Model Performed Research Articles

Visual field prediction based on temporal-spatial feature learning

Glaucoma stands as the leading irreversible cause of blindness worldwide. Regular visual field examinations play a crucial role in both diagnosing and treating glaucoma. Predicting future visual field changes can assist clinicians in making timely interventions to manage the progression of this disease. To integrate temporal and spatial features from past visual field examination results and enhance visual field prediction, a convolutional long short-term memory (ConvLSTM) network was employed to construct a predictive model. The predictive performance of the ConvLSTM model was validated and compared with other methods using a dataset of perimetry tests from the Humphrey field analyzer at the University of Washington (UWHVF). Compared to traditional methods, the ConvLSTM model demonstrated higher prediction accuracy. Additionally, the relationship between visual field series length and prediction performance was investigated. In predicting the visual field using the previous three visual field results of past 1.5~6.0 years, it was found that the ConvLSTM model performed better, achieving a mean absolute error of 2.255 dB, a root mean squared error of 3.457 dB, and a coefficient of determination of 0.960. The experimental results show that the proposed method effectively utilizes existing visual field examination results to achieve more accurate visual field prediction for the next 0.5~2.0 years. This approach holds promise in assisting clinicians in diagnosing and treating visual field progression in glaucoma patients.

Conceptualizing bias in EHR data: A case study in performance disparities by demographic subgroups for a pediatric obesity incidence classifier.

Electronic Health Records (EHRs) are increasingly used to develop machine learning models in predictive medicine. There has been limited research on utilizing machine learning methods to predict childhood obesity and related disparities in classifier performance among vulnerable patient subpopulations. In this work, classification models are developed to recognize pediatric obesity using temporal condition patterns obtained from patient EHR data in a U.S. study population. We trained four machine learning algorithms (Logistic Regression, Random Forest, Gradient Boosted Trees, and Neural Networks) to classify cases and controls as obesity positive or negative, and optimized hyperparameter settings through a bootstrapping methodology. To assess the classifiers for bias, we studied model performance by population subgroups then used permutation analysis to identify the most predictive features for each model and the demographic characteristics of patients with these features. Mean AUC-ROC values were consistent across classifiers, ranging from 0.72-0.80. Some evidence of bias was identified, although this was through the models performing better for minority subgroups (African Americans and patients enrolled in Medicaid). Permutation analysis revealed that patients from vulnerable population subgroups were over-represented among patients with the most predictive diagnostic patterns. We hypothesize that our models performed better on under-represented groups because the features more strongly associated with obesity were more commonly observed among minority patients. These findings highlight the complex ways that bias may arise in machine learning models and can be incorporated into future research to develop a thorough analytical approach to identify and mitigate bias that may arise from features and within EHR datasets when developing more equitable models.

Applying a Theory of Change for Human Resources Development in Public Health Supply Chains in Rwanda.

The health supply chain (SC) system in Rwanda experienced a number of workforce-related challenges, including insufficient skilled supply chain management (SCM) professionals with the necessary competencies. The Human Resources for Supply Chain Management (HR4SCM) Theory of Change (TOC) provides a methodology to assess human resources (HR) management systems by explaining the preconditions required to achieve optimized workforce performance. We applied this model to design interventions to strengthen the Rwanda health SC workforce. We compared conditions in the health SC HR system in Rwanda with the 60 outcomes described as necessary for optimized workforce performance in the HR4SCM TOC model. We used a survey and participatory workshop at the central level, followed by structured interviews (N=35) with SC professionals in health centers, hospitals, and regional warehouses (N=20) in Southern Province and Kigali City to identify which outcomes already existed in the Rwandan HR system and which outcomes required strengthening through targeted interventions. We used focus groups (N=2) to refine interventions. We identified that 31 of the 60 outcomes were not sufficiently in place in the Rwandan health SC HR system. SCM workers had gaps in the technical and managerial competencies and did not have access to adequate training and professional development opportunities for certain required competencies. An SCM career path did not exist, and education was not available for all required SCM qualifications. Fourteen of these outcomes were prioritized for strengthening. We designed 20 workforce interventions with the Ministry of Health to address these deficiencies and selected indicators to monitor the interventions. Applying this HR TOC model enabled a systematic process to identify gaps, develop and prioritize interventions, and select indicators. Practitioners designing and evaluating SC workforce interventions should consider applying this methodology to design more effective, theory-driven interventions to improve SC workforce performance.

Forecasting Orange Juice Futures: LSTM, ConvLSTM, and Traditional Models Across Trading Horizons

This study evaluated the forecasting accuracy of various models over 5-day and 10-day trading horizons to predict the prices of orange juice futures (OJ = F). The analysis included traditional models like Autoregressive Integrated Moving Average (ARIMA) and advanced neural network models such as Long Short-Term Memory (LSTM), Recurrent Neural Network (RNN), Backpropagation Neural Network (BPNN), Support Vector Regression (SVR), and Convolutional Long Short-Term Memory (ConvLSTM), incorporating factors like the Commodities Index and the S&P500 Index. We employed loss function metrics and various tests to assess model performance. The results indicated that for the 5-day horizon, the LSTM and ConvLSTM consistently outperformed the other models. LSTM achieved the lowest error rates and demonstrated superior capability in capturing temporal dependencies, especially in single-factor and S&P500 Index predictions. ConvLSTM also performed strongly, effectively modeling spatial and temporal data patterns. In the 10-day horizon, similar trends were observed. LSTM and ConvLSTM models had significantly lower errors and better alignment with actual values. The BPNN model performed well when all factors were included, and the SVR model maintained consistent accuracy, particularly for single-factor predictions. The Diebold–Mariano (DM) test indicated significant differences in forecasting accuracy, favoring advanced neural network models. In addition, incorporating multiple influencing factors further improved predictive performance, enhancing investment outcomes and reducing risk.

Performance Evaluation of CMIP6 Climate Model Projections for Precipitation and Temperature in the Upper Blue Nile Basin, Ethiopia

The projection and identification of historical and future changes in climatic systems is crucial. This study aims to assess the performance of CMIP6 climate models and projections of precipitation and temperature variables over the Upper Blue Nile Basin (UBNB), Northwestern Ethiopia. The bias in the CMIP6 model data was adjusted using data from meteorological stations. Additionally, this study uses daily CMIP6 precipitation and temperature data under SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios for the near (2015–2044), mid (2045–2074), and far (2075–2100) periods. Power transformation and distribution mapping bias correction techniques were used to adjust biases in precipitation and temperature data from seven CMIP6 models. To validate the model data against observed data, statistical evaluation techniques were employed. Mann–Kendall (MK) and Sen’s slope estimator were also performed to identify trends and magnitudes of variations in rainfall and temperature, respectively. The performance evaluation revealed that the INM-CM5-0 and INM-CM4-8 models performed best for precipitation and temperature, respectively. The precipitation projections in all agro-climatic zones under SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios show a significant (p < 0.01) positive trend. The mean annual maximum temperature over UBNB is estimated to increase by 1.8 °C, 2.1 °C, and 2.8 °C under SSP1-2.6, SSP2-4.5, and SSP5-8.5 between 2015 and 2100, respectively. Similarly, the mean annually minimum temperature is estimated to increase by 1.5 °C, 2.1 °C, and 3.1 °C under SSP1-2.6, SSP2-4.5, and SSP5-8.5, respectively. These significant changes in climate variables are anticipated to alter the incidence and severity of extremes. Hence, communities should adopt various adaptation practices to mitigate the effects of rising temperatures.

Spatial memory in Alzheimer's disease 5XFAD mice is enhanced by XPO1 inhibitor KPT-330.

The proteostatic decline in Alzheimer's disease is well established and improvement in proteostasis could potentially delay cognitive impairment. One emerging entry point to modulate proteostasis is the regulation of nucleo-cytoplasmic partitioning of proteins across the nuclear pore via karyopherins. The nuclear exportin XPO1 is a key regulator of proteostasis by driving the assembly of ribosomes and by modulating the process of autophagy. We recently found that XPO1 inhibitor KPT-330 (Selinexor), an FDA approved drug against multiple myelomas, enhances proteostasis, leading to benefits in models of neurodegenerative diseases in C. elegans and Drosophila . Here, we find that KPT-330 increases autophagy in murine neuronal cells and improves spatial memory performance in a murine model of Alzheimer's disease (5XFAD). Unexpectedly, general amyloid deposition in several brain regions was significantly increased by KPT-330, but specific regions, especially the thalamus, displayed significantly lower deposition, suggesting that XPO1 inhibition has regional-specific effects on proteostasis and amyloid plaque formation. Altogether, we conclude that XPO1 inhibition can improve cognition via spatially-specific reductions in amyloid deposition.

Evaluating Hydrologic Model Performance for Characterizing Streamflow Drought in the Conterminous United States

Hydrologic models are the primary tools that are used to simulate streamflow drought and assess impacts. However, there is little consensus about how to evaluate the performance of these models, especially as hydrologic modeling moves toward larger spatial domains. This paper presents a comprehensive multi-objective approach to systematically evaluating the critical features in streamflow drought simulations performed by two widely used hydrological models. The evaluation approach captures how well a model classifies observed periods of drought and non-drought, quantifies error components during periods of drought, and assesses the models’ simulations of drought severity, duration, and intensity. We apply this approach at 4662 U.S. Geological Survey streamflow gages covering a wide range of hydrologic conditions across the conterminous U.S. from 1985 to 2016 to evaluate streamflow drought using two national-scale hydrologic models: the National Water Model (NWM) and the National Hydrologic Model (NHM); therefore, a benchmark against which to evaluate additional models is provided. Using this approach, we find that generally the NWM better simulates the timing of flows during drought, while the NHM better simulates the magnitude of flows during drought. Both models performed better in wetter eastern regions than in drier western regions. Finally, each model showed increased error when simulating the most severe drought events.

Vis-NIR and NIR hyperspectral imaging combined with convolutional neural network with attention module for flaxseed varieties identification

The screening and identifying flax germplasm resources are critical for achieving precise flax breeding and variety enhancement. This study integrates hyperspectral imaging (HSI) technology with deep learning to identify flaxseed varieties. Hyperspectral images were captured for 15 flaxseed varieties across two spectral ranges: Vis-NIR (380–1018 nm) and NIR (870–1709 nm). PCA and LDA were utilized to visually cluster these varieties. To automatically learn the spectral features and improve model performance, a one-dimensional convolutional neural network (CAM-TM-1DCNN) embedded with a channel attention module (CAM) and transformer module (TM) was developed for rapid recognition of flaxseed varieties. Experimental results validate the model's efficacy. Compared with ELM, BPNN, LSTM and 1DCNN classification models, the CAM-TM-1DCNN demonstrated superior classification performance in the NIR spectral range, achieving a test accuracy of 95.26 %. Moreover, all models performed better in the NIR spectral range compared to the Vis-NIR spectral range. The study also evaluated the impact of SPA and CARS feature selection algorithms on the classification models, confirming that the full-spectrum-based CAM-TM-1DCNN model outperformed others. These findings suggest that the CAM-TM-1DCNN model can effectively identify flaxseed varieties, providing a novel strategy and viable technical approach for future flaxseed variety recognition based on HSI technology.

Feasibility of Transformer Model for User Authentication Using Electromyogram Signals

Transformer models are widely used in natural language processing (NLP) and time-series data analysis. Applications of these models include prediction systems and hand gesture recognition using electromyogram (EMG) signals. However, in the case of time-series analysis, the models perform similarly to traditional networks, contrary to expectations. This study aimed to compare the performance of the transformer model and its various modified versions in terms of accuracy through a user authentication system using EMG signals, which exhibit significant variability and pose challenges in feature extraction. A Siamese network was employed to distinguish subtle differences in the EMG signals between users, using Euclidean distance. Data from 100 individuals were used to create a challenging scenario while ensuring accuracy. Three scenarios were considered: data preprocessing, integration with existing models, and the modification of the internal structure of the transformer model. The method that achieved the highest accuracy was the bidirectional long short-term memory (BiLSTM)–transformer approach. Based on this, a network was further constructed and optimized, resulting in a user authentication accuracy of 99.7% using EMG data from 100 individuals.

Assessing Retrieval-Augmented Large Language Model Performance in Emergency Department ICD-10-CM Coding Compared to Human Coders.

Accurate medical coding is essential for clinical and administrative purposes but complicated, time-consuming, and biased. This study compares Retrieval-Augmented Generation (RAG)-enhanced LLMs to provider-assigned codes in producing ICD-10-CM codes from emergency department (ED) clinical records. Retrospective cohort study using 500 ED visits randomly selected from the Mount Sinai Health System between January and April 2024. The RAG system integrated past 1,038,066 ED visits data (2021-2023) into the LLMs' predictions to improve coding accuracy. Nine commercial and open-source LLMs were evaluated. The primary outcome was a head-to-head comparison of the ICD-10-CM codes generated by the RAG-enhanced LLMs and those assigned by the original providers. A panel of four physicians and two LLMs blindly reviewed the codes, comparing the RAG-enhanced LLM and provider-assigned codes on accuracy and specificity. RAG-enhanced LLMs demonstrated superior performance to provider coders in both the accuracy and specificity of code assignments. In a targeted evaluation of 200 cases where discrepancies existed between GPT-4 and provider-assigned codes, human reviewers favored GPT-4 for accuracy in 447 instances, compared to 277 instances where providers' codes were preferred (p<0.001). Similarly, GPT-4 was selected for its superior specificity in 509 cases, whereas human coders were preferred in only 181 cases (p<0.001). Smaller open-access models, such as Llama-3.1-70B, also demonstrated substantial scalability when enhanced with RAG, with 218 instances of accuracy preference compared to 90 for providers' codes. Furthermore, across all models, the exact match rate between LLM-generated and provider-assigned codes significantly improved following RAG integration, with Qwen-2-7B increasing from 0.8% to 17.6% and Gemma-2-9b-it improving from 7.2% to 26.4%. RAG-enhanced LLMs improve medical coding accuracy in EDs, suggesting clinical workflow applications. These findings show that generative AI can improve clinical outcomes and reduce administrative burdens. This work was supported in part through the computational and data resources and staff expertise provided by Scientific Computing and Data at the Icahn School of Medicine at Mount Sinai and supported by the Clinical and Translational Science Awards (CTSA) grant UL1TR004419 from the National Center for Advancing Translational Sciences. Research reported in this publication was also supported by the Office of Research Infrastructure of the National Institutes of Health under award number S10OD026880 and S10OD030463. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The funders played no role in study design, data collection, analysis and interpretation of data, or the writing of this manuscript. A study showed AI models with retrieval-augmented generation outperformed human doctors in ED diagnostic coding accuracy and specificity. Even smaller AI models perform favorably when using RAG. This suggests potential for reducing administrative burden in healthcare, improving coding efficiency, and enhancing clinical documentation.

Frugal engineering-inspired wearable augmented reality goggle system enables fluorescence-guided cancer surgery

Disparities in surgical outcomes often result from subjective decisions dictated by surgical training, experience, and available resources. To improve outcomes, surgeons have adopted advancements in robotics, endoscopy, and intra-operative imaging including fluorescence-guided surgery (FGS), which highlights tumors and anatomy in real-time. However, technical, economic, and logistic challenges hinder widespread adoption of FGS beyond high-resource centers. To overcome these impediments, we combined laser diodes, Raspberry Pi cameras and computers, off-the-shelf optical components, and 3D-printed parts to make a battery-powered, compact, dual white light and NIR imaging system that has comparable performance to existing bulkier, pricier, and wall-powered technologies. We combined these components with off-the-shelf augmented reality (AR) glasses to create a fully-wearable fluorescence imaging AR Raspberry Pi-based goggle system (FAR-Pi) and validated performance in a pre-clinical cancer surgery model. Novel device design ensures distance-independent coalignment between real and augmented views. As an open-source, affordable, and adaptable system, FAR-Pi is poised to democratize access to FGS and improve health outcomes worldwide.

Competing and Noncompeting Risk Models for Predicting Kidney Allograft Failure.

Prognostic models are becoming increasingly relevant in clinical trials as potential surrogate endpoints, and for patient management as clinical decision support tools. However, the impact of competing risks on model performance remains poorly investigated. We aimed to carefully assess the performance of competing risk and noncompeting risk models in the context of kidney transplantation, where allograft failure and death with a functioning graft are two competing outcomes. We included 11,046 kidney transplant recipients enrolled in 10 countries. We developed prediction models for long-term kidney graft failure prediction, without accounting (i.e., censoring) and accounting for the competing risk of death with a functioning graft, using Cox, Fine-Gray, and cause-specific Cox regression models. To this aim, we followed a detailed and transparent analytical framework for competing and noncompeting risk modelling, and carefully assessed the models' development, stability, discrimination, calibration, overall fit, clinical utility, and generalizability in external validation cohorts and subpopulations. More than 15 metrics were used to provide an exhaustive assessment of model performance. Among 11,046 recipients in the derivation and validation cohorts, 1,497 (14%) lost their graft and 1,003 (9%) died with a functioning graft after a median follow-up post-risk evaluation of 4.7 years (IQR 2.7-7.0). The cumulative incidence of graft loss was similarly estimated by Kaplan-Meier and Aalen-Johansen methods (17% versus 16% in the derivation cohort). Cox and competing risk models showed similar and stable risk estimates for predicting long-term graft failure (average mean absolute prediction error of 0.0140, 0.0138 and 0.0135 for Cox, Fine-Gray, and cause-specific Cox models, respectively). Discrimination and overall fit were comparable in the validation cohorts, with concordance index ranging from 0.76 to 0.87. Across various subpopulations and clinical scenarios, the models performed well and similarly, although in some high-risk groups (such as donors over 65 years old), the findings suggest a trend towards moderately improved calibration when using a competing risk approach. Competing and noncompeting risk models performed similarly in predicting long-term kidney graft failure.

Enhancing heart disease classification with M2MASC and CNN-BiLSTM integration for improved accuracy

Heart disease is a leading cause of death globally; therefore, accurate detection and classification are prominent, and several DL and ML methods have been developed over the last decade. However, the classical approaches may be prone to overfitting and under fitting issues, and the model performance may lag due to the unavailability of annotated datasets. To overcome these issues, the research proposed a model for heart disease detection and classification by integrating blockchain technology with a Modified mixed attention-enabled search optimizer-based CNN-Bidirectional Long Short-Term Memory (M2MASC enabled CNN-BiLSTM) model. The novel model incorporates a pre-trained VGG16 model to enhance feature extraction and improve the overall predictive accuracy. Leveraging the continuous monitoring capabilities of IoT devices, patient data is collected in real-time, providing a dynamic source to the CNN-BiLSTM model. Blockchain integration ensures stored health data’s security, transparency, and immutability, addresses privacy concerns, and promotes trust in the predictive system. The classifier parameters are tuned using the modified mixed attention and search optimization. The M2MASC-enabled CNN-BiLSTM model performs better than traditional methods of accuracy 98.25%, precision 99.57%, and recall 97.53% for TP 80 with the MIT-BIH dataset.

Evaluation of Groundwater Potential Zones Using GIS‐Based Machine Learning Ensemble Models in the Gidabo Watershed, Ethiopia

AbstractThe main objective of this study is to map and evaluate groundwater potential zones (GWPZs) using advanced ensemble machine learning (ML) models, notably Random Forest (RF) and Support Vector Machine (SVM). GWPZs are identified by considering essential factors such as geology, drainage density, slope, land use/land cover (LULC), rainfall, soil, and lineament density. This is combined with datasets used for training and validating the RF and SVM models, which consisted of 75 potential sites (boreholes and springs), 22 non‐potential sites (bare lands and settlement areas), and 20 potential sites (water bodies). Each dataset is randomly partitioned into two sets: training (70%) and validation (30%). The model's performance is evaluated using the area under the receiver operating characteristic curve (AUC‐ROC). The AUC of the RF model is 0.91, compared to 0.88 for the SVM model. Both models classified GWPZs effectively, but the RF model performed slightly better. The classified GWPZ map shows that high GWPZs are typically located within water bodies, natural springs, low‐lying regions, and forested areas. In contrast, low GWPZs are primarily found in shrubland and grassland areas. This study is vital for decision‐makers as it promotes sustainable groundwater use and ensures water security in the studied area.

A Novel Neural Network Prediction Model for Mineral Content Based on the Data of Elemental Logging and Well Logging

The mineral content of shale plays an important role in the study of reservoir characteristics. Obtaining high-precision mineral content data from a simple and straightforward method is helpful for the study of reservoir properties. Due to the complex and varied elemental combination of minerals and the lithological diversity in different formations, there are limitations in the traditional statistical methods for predicting mineral content. In this study, a multi-source fusion dataset is constructed first using a detailed workflow to process multi-source data from elemental logging and conventional well logging from the Funing shale formations in China. A hybrid neural network model consisting of a back-propagation artificial neural network (BP-ANN) and gated recurrent unit (GRU) structure, with a custom constraint loss function, is developed to predict the mineral contents. The results show that the predictive performance of this hybrid multi-source model performs statistically significantly better than that of the BP-ANN and GRU models using single source data. In addition, the accuracy of the hybrid model is further improved by using the custom constraint loss function. This method provides high-resolution distributions of mineral content compared to the X-ray powder diffraction testing method. It will help enhance the understanding of the formation.

MicroRNA associated with hepatocyte injury and systemic inflammation may predict adverse outcomes in cirrhotic patients

As the global prevalence of chronic liver disease continues to rise, the need to determine which patients will develop end-stage liver disease and require liver transplantation is increasingly important. However, current prognostic models perform sub-optimally. We aim to determine microRNA profiles associated with clinical decompensation and mortality/transplantation within 1 year. We examined microRNA expression profiles in plasma samples from patients across the spectrum of cirrhosis (n = 154), acute liver failure (ALF) (n = 22), sepsis (n = 20) and healthy controls (HC) (n = 20). We demonstrated that a microRNA-based model (miR-24 and -27a) associated with systemic inflammation differentiated decompensated cirrhosis states from compensated cirrhosis and HC (AUC 0.77 (95% CI 0.69–0.85)). 6 patients within the compensated cirrhosis group decompensated the subsequent year and their exclusion improved model performance (AUC 0.81 (95% CI 0.71–0.89)). miR-191 (associated with liver injury) predicted risk of mortality across the cohort when acutely decompensated and acute-on-chronic-liver failure patients were included. When they were excluded miR-24 (associated with systemic inflammation) predicted risk of mortality. Our findings demonstrate that microRNA associated with systemic inflammation and liver injury predict adverse outcomes in cirrhosis. miR-24 and -191 require further investigation as prognostic biomarkers and therapeutic targets for patients with liver disease.

Impacts of age and BMI on vancomycin model choice in a Bayesian software: Lessons from a very large multi-site retrospective study.

Model-informed precision dosing (MIPD) optimizes drug doses based on pharmacokinetic (PK) model predictions, necessitating careful selection of models tailored to patient characteristics. This study evaluates the predictive performance of various vancomycin PK models across diverse age and BMI categories, drawing insights from a large multi-site database. Adults receiving vancomycin intravenous therapy at United States health systems between January 1, 2022, and December 31, 2023, were included. Patient demographics, vancomycin administration records, and therapeutic drug monitoring levels (TDMs) were collected from the InsightRX database. Age and body mass index (BMI)-based subgroups were formed to assess model performance, with predictions made iteratively. The optimal model for each age-BMI subgroup was chosen based on predefined criteria: models were filtered for mean percentage error (MPE) ≤ 20% and normalized root mean squared error (RMSE) < 8 mg/L, and then the most accurate among them was selected. A total of 384,876 treatment courses across 155 US health systems were analyzed, contributing 841,604 TDMs. Eleven models were compared, showing varying accuracy across age-BMI categories (41%-73%), with higher accuracy observed once TDMs were available for Bayesian estimates of individual PK parameters. Models performed more poorly in younger adults compared to older adults, and the optimal model differed depending on age-BMI categories and prediction methods. Notably, in the a priori period, the Colin model performed best in adults aged 18-64 years across most BMI categories; the Goti/Tong model performed best in the older, non-obese adults; and the Hughes model performed best in many of the obese categories. Our study identifies specific vancomycin PK models that demonstrate superior predictions across age-BMI categories in MIPD applications. Our findings underscore the importance of tailored model selection for vancomycin management, especially highlighting the need for improved models in younger adult patients. Further research into the clinical implications of model performance is warranted to enhance patient care outcomes.

A mental health monitoring system based on intelligent algorithms and biosensors: Algorithm behaviour analysis and intervention strategies

Introduction: Mental health monitoring encompasses the systematic observation and assessment of an individual’s psychological well-being, aiming to detect, understand and manage mental health conditions. It involves various techniques and interventions tailored to support individuals in achieving and maintaining optimal mental wellness. Recent advancements include the use of biosensors and biomechanics to analyze physiological signals that correlate with mental states, providing a more comprehensive understanding of psychological well-being. Aim: The aim of this research is to construct an innovative mental health monitoring system established on intelligent algorithms and biomechanical data through behaviour analysis and intervention strategies. Methodology: We propose a novel Snow Ablation-driven Bi-directional Fine-tuned Recurrent Neural Network (SA-BFRNN) to identify the state of mental health. In addition to behavioural data, biosensors are employed to collect real-time physiological signals such as heart rate variability, skin conductance, and muscle tension, offering objective markers of mental stress and anxiety. These biomechanical inputs are integrated into the system for multi-modal analysis. We employ the SA algorithm, iteratively removing less influential connections and nodes based on their impact on model performance. This process enhances network efficiency and generalization capabilities, refining the BFRNN for mental health state identification. Utilizing a questionnaire with 25 questions, administered to a selected group of 756 individuals, we validate our proposed model. Biosensor data is synchronized with questionnaire responses to improve the precision of mental state identification. Clustering-derived labels are validated with mean opinion score. These labels inform classifiers for individual mental health prediction, aligning with our objective of robust mental health assessment through data-driven approaches. SA-BFRNN integrates both forward and backward temporal information, enhancing its ability to discern subtle patterns in behaviour. Through iterative fine-tuning, our network learns to adapt to diverse datasets, enabling precise identification of mental health states. Research findings: In the result evaluation phase, we thoroughly examine how well our proposed SA-BFRNN model recognizes various states of mental health across different parameters. Our findings also highlight the significance of incorporating biomechanics, where biosensor data showed a strong correlation with mental health indicators, thereby augmenting the accuracy of the system. Our findings emphasize the efficacy of the SA-BFRNN technique, as demonstrated by its overall performance in terms of recall (92.56%), accuracy (90.13%), F1-score (88.16%) and precision (89.23%). Our experimental results unequivocally demonstrate that our proposed model performed better than other traditional approaches in classifying contents from multimodal data, showing notable enhancements in accuracy and robustness, particularly under dynamic conditions.

Geographic traceability of Gastrodia elata Blum based on combination of NIRS and Chemometrics

The content of the active ingredient in G. elata Bl. is affected by the soil and climate of different regions, so geographical traceability is essential to ensure its quality, commercial value. This study used a combination of NIRS and various chemometric methods to establish an effective geotraceability method for G. elata Bl.. Firstly, a traditional machine learning model was built based on the SF dataset NIRS, and a ResNet model was built based on NIRS generated 2DCOS images and 3DCOS images. Secondly, the model performance was validated using the ZT dataset. The results show that the 3DCOS-ResNet model performs the best with 100.00 % and 95.45 % test set and EV accuracy, respectively. This study provides a theoretical basis for regulators to quickly ensure the authenticity of G. elata Bl. sources. However, more data and in-depth studies are needed in the future to validate and improve the applicability of the model.

FluencyBank Timestamped: An Updated Data Set for Disfluency Detection and Automatic Intended Speech Recognition.

This work introduces updated transcripts, disfluency annotations, and word timings for FluencyBank, which we refer to as FluencyBank Timestamped. This data set will enable the thorough analysis of how speech processing models (such as speech recognition and disfluency detection models) perform when evaluated with typical speech versus speech from people who stutter (PWS). We update the FluencyBank data set, which includes audio recordings from adults who stutter, to explore the robustness of speech processing models. Our update (semi-automated with manual review) includes new transcripts with timestamps and disfluency labels corresponding to each token in the transcript. Our disfluency labels capture typical disfluencies (filled pauses, repetitions, revisions, and partial words), and we explore how speech model performance compares for Switchboard (typical speech) and FluencyBank Timestamped. We present benchmarks for three speech tasks: intended speech recognition, text-based disfluency detection, and audio-based disfluency detection. For the first task, we evaluate how well Whisper performs for intended speech recognition (i.e., transcribing speech without disfluencies). For the next tasks, we evaluate how well a Bidirectional Embedding Representations from Transformers (BERT) text-based model and a Whisper audio-based model perform for disfluency detection. We select these models, BERT and Whisper, as they have shown high accuracies on a broad range of tasks in their language and audio domains, respectively. For the transcription task, we calculate an intended speech word error rate (isWER) between the model's output and the speaker's intended speech (i.e., speech without disfluencies). We find isWER is comparable between Switchboard and FluencyBank Timestamped, but that Whisper transcribes filled pauses and partial words at higher rates in the latter data set. Within FluencyBank Timestamped, isWER increases with stuttering severity. For the disfluency detection tasks, we find the models detect filled pauses, revisions, and partial words relatively well in FluencyBank Timestamped, but performance drops substantially for repetitions because the models are unable to generalize to the different types of repetitions (e.g., multiple repetitions and sound repetitions) from PWS. We hope that FluencyBank Timestamped will allow researchers to explore closing performance gaps between typical speech and speech from PWS. Our analysis shows that there are gaps in speech recognition and disfluency detection performance between typical speech and speech from PWS. We hope that FluencyBank Timestamped will contribute to more advancements in training robust speech processing models.