and lateral knee radiographs, were used for model training (70% of the images), validation (10%), and testing (20%). The 865 test set data were used to evaluate the effectiveness of the AI model, and an AI QC result, QC4, was automatically generated by the model after training. Finally, using a double-blind method, the senior QC expert reviewed the final QC results of the test set with reference to the results QC3 and QC4 and used them as a reference standard to evaluate the performance of the model. The precision and mean absolute error (MAE) were used to evaluate the quality of all the labels in relation to the reference standard. For the 16 "acquisition technique" features, QC4 exhibited the highest weighted average precision (98.42%±0.81%), followed by QC3 (91.39%±1.35%), QC2 (87.84%±1.68%), and QC1 (87.35%±1.71%). For the image clarity features, the MAEs between QC1, QC2, QC3, and QC4 and the reference standard were 0.508±0.021, 0.475±0.019, 0.237±0.016, and 0.303±0.018, respectively. The experimental results show that our automated quality assessment system performed well in classifying the acquisition technique used for knee radiographs. The image clarity quality evaluation accuracy of the model must be further improved but is generally close to that of radiographic technologists. Intelligent QC methods using knowledge graphs and convolutional neural networks have the potential for clinical applications.

Environmental performance and trends of the world's semiconductor foundry industry

AbstractThe semiconductor foundry industry faces the challenge of reducing its high environmental impact, mainly due to its energy‐ and water‐intensive processes and significant generation of waste. To date, no other study has focused on the assessment of the environmental performance and related historical trends of this industry as a whole. Methodologically, the first step was to analyze and process a large quantity of economic, production, and environmental data, available in the Corporate Social Responsibility reports of a companies’ sample, highly representative of the entire world's foundry industry (about 70% of the global revenue of the related sector). It was thus possible to calculate, using a common manufacturing index (MI) and after appropriate data processing, some key performance indicators, along a significant decade (2012–2021), marked by deep political, economic, and health crises. Some of the main findings of this study are that, over this 10‐year period, the increases in technological capacity (patents), wafer production, and revenue (400%, 183%, and 172%, respectively) are matched by a significant increase in hazardous waste generation per MI (20%; 239% in absolute value) and a much larger increase in general waste generation per MI (135%; 568% in absolute value). The indicators of energy, water, and revenue per MI are substantially unchanged. A substantial decrease occurs in GHG1&2 emissions per MI (−32%), mainly due to significant investments in renewable energy sources. The findings of this research could help and guide upcoming sustainability policy decisions and encourage business‐to‐business collaboration and the adoption of better environmental production practices.

Low Resource Malay Dialect Automatic Speech Recognition Modeling Using Transfer Learning from a Standard Malay Model

Approaches to automatic speech recognition have transited from Hidden Markov Model (HMM)-based ASR to deep neural networks. The advantages of deep neural network approaches are that they can be developed quickly and perform better given large language resources. Nevertheless, dialect speech recognition is still challenging due to the limited resources. Transfer learning approaches have been proposed to improve speech recognition for low resources. In the first approach, the model is pre-trained on a large and diverse labeled dataset to learn the acoustic and language patterns from the speech signal. Then, the model parameters are updated with a new dataset, and the pre-trained model is fine-tuned on a low-resource language dataset. The fine-tuning process is usually completed by freezing the pre-trained layers and training the remaining layers of the model on the low-resource language corpus. Another approach is to use a pre-trained model to capture the compact and meaningful features as input to the encoder. Pre-training in this approach usually involves using unsupervised learning methods to train models on a corpus of large amounts of unmarked data. It enables the model to learn the general patterns and relationships between the input speech signals. This paper proposes a training recipe using transfer learning and Standard Malay models to improve automatic speech recognition for Kelantan and Sarawak Malay dialects.

Old meets new: Innovative and evolving uses of herbaria over time as revealed by a literature review

Societal Impact StatementHerbaria, as collections of preserved plants, contain large amounts of data both in the labels and the specimens themselves, which can be applied in different study fields. A literature review was conducted to discover how the uses of herbaria have evolved over time since records began. This analysis revealed insights into how herbaria are presently used. Uses include traditional taxonomic review, as well as advanced technological tools, which are being applied to herbaria material to address societal and global challenges and therefore contribute to decision‐making in conservation.SummaryHerbaria as collections of preserved plants contain large amounts of data and prominent information, both on the labels and on the specimens themselves. There are 400 million specimens preserved in herbaria worldwide, with great potential for studies in conservation and effects of global change on plants. (1) In this paper, we investigate the array of herbaria uses through a systematic review of the scientific literature in SCOPUS covering the period 1842–2022. (2) We reviewed a total of 2900 papers and classified them in different areas of knowledge, as well as the taxonomic level studied. (3) Our results show that taxonomic use is the most relevant over time. This taxonomic use, together with the use as primary source of plant diversity data, is essential for documenting, planning, and acting on the conservation of threatened plants. Furthermore, our results have shown that new and diverse uses have emerged since 1990, mostly related to ecological issues, as herbaria provide a historical record of plant diversity and distribution, as well as their ecological and evolutionary responses, allowing scientists to track changes over time. (4) This contributes to improve the knowledge of biodiversity and to increase the effectiveness of conservation strategies and policies, which are a priority to address on going global change. Therefore, our study shows the relevance and potential of herbaria in ecology, including new or forthcoming uses, which are different from the uses originally intended by collectors. Thus, their preservation is critical as the unique and exceptional record for the knowledge of changes in biodiversity over space and time.

Coral spawning patterns on the Luhuitou fringing reef in Hainan Island of the northern South China Sea

Observation of coral spawning times is valuable to detect spawning patterns and identify the potential mechanisms behind coral reproduction. Although large amount of data on global coral spawning records have become available over the past decades, information on coral spawning in the northern South China Sea remains scarce. In this study, we investigated the spawning patterns of scleractinian corals in Luhuitou fringing reef, Hainan Island, China, from 2009 to 2021 in relation to lunar cycles (month and day). The spawning times of 22 coral species from five genera (Acropora, Montipora, Platygyra, Dipsastraea, and Galaxea) within three families (Acroporidae, Merulinidae, and Euphylliidae) were recorded, with spawning occurring from lunar February to May 2009–2021. Recorded spawning events started at a period of increasing seawater temperature. Acropora, the most documented genus, spawned between lunar February and April, but primarily in lunar March. Importantly, the spawning time of Acropora was delayed for one lunar month in 2016 most likely due to a rapid decrease in monthly mean seawater temperature in lunar February. Spawning lunar days before, on or after full moon of corals in the Luhuitou Reef, including those of the Acropora species, were highly variable between years even for the same species. No predictable pattern of spawning times can thus be detected. Nonetheless, our results fill up an information gap on coral spawning patterns in the northern South China Sea that may be useful for further understanding of the reproductive biology of scleractinian corals throughout the Indo–West Pacific.

Predicting pharmaceutical prices. Advances based on purchase-level data and machine learning

BackgroundIncreased costs in the health sector have put considerable strain on the public budgets allocated to pharmaceutical purchases. Faced with such pressures amplified by financial crises and pandemics, national purchasing authorities are presented with a puzzle: how to procure pharmaceuticals of the highest quality for the lowest price. The literature explored a range of impactful factors using data on producer and reference prices, but largely foregone the use of data on individual purchases by diverse public buyers.MethodsLeveraging the availability of open data in public procurement from official government portals, the article examines the relationship between unit prices and a host of predictors that account for policies that can be amended nationally or locally. The study uses traditional linear regression (OLS) and a machine learning model, random forest, to identify the best models for predicting pharmaceutical unit prices. To explore the association between a wide variety of predictors and unit prices, the study relies on more than 200,000 purchases in more than 800 standardized pharmaceutical product categories from 10 countries and territories.ResultsThe results show significant price variation of standardized products between and within countries. Although both models present substantial potential for predicting unit prices, the random forest model, which can incorporate non-linear relationships, leads to higher explained variance (R2 = 0.85) and lower prediction error (RMSE = 0.81).ConclusionsThe results demonstrate the potential of i) tapping into large quantities of purchase-level data in the health care sector and ii) using machine learning models for explaining and predicting pharmaceutical prices. The explanatory models identify data-driven policy interventions for decision-makers seeking to improve value for money.

Natural language processing of social network data for the evaluation of agricultural and rural policies

Global sustainable development challenges affect the agricultural sector, and many innovations aimed at addressing these challenges have been introduced in the agri-food sector. In this complex context, new agricultural policies are being implemented in Europe. Their success depends on their potential to adapt to new realities, responding to the opinions and demands of the European population. Given the rapid rise of social media as an important part of people's daily lives, public administrations have introduced digitalization and communication strategies through social media sites. Social media can provide policymakers with large amounts of data on user opinions. Given the value of social media as a rich source of data on public views and opinions, the aim of this paper is twofold: (i) to use natural language processing (NLP) to identify the events that have led to negative or positive opinion about European Common Agricultural Policy (CAP) reform and (ii) to evaluate the ability of NLP to study users' opinions on Twitter/X. The findings show that issues such as Brexit, the European Green Deal, the role of CAP in the environment, livestock farming, food safety, and illegal practices and corruption in the distribution of CAP funds have crucial implications for the design and application of the new CAP. Moreover, the study also suggests that NLP techniques can provide opportunities to integrate agricultural policies and instruments in the agri-food sector by assessing society's opinions. Sentiment analysis, even considering its limitations, could support sound and inclusive policymaking approaches anticipating public opinion in cases of risk of social unrest.

Mobile VOC measurements in Commerce City, CO reveal the emissions from different sources

ABSTRACT Source attribution of volatile organic compounds (VOCs) can be challenging in urban areas, which have many point sources. Mobile laboratories using time-of-flight mass spectrometers (TOF-MS) can take measurements throughout areas of concern, resulting in data with high spatial resolution that can be used to more easily identify these sources. However, emissions in heavily polluted areas still undergo significant mixing over short distances, making source attribution of some compounds challenging. Positive matrix factorization (PMF) has been widely used for attributing pollutants to different sources when taking stationary measurements due to its ability to process large amounts of data into generally interpretable results. However, some limitations of PMF can impact its usefulness to mobile data; PMF is a computationally intensive process, requires some user choices in attributing factors to emissions sources, and results can be significantly impacted by chemical transformations after emission. Here, both PMF and a simpler comparative analysis method are evaluated in analyzing measurements taken in the Elyria Swansea neighborhood of Commerce City, CO. This neighborhood is located near an oil refinery, a wastewater treatment plant, local industrial shops, and major highways. PMF failed to differentiate between oil refinery emissions and traffic emissions, and had difficulties recognizing other key sources. A simpler comparative analysis showed that the refinery contributed significantly to VOC concentrations throughout the neighborhood, including air toxics such as benzene. A wastewater treatment plant contributed to methanethiol and dimethyl sulfide. Finally, a small woodshop was identified as a hyperlocal VOC source, and contributed high amounts of some VOCs, such as toluene and other solvents, in its immediate surroundings. Implications Statement: This work discusses mobile measurements of VOCs around Commerce City, CO, a heavily polluted urban area north of Denver, using a PTR-TOF-MS. Two different source attribution methods, positive matrix factorization (PMF) and comparative analysis, were evaluated in the context of mobile measurements. The results show that an oil refinery and a woodshop contributed greatly to many VOC concentrations in the Elyria Swansea residential area of Commerce City. Additional sources, such as a wastewater treatment plant, also contributed to some odorous VOCs. PMF was unable to fully describe sources based on the mobile data. Comparative analysis was useful in attributing more VOCs to different sources, but quantitative results were influenced by how the analysis is set up. These findings are relevant to the residents of Denver and regulatory bodies to better understand Denver air pollution, as well as to other mobile studies doing source attribution of VOCs.

Uncertainty quantification for probabilistic machine learning in earth observation using conformal prediction

Machine learning is increasingly applied to Earth Observation (EO) data to obtain datasets that contribute towards international accords. However, these datasets contain inherent uncertainty that needs to be quantified reliably to avoid negative consequences. In response to the increased need to report uncertainty, we bring attention to the promise of conformal prediction within the domain of EO. Unlike previous uncertainty quantification methods, conformal prediction offers statistically valid prediction regions while concurrently supporting any machine learning model and data distribution. To support the need for conformal prediction, we reviewed EO datasets and found that only 22.5% of the datasets incorporated a degree of uncertainty information, with unreliable methods prevalent. Current open implementations require moving large amounts of EO data to the algorithms. We introduced Google Earth Engine native modules that bring conformal prediction to the data and compute, facilitating the integration of uncertainty quantification into existing traditional and deep learning modelling workflows. To demonstrate the versatility and scalability of these tools we apply them to valued EO applications spanning local to global extents, regression, and classification tasks. Subsequently, we discuss the opportunities arising from the use of conformal prediction in EO. We anticipate that accessible and easy-to-use tools, such as those provided here, will drive wider adoption of rigorous uncertainty quantification in EO, thereby enhancing the reliability of downstream uses such as operational monitoring and decision-making.

Novel mining conveyor monitoring system based on quasi-distributed optical fiber accelerometer array and self-supervised learning

Belt conveyors in mining are crucial, with downtime leading to significant losses and safety hazards. Unplanned shutdowns often result from idler failures. To address this, an online monitoring system for continuous idler health assessment is proposed. Considering the large number and dense spatial distribution of idlers over long distances, this work presents a system that utilizes a quasi-distributed optical fiber accelerometer array. This array incorporates phase-sensitive optical time domain reflectometry (Phase-OTDR) interrogation technology and ultra-weak fiber Bragg gratings (UWFBGs) to effectively capture idler vibrations. The designed array achieves high-sensitivity vibration sensing with a sensitivity of 2.4 rad/g and a resolution of 1.7 mg/Hz. After collecting the vibrations of idlers by the designed accelerometer array, an automatic fault classification algorithm based on self-supervised learning (SSL) is introduced, which requires only a small number of labeled samples. By leveraging large amount of unlabeled data in the pretext task, the algorithm efficiently extracts latent features from the quasi-distributed accelerometer array. A diagnosis accuracy of 95.37 % can be achieved on a seven-class classification task with only 3.6 % labeled data (16 samples/class). This system offers a promising solution for idler monitoring, combining high sensitivity, distributed measurement capabilities, enhanced security, and superior fault detection accuracy.

Multi-dimensional early warning of the entire supply chain of power materials based on RFID technology

Early warning system needs to process a large amount of real-time data, and carry out in-depth analysis and mining of these data to identify potential risks and hidden dangers. However, existing data processing and analysis capabilities may not be able to meet this demand. To this end, a multi-dimensional early warning of the entire supply chain of power materials based on RFID technology is designed. According to the real-time failure probability of power materials, the probability of early warning accidents is calculated and classified, and the risk is graded based on these probabilities. Through the methods of bonus points, deduction points and grade evaluation, the risk early warning indicators of different stages in the whole supply chain of power materials were quantified. The objective and rationality of risk assessment can be ensured by means of comprehensive weight. A multi-dimensional early warning system based on RFID technology is established, combining multiple linear regression model and particle swarm optimization algorithm to determine the time window of multi-dimensional early warning, and carry out dynamic monitoring and early warning of supply chain. The experimental results show that the early warning effect of the design method can reach 95% and the highest early warning effect can reach 98% at 10 seconds. The average warning error is only 2.91%, and the average warning time is only 1.34 seconds, which is more accurate in identifying the number of first-level risks, second-level risks and third-level risks.

A review on emotion detection by using deep learning techniques

Along with the growth of Internet with its numerous potential applications and diverse fields, artificial intelligence (AI) and sentiment analysis (SA) have become significant and popular research areas. Additionally, it was a key technology that contributed to the Fourth Industrial Revolution (IR 4.0). The subset of AI known as emotion recognition systems facilitates communication between IR 4.0 and IR 5.0. Nowadays users of social media, digital marketing, and e-commerce sites are increasing day by day resulting in massive amounts of unstructured data. Medical, marketing, public safety, education, human resources, business, and other industries also use the emotion recognition system widely. Hence it provides a large amount of textual data to extract the emotions from them. The paper presents a systematic literature review of the existing literature published between 2013 to 2023 in text-based emotion detection. This review scrupulously summarized 330 research papers from different conferences, journals, workshops, and dissertations. This paper explores different approaches, methods, different deep learning models, key aspects, description of datasets, evaluation techniques, Future prospects of deep learning, challenges in existing studies and presents limitations and practical implications.

The seasonal model of chili price movement with the effect of long memory and exogenous variables for improving time series model accuracy

This study aimed to explore big-time series data on agricultural commodities with an autocorrelation model comprising long-term processes, seasonality, and the impact of exogenous variables. Among the agricultural commodities with a large amount of data, chili prices exemplified criteria for long-term memory, seasonality, and the impact of various factors on production as an exogenous variable. These factors included the month preceding the new year and the week before the Eid al-Fitr celebration in Indonesia. To address the factors affecting price fluctuations, the Seasonal Autoregressive Fractionally Integrated Moving Average (SARFIMA) model was used to manage seasonality and long-term memory effects in the big data analysis. It improved with the addition of exogenous variables called SARFIMAX (SARFIMA with exogenous variables is known as SARFIMAX). After comparing the accuracy of both models, it was discovered that the SARFIMAX performed better, indicating the influence of seasonality and previous chili prices for an extended period in conjunction with exogenous variables. The SARFIMAX model gives an improvement in model accuracy by adding the effect of exogenous variables. Consequently, this observation concerning price dynamics established the cornerstone for maintaining the sustainability of chili supply even with the big data case.

Development of a multi-scanner facility for data acquisition for digital pathology artificial intelligence.

Whole slide imaging (WSI) of pathology glass slides using high-resolution scanners has enabled the large-scale application of artificial intelligence (AI) in pathology, to support the detection and diagnosis of disease, potentially increasing efficiency and accuracy in tissue diagnosis. Despite the promise of AI, it has limitations. 'Brittleness' or sensitivity to variation in inputs necessitates that large amounts of data are used for training. AI is often trained on data from different scanners but not usually by replicating the same slide across scanners. The utilisation of multiple WSI instruments to produce digital replicas of the same slides will make more comprehensive datasets and may improve the robustness and generalisability of AI algorithms as well as reduce the overall data requirements of AI training. To this end, the National Pathology Imaging Cooperative (NPIC) has built the AI FORGE (Facilitating Opportunities for Robust Generalisable data Emulation), a unique multi-scanner facility embedded in a clinical site in the NHS to (1) compare scanner performance, (2) replicate digital pathology image datasets across WSI systems, and (3) support the evaluation of clinical AI algorithms. The NPIC AI FORGE currently comprises 15 scanners from nine manufacturers. It can generate approximately 4,000 WSI images per day (approximately 7 TB of image data). This paper describes the process followed to plan and build such a facility. © 2024 The Author(s). The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.

Development of a multi-scanner facility for data acquisition for digital pathology artificial intelligence.

Whole slide imaging (WSI) of pathology glass slides using high-resolution scanners has enabled the large-scale application of artificial intelligence (AI) in pathology, to support the detection and diagnosis of disease, potentially increasing efficiency and accuracy in tissue diagnosis. Despite the promise of AI, it has limitations. 'Brittleness' or sensitivity to variation in inputs necessitates that large amounts of data are used for training. AI is often trained on data from different scanners but not usually by replicating the same slide across scanners. The utilisation of multiple WSI instruments to produce digital replicas of the same slides will make more comprehensive datasets and may improve the robustness and generalisability of AI algorithms as well as reduce the overall data requirements of AI training. To this end, the National Pathology Imaging Cooperative (NPIC) has built the AI FORGE (Facilitating Opportunities for Robust Generalisable data Emulation), a unique multi-scanner facility embedded in a clinical site in the NHS to (1) compare scanner performance, (2) replicate digital pathology image datasets across WSI systems, and (3) support the evaluation of clinical AI algorithms. The NPIC AI FORGE currently comprises 15 scanners from nine manufacturers. It can generate approximately 4,000 WSI images per day (approximately 7 TB of image data). This paper describes the process followed to plan and build such a facility. © 2024 The Author(s). The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.

Research on virtual entertainment robots based on machine learning algorithms providing psychological health services for college students

In modern society, college students are facing increasing psychological pressure and mental health problems. In this context, virtual entertainment robots have become a promising form of mental health services, which can utilize machine learning algorithms to provide personalized psychological support and guidance by analyzing a large amount of psychological data and user information. Study the use of sample calculation and screening methods to determine the number of samples and perform feature selection to improve algorithm performance. Then analyze the detection effect and evaluate the effectiveness of the algorithm. By designing the architecture of a virtual entertainment robot and adopting anti-interference strategies to ensure that the robot can accurately recognize mental health information, text recognition technology was implemented, its effectiveness was evaluated, and further multi-source information recognition was carried out to improve recognition accuracy. Finally, a psychological health evaluation system for college students was constructed, and corresponding psychological health service strategies were proposed to meet the needs of college students. The results of this study indicate that virtual entertainment robots based on machine learning algorithms can effectively provide mental health services, providing support and guidance for the mental health problems of college students.