Cross validation of pharmacokinetic bioanalytical methods: Experimental and statistical design.

Pharmacokinetic (PK) analysis is an integral part of drug development. Health agency guidance provides development and validation recommendations for PK bioanalytical methods run in one laboratory. However, as a drug development program progresses, a PK bioanalytical method may need to be run in more than one laboratory. Additionally, a PK bioanalytical method format may change and a new method platform may be validated and implemented during the drug development cycle. Here we describe the cross validation strategy for comparisons of two validated bioanalytical methods used to generate PK data within the same study or across different studies. Current guidance for cross validations is limited and, therefore, Genentech, Inc. has developed a cross validation experimental strategy that utilizes incurred samples along with a comprehensive statistical analysis. One hundred incurred study samples over the applicable range of concentrations are selected based on four quartiles (Q) of in-study concentration levels. The samples are assayed once in the two bioanalytical methods. Bioanalytical method equivalency is assessed for the 100 samples based on pre-specified acceptability criterion: the two methods are considered equivalent if the percent differences in the lower and upper bound limits of the 90 % confidence interval (CI) are both within ±30 %. Quartile by concentration analysis using the same criterion may also need to be performed. A Bland-Altman plot of the percent difference of sample concentrations versus the mean concentration of each sample is also created to help further characterize the data. This strategy is a robust assessment of PK bioanalytical method equivalency and includes subgroup analyses by concentration to assess for biases. This strategy was implemented in two case studies: 1) two different laboratories using the same bioanalytical method and 2) a bioanalytical method platform change from enzyme-linked immunosorbent assay (ELISA) to multiplexing immunoaffinity (IA) liquid chromatography tandem mass spectrometry (IA LC-MS/MS).

SmartScanPCOS: A feature-driven approach to cutting-edge prediction of Polycystic Ovary Syndrome using Machine Learning and Explainable Artificial Intelligence.

PolyCystic Ovarian Syndrome (PCOS) poses significant challenges to women's reproductive health due to its diagnostic complexity arising from a variety of symptoms, including hirsutism, anovulation, pain, obesity, hyperandrogenism, and oligomenorrhea, necessitating multiple clinical tests. Leveraging Artificial Intelligence (AI) in healthcare offers several benefits that can significantly impact patient care, streamline operations, and improve medical outcomes overall. This study presents an Explainable Artificial Intelligence (XAI)-driven PCOS smart predictor, structured as a hierarchical ensemble consisting of two tiers of Random Forest classifiers following extensive analysis of seven conventional classifiers and two additional stacking ensemble classifiers. An open-source data set comprising numerical parametric features linked to PCOS for classifier training was used. Moreover, to identify essential features for PCOS prediction three feature selection methods: Threshold-driven Optimized Principal Component Analysis (TOPCA), Optimized Salp Swarm (OSSM), and Threshold-driven Optimized Mutual Information Method (TOMIM) were fine-tuned through thresholding and improvisation to detect diverse attribute sets with varying numbers and combinations. Notably, the two-level Random Forest classifier model outperformed others with a remarkable 99.31 % accuracy by employing the top 17 features selected through the Threshold-driven Optimized Mutual Information Method (TOMIM) along with anoverallaccuracy of 99.32 % with 8 fold cross validation for 25 runs. The Smart predictor, constructed using Shapash - a Python library for Explainable Artificial Intelligence - was utilized to deploy the two-level Random Forest classifier model. Ensuring transparency and result reliability, visualizations from robust Explainable AI libraries were employed at different prediction stages for all considered classifiers in this study.

iDLB-Pred: identification of disordered lipid binding residues in protein sequences using convolutional neural network.

Proteins, nucleic acids, and lipids all interact with intrinsically disordered protein areas. Lipid-binding regions are involved in a variety of biological processes as well as a number of human illnesses. The expanding body of experimental evidence for these interactions and the dearth of techniques to anticipate them from the protein sequence serve as driving forces. Although large-scale laboratory techniques are considered to be essential for equipment for studying binding residues, they are time consuming and costly, making it challenging for researchers to predict lipid binding residues. As a result, computational techniques are being looked at as a different strategy to overcome this difficulty. To predict disordered lipid-binding residues (DLBRs), we proposed iDLB-Pred predictor utilizing benchmark dataset to compute feature through extraction techniques to identify relevant patterns and information. Various classification techniques, including deep learning methods such as Convolutional Neural Networks (CNNs), Deep Neural Networks (DNNs), Multilayer Perceptrons (MLPs), Recurrent Neural Networks (RNNs), Long Short-Term Memory (LSTM) networks, and Gated Recurrent Units (GRUs), were employed for model training. The proposed model, iDLB-Pred, was rigorously validated using metrics such as accuracy, sensitivity, specificity, and Matthew's correlation coefficient. The results demonstrate the predictor's exceptional performance, achieving accuracy rates of 81% on an independent dataset and 86% in 10-fold cross-validation.

Implementation of a community-based LC-UV drug checking service: promising preliminary findings on feasibility and validity.

BACKGROUND: The increasing diversity of psychoactive substances on the unregulated drug market poses significant health, psychological, and social risks to people who use drugs (PWUD). To address these risks, various harm reduction (HR) policies have been implemented, including drug checking services (DCS). Many analytical methods are used for DCS. While qualitative methods (e.g., thin layer chromatography, spectroscopy) are easier to implement, they are not as accurate as quantitative methods (e.g., LC-UV, LC-MS). Some HR programmes have implemented high-performance liquid chromatography coupled with UV detection (LC-UV). This article presents the cross-validation of this quantitative method with a reference liquid chromatography coupled with high resolution mass spectrometry (LC-HRMS) method. METHODS: Drug samples were provided by PWUD to a DCS called DrugLab in Marseille, France. The samples were weighed and prepared through dissolution in methanol, followed by ultrasonic bathing. Samples were analysed onsite using LC-UV analysis. They were then subsequently analysed with the reference LC-HRMS method. The LC-UV instrument in DrugLab was calibrated after being purchased; analysis of standard solutions was routinely performed once a month and after maintenance operations. For the LC-HRMS instrument, calibration and quality control procedures followed European Medicines Agency (EMA) guidelines. Statistical analyses were conducted including Spearman correlation tests using IBM® SPSS® Statistics version 20. RESULTS: A total of 102 samples representing different product classes and cutting agents were cross-validated. Differences between both analyses methods for each molecule analysed were ≤ 20%, with significant correlations between both methods' results for most substances. Notably, LC-HRMS provided lower concentration values for cocaine and acetaminophen, whereas it provided higher values for other substances. Correlations were significant for cocaine, ketamine, MDMA, heroin, amphetamine, caffeine, acetaminophen, and levamisole. CONCLUSIONS: This study demonstrates that the results provided by DrugLab were accurate and reliable, making LC-UV an adaptable, stable, and suitable analytical method for simple matrices like drugs in a DCS context. However, this cross validation does not guarantee accuracy over time. A proficiency test project in HR laboratories across France is currently under development in order to address potential drifts in LC-UV accuracy.

Empirical investigation of multi-source cross-validation in clinical ECG classification.

Traditionally, machine learning-based clinical prediction models have been trained and evaluated on patient data from a single source, such as a hospital. Cross-validation methods can be used to estimate the accuracy of such models on new patients originating from the same source, by repeated random splitting of the data. However, such estimates tend to be highly overoptimistic when compared to accuracy obtained from deploying models to sources not represented in the dataset, such as a new hospital. The increasing availability of multi-source medical datasets provides new opportunities for obtaining more comprehensive and realistic evaluations of expected accuracy through source-level cross-validation designs. In this study, we present a systematic empirical evaluation of standard K-fold cross-validation and leave-source-out cross-validation methods in a multi-source setting. We consider the task of electrocardiogram based cardiovascular disease classification, combining and harmonizing the openly available PhysioNet/CinC Challenge 2021 and the Shandong Provincial Hospital datasets for our study. Our results show that K-fold cross-validation, both on single-source and multi-source data, systemically overestimates prediction performance when the end goal is to generalize to new sources. Leave-source-out cross-validation provides more reliable performance estimates, having close to zero bias though larger variability. The evaluation highlights the dangers of obtaining misleading cross-validation results on medical data and demonstrates how these issues can be mitigated when having access to multi-source data.

Antecedent Effect Models as an Exploratory Tool to Link Climate Drivers to Herbaceous Perennial Population Dynamics Data.

Understanding mechanisms and predicting natural population responses to climate is a key goal of Ecology. However, studies explicitly linking climate to population dynamics remain limited. Antecedent effect models are a set of statistical tools that capitalize on the evidence provided by climate and population data to select time windows correlated with a response (e.g., survival, reproduction). Thus, these models can serve as both a predictive and exploratory tool. We compare the predictive performance of antecedent effect models against simpler models and showcase their exploratory analysis potential by selecting a case study with high predictive power. We fit three antecedent effect models: (1) weighted mean models (WMM), which weigh the importance of monthly anomalies based on a Gaussian curve, (2) stochastic antecedent models (SAM), which weigh the importance of monthly anomalies using a Dirichlet process, and (3) regularized regressions using the Finnish horseshoe model (FHM), which estimate a separate effect size for each monthly anomaly. We compare these approaches to a linear model using a yearly climatic predictor and a null model with no predictors. We use demographic data from 77 natural populations of 34 plant species ranging between seven and 36 years in length. We then fit models to the asymptotic population growth rate (λ) and its underlying vital rates: survival, development, and reproduction. We find that models including climate do not consistently outperform null models. We hypothesize that the effect of yearly climate is too complex, weak, and confounded by other factors to be easily predicted using monthly precipitation and temperature data. On the other hand, in our case study, antecedent effect models show biologically sensible correlations between two precipitation anomalies and multiple vital rates. We conclude that, in temporal datasets with limited sample sizes, antecedent effect models are better suited as exploratory tools for hypothesis generation.

Evaluation of four machine learning methods in predicting orthodontic extraction decision from clinical examination data and analysis of feature contribution.

Introduction: The study aims to predict tooth extraction decision based on four machine learning methods and analyze the feature contribution, so as to shed light on the important basis for experts of tooth extraction planning, providing reference for orthodontic treatment planning. Methods: This study collected clinical information of 192 patients with malocclusion diagnosis and treatment plans. This study used four machine learning strategies, including decision tree, random forest, support vector machine (SVM) and multilayer perceptron (MLP) to predict orthodontic extraction decisions on clinical examination data acquired during initial consultant containing Angle classification, skeletal classification, maxillary and mandibular crowding, overjet, overbite, upper and lower incisor inclination, vertical growth pattern, lateral facial profile. Among them, 30% of the samples were randomly selected as testing sets. We used five-fold cross-validation to evaluate the generalization performance of the model and avoid over-fitting. The accuracy of the four models was calculated for the training set and cross-validation set. The confusion matrix was plotted for the testing set, and 6 indicators were calculated to evaluate the performance of the model. For the decision tree and random forest models, we observed the feature contribution. Results: The accuracy of the four models in the training set ranges from 82% to 90%, and in the cross-validation set, the decision tree and random forest had higher accuracy. In the confusion matrix analysis, decision tree tops the four models with highest accuracy, specificity, precision and F1-score and the other three models tended to classify too many samples as extraction cases. In the feature contribution analysis, crowding, lateral facial profile, and lower incisor inclination ranked at the top in the decision tree model. Conclusion: Among the machine learning models that only use clinical data for tooth extraction prediction, decision tree has the best overall performance. For tooth extraction decisions, specifically, crowding, lateral facial profile, and lower incisor inclination have the greatest contribution.

Stacking based ensemble learning framework for identification of nitrotyrosine sites.

Protein nitrotyrosine is an essential post-translational modification that results from the nitration of tyrosine amino acid residues. This modification is known to be associated with the regulation and characterization of several biological functions and diseases. Therefore, accurate identification of nitrotyrosine sites plays a significant role in the elucidating progress of associated biological signs. In this regard, we reported an accurate computational tool known as iNTyro-Stack for the identification of protein nitrotyrosine sites. iNTyro-Stack is a machine-learning model based on a stacking algorithm. The base classifiers in stacking are selected based on the highest performance. The feature map employed is a linear combination of the amino composition encoding schemes, including the composition of k-spaced amino acid pairs and tri-peptide composition. The recursive feature elimination technique is used for significant feature selection. The performance of the proposed method is evaluated using k-fold cross-validation and independent testing approaches. iNTyro-Stack achieved an accuracy of 86.3% and a Matthews correlation coefficient (MCC) of 72.6% in cross-validation. Its generalization capability was further validated on an imbalanced independent test set, where it attained an accuracy of 69.32%. iNTyro-Stack outperforms existing state-of-the-art methods across both evaluation techniques. The github repository is create to reproduce the method and results of iNTyro-Stack, accessible on: https://github.com/waleed551/iNTyro-Stack/.

Fusion of spectral and topographic features for land use mapping using a machine learning framework for a regional scale application.

This study investigated the dynamics of land use and land cover (LULC) modelling, mapping, and assessment in the Kegalle District of Sri Lanka, where policy decision-making is crucial in agricultural development where LULC temporal datasets are not readily available. Employing remotely sensed datasets and machine learning algorithms, the work presented here aims to compare the accuracy of three classification approaches in mapping LULC categories across the time in the study area primarily using the Google Earth Engine (GEE). Three classifiers namely random forest (RF), support vector machines (SVM), and classification and regression trees (CART) were used in LULC modelling, mapping, and change analysis. Different combinations of input features were investigated to improve classification performance. Developed models were optimised using the grid search cross-validation (CV) hyperparameter optimisation approach. It was revealed that the RF classifier constantly outstrips SVM and CART in terms of accuracy measures, highlighting its reliability in classifying the LULC. Land cover changes were examined for two periods, from 2001 to 2013 and 2013 to 2022, implying major alterations such as the conversion of rubber and coconut areas to built-up areas and barren lands. For suitable classification with higher accuracy, the study suggests utilising high spatial resolution satellite data, advanced feature selection approaches, and a combination of several spatial and spatial-temporal data sources. The study demonstrated practical applications of derived temporal LULC datasets for land management practices in agricultural development activities in developing nations.

IPECAD Modeling Workshop 2023 Cross Comparison Challenge on Cost-Effectiveness Models in Alzheimer's Disease.

OBJECTIVES: Decision-analytic models assessing the value of emerging Alzheimer's disease (AD) treatments are challenged by limited evidence on short-term trial outcomes and uncertainty in extrapolating long-term patient-relevant outcomes. To improve understanding and foster transparency and credibility in modeling methods, we cross-compared AD decision models in a hypothetical context of disease-modifying treatment for mild cognitive impairment (MCI) due to AD. METHODS: A benchmark scenario (US setting) was used with target population MCI due to AD and a set of synthetically generated hypothetical trial efficacy estimates. Treatment costs were excluded. Model predictions (10-year horizon) were assessed and discussed during a 2-day workshop. RESULTS: Nine modeling groups provided model predictions. Implementation of treatment effectiveness varied across models based on trial efficacy outcome selection (CDR-SB, CDR-global, MMSE, FAQ) and analysis method (observed severity transitions, change from baseline, progression hazard ratio, or calibration to these). Predicted mean time in MCI ranged from 2.6-5.2 years for control strategy, and from 0.1-1.0 years for difference between intervention and control strategies. Predicted quality-adjusted life-year gains ranged from 0.0-0.6 and incremental costs (excluding treatment costs) from -US$66,897 to US$11,896. CONCLUSIONS: Trial data can be implemented in different ways across health-economic models leading to large variation in model predictions. We recommend 1) addressing the choice of outcome measure and treatment effectiveness assumptions in sensitivity analysis, 2) a standardized reporting table for model predictions, and 3) exploring the use of registries for future AD treatments measuring long-term disease progression to reduce uncertainty of extrapolating short-term trial results by health economic models.

Machine Learning-Driven Prediction of Brain Age for Alzheimer's Risk: APOE4 Genotype and Gender Effects.

Background: Alzheimer's disease (AD) is a leading cause of dementia, and it is significantly influenced by the apolipoprotein E4 (APOE4) gene and gender. This study aimed to use machine learning (ML) algorithms to predict brain age and assess AD risk by considering the effects of the APOE4 genotype and gender. Methods: We collected brain volumetric MRI data and medical records from 1100 cognitively unimpaired individuals and 602 patients with AD. We applied three ML regression models-XGBoost, random forest (RF), and linear regression (LR)-to predict brain age. Additionally, we introduced two novel metrics, brain age difference (BAD) and integrated difference (ID), to evaluate the models' performances and analyze the influences of the APOE4 genotype and gender on brain aging. Results: Patients with AD displayed significantly older brain ages compared to their chronological ages, with BADs ranging from 6.5 to 10 years. The RF model outperformed both XGBoost and LR in terms of accuracy, delivering higher ID values and more precise predictions. Comparing the APOE4 carriers with noncarriers, the models showed enhanced ID values and consistent brain age predictions, improving the overall performance. Gender-specific analyses indicated slight enhancements, with the models performing equally well for both genders. Conclusions: This study demonstrates that robust ML models for brain age prediction can play a crucial role in the early detection of AD risk through MRI brain structural imaging. The significant impact of the APOE4 genotype on brain aging and AD risk is also emphasized. These findings highlight the potential of ML models in assessing AD risk and suggest that utilizing AI for AD identification could enable earlier preventative interventions.

Semi-automated reproducible target transfer for cardiac radioablation - A multi-center cross-validation study within the RAVENTA trial.

BACKGROUND: Stereotactic arrhythmia radioablation (STAR) is a therapeutic option for ventricular tachycardia (VT) where catheter-based ablation is not feasible or has previously failed. Target definition and its transfer from electro-anatomic maps (EAM) to radiotherapy treatment planning systems (TPS) is challenging and operator-dependent. Software solutions have been developed to register EAM with cardiac CT and semi-automatically transfer 2D target surface data into 3D CT volume coordinates. Results of a cross-validation study of two conceptually different software solutions using data from the RAVENTA trial (NCT03867747) are reported. METHODS: Clinical Target Volumes (CTVs) were created from target regions delineated on EAM using two conceptually different approaches by separate investigators on data of 10 patients, blinded to each other's results. Targets were transferred using 3D-3D registration and 2D-3D registration, respectively. The resulting CTVs were compared in a core-lab using two complementary analysis software packages for structure similarity and geometric characteristics. RESULTS: Volumes and surface areas of the CTVs created by both methods were comparable: 14.88 ± 11.72 ml versus 15.15 ± 11.35 ml and 44.29 ± 33.63 cm2 versus 46.43 ± 35.13 cm2. The Dice-coefficient was 0.84 ± 0.04; median surface-distance and Hausdorff-distance were 0.53 ± 0.37 mm and 6.91 ± 2.26 mm, respectively. The 3D-center-of-mass difference was 3.62 ± 0.99 mm. Geometrical volume similarity was 0.94 ± 0.05 %. CONCLUSION: The STAR targets transferred from EAM to TPS using both software solutions resulted in nearly identical 3D structures. Both solutions can be used for QA (quality assurance) and EAM-to-TPS transfer of STAR-targets. Semi-automated methods could potentially help to avoid mistargeting in STAR and offer standardized workflows for methodically harmonized treatments.

ICH M10 Bioanalytical Method Validation Guideline-1 year Later.

The International Council for Harmonization of Technical Requirements for Pharmaceuticals for Human Use (ICH) adopted Guideline M10 entitled "Bioanalytical Method Validation and Study Sample Analysis" in May 2022. In October 2023, approximately one year after the adoption of the ICH M10 guideline, a "Hot Topic" session was held during the AAPS PharmSci 360 meeting to discuss the implementation of the guideline. The session focused on items the bioanalytical community felt were challenging to implement or ambiguous within the guideline. These topics included cross-validation, parallelism, comparative bioavailability studies, combination drug stability, endogenous analyte bioanalysis, and dilution QCs. In addition, the regulatory perspective on the guideline was presented. This report provides a summary of the Hot Topic session.

High-Throughput Image-Based Assay for Identifying In Vitro Hepatocyte Microtubule Disruption.

Disruption of microtubule stability in mammalian cells may lead to genotoxicity and carcinogenesis. The ability to screen for microtubule destabilization or stabilization is therefore a useful and efficient approach to aid in the design of molecules that are safe for human health. In this study, we developed a high-throughput 384-well assay combining immunocytochemistry with high-content imaging to assess microtubule disruption in the metabolically competent human liver cell line: HepaRG. To enhance analysis throughput, we implemented a supervised machine learning approach using a curated training library of 180 compounds. A majority voting ensemble of eight machine learning classifiers was employed for predicting microtubule disruptions. Our prediction model achieved over 99.0% accuracy and a 98.4% F1 score, which reflects the balance between precision and recall for in-sample validation and 93.5% accuracy and a 94.3% F1 score for out-of-sample validation. This automated image-based testing can provide a simple, high-throughput screening method for early stage discovery compounds to reduce the potential risk of genotoxicity for crop protection product development.

Comparison of machine learning algorithms and multiple linear regression for live weight estimation of Akkaraman lambs.

This study was designed to predict the post-weaning weights of Akkaraman lambs reared on different farms using multiple linear regression and machine learning algorithms. The effect of factors the age of the dam, gender, type of lambing, enterprise, type of flock, birth weight, and weaning weight was analyzed. The data was collected from a total of 25,316 Akkaraman lambs raised at multiple farms in the Çiftlik District of Nigde province. Comparative analysis was conducted by using multiple linear regression, Random Forest, Support Vector Machines (and Support Vector Regression), Extreme Gradient Boosting (XGBoost) (and Gradient Boosting), Bayesian Regularized Neural Network, Radial Basis Function Neural Network, Classification and Regression Trees, Exhaustive Chi-squared Automatic Interaction Detection (and Chi-squared Automatic Interaction Detection), and Multivariate Adaptive Regression Splines algorithms. In this study, the test dataset was divided into five layers using the K-fold cross-validation method. The performance of models was compared using performance criteria such as Adjusted R-squared (Adj-[Formula: see text]), Root Mean Square Error (RMSE), Mean Absolute Deviation (MAD), and Mean Absolute Percentage Error (MAPE) by utilizing test populations in the predicted models. Additionally, the presence of low standard deviations for these criteria indicates the absence of an overfitting problem. [Formula: see text]The comparison results showed the Random Forest algorithm had the best predictive performance compared to other algorithms with Adj-[Formula: see text], RMSE, MAD, and MAPE values of 0.75, 3.683, 2.876, and 10.112, respectively. In conclusion, the results obtained through Multiple Linear Regression for the live weights of Akkaraman lambs were less accurate than the results obtained through artificial neural network analysis.

Refining penalty parameter selection in whole-body PET image reconstruction for lung cancer patients using the cross-validation log-likelihood method.

Objective.Penalty parameters in penalized likelihood positron emission tomography (PET) reconstruction are typically determined empirically. The cross-validation log-likelihood (CVLL) method has been introduced to optimize these parameters by maximizing a CVLL function, which assesses the likelihood of reconstructed images using one subset of a list-mode dataset based on another subset. This study aims to validate the efficacy of the CVLL method in whole-body imaging for cancer patients using a conventional clinical PET scanner.Approach.Fifteen lung cancer patients were injected with 243.7 ± 23.8 MBq of [18F]FDG and underwent a 22 min PET scan on a Biograph mCT PET/CT scanner, starting at 60 ± 5 min post-injection. The PET list-mode data were partitioned by subsampling without replacement, with 20 minutes of data for image reconstruction using an in-house ordered subset expectation maximization algorithm and the remaining 2 minutes of data for cross-validation. Two penalty parameters, penalty strengthßand Fair penalty function parameterδ, were subjected to optimization. Whole-body images were reconstructed, and CVLL values were computed across various penalty parameter combinations. The optimal image corresponding to the maximum CVLL value was selected by a grid search for each patient.Main results.Theδvalue required to maximize the CVLL value was notably small (⩽10-6in this study). The influences of voxel size and scan duration on image optimization were investigated. A correlation analysis revealed a significant inverse relationship between optimalßand scan count level, with a correlation coefficient of -0.68 (p-value = 3.5 × 10-5). The optimal images selected by the CVLL method were compared with those chosen by two radiologists based on their diagnostic preferences. Differences were observed in the selection of optimal images.Significance.This study demonstrates the feasibility of incorporating the CVLL method into routine imaging protocols, potentially allowing for a wide range of combinations of injected radioactivity amounts and scan durations in modern PET imaging.

Quantitative Prediction and Analysis of Rattle Index Using DNN on Sound Quality of Synthetic Sources with Gaussian Noise.

This study researched the prediction of the BSR noise evaluation quantitative index, Loudness N10, for sound sources with noise using statistics and machine learning. A total of 1170 data points was obtained from 130 automotive seats measured at 9-point positions, with Gaussian noise integrated to construct synthetic sound data. Ten physical quantities related to sound quality and sound pressure were used and defined as dB and fluctuation strength, considering statistical characteristics and Loudness N10. BSR quantitative index prediction was performed using regression analysis with K-fold cross-validation, DNN in hold-out, and DNN in K-fold cross-validation. The DNN in the K-fold cross-validation model demonstrated relatively superior prediction accuracy, especially when the data quantity was relatively small. The results demonstrate that applying machine learning to BSR prediction allows for the prediction of quantitative indicators without complex formulas and that specific physical quantities can be easily estimated even with noise.

Enhanced Hand Gesture Recognition with Surface Electromyogram and Machine Learning.

This study delves into decoding hand gestures using surface electromyography (EMG) signals collected via a precision Myo-armband sensor, leveraging machine learning algorithms. The research entails rigorous data preprocessing to extract features and labels from raw EMG data. Following partitioning into training and testing sets, four traditional machine learning models are scrutinized for their efficacy in classifying finger movements across seven distinct gestures. The analysis includes meticulous parameter optimization and five-fold cross-validation to evaluate model performance. Among the models assessed, the Random Forest emerges as the top performer, consistently delivering superior precision, recall, and F1-score values across gesture classes, with ROC-AUC scores surpassing 99%. These findings underscore the Random Forest model as the optimal classifier for our EMG dataset, promising significant advancements in healthcare rehabilitation engineering and enhancing human-computer interaction technologies.

A novel nomogram for predicting overall survival in patients with tongue squamous cell carcinoma using clinical features and MRI radiomics data: a pilot study.

OBJECTIVE: Tongue squamous cell carcinoma (TSCC) accounts for 43.4% of oral cancers in China and has a poor prognosis. This study aimed to explore whether radiomics features extracted from preoperative magnetic resonance imaging (MRI) could predict overall survival (OS) in patients with TSCC. METHODS: The clinical imaging data of 232 patients with pathologically confirmed TSCC at Xiangyang No. 1 People's Hospital were retrospectively analyzed from February 2010 to October 2022. Based on 2-10 years of follow-up, patients were categorized into two groups: control (healthy survival, n = 148) and research (adverse events: recurrence or metastasis-related death, n = 84). A training and a test set were established using a 7:3 ratio and a time node. Radiomics features were extracted from axial T2-weighted imaging, contrast-enhanced T1-weighted imaging, and diffusion-weighted imaging (DWI) sequences. The corresponding radiomics scores were generated using the least absolute shrinkage and selection operator algorithm. Kaplan-Meier and multivariate Cox regression analyses were used to screen for independent factors affecting adverse events in patients with TSCC using clinical and pathological results. A novel nomogram was established to predict the probability of adverse events and OS in patients with TSCC. RESULTS: The incidence of adverse events within 2-10 years after surgery was 36.21%. Kaplan-Meier analysis revealed that hot pot consumption, betel nut chewing, platelet-lymphocyte ratio, drug use, neutrophil-lymphocyte ratio, Radscore, and other factors impacted TSCC survival. Multivariate Cox regression analysis revealed that the clinical stage (P < 0.001), hot pot consumption (P < 0.001), Radscore 1 (P = 0.01), and Radscore 2 (P < 0.001) were independent factors affecting TSCC-OS. The same result was validated by the XGBoost algorithm. The nomogram based on the aforementioned factors exhibited good discrimination (C-index 0.86/0.81) and calibration (P > 0.05) in the training and test sets, accurately predicting the risk of adverse events and survival. CONCLUSION: The nomogram constructed using clinical data and MRI radiomics parameters may accurately predict TSCC-OS noninvasively, thereby assisting clinicians in promptly modifying treatment strategies to improve patient prognosis.

Selection of optimal knot point and best geographic weighted on geographically weighted spline nonparametric regression model.

This study proposes the development of a nonparametric regression model combined with geographically weighted regression. The regression model considers geographical factors and has a data pattern that does not follow a parametric form to overcome the problem of spatial heterogeneity and unknown regression functions. This study aims to model provincial food security index data in Indonesia with the GWSNR model, so finding the optimal knot point and the best geographic weighting is necessary. We propose the selection of optimal knot points using the Cross Validation (CV) and Generalized Cross Validation (GCV) methods. The optimal knot point will control the accuracy of the regression curve as we also consider the MSE value in showing the ability of the model. In addition, we determine the best geographic weighting and test the significance of the model parameters. We demonstrate the GWSNR model on food security index data. The best GWSNR model uses the Gaussian kernel weighting function and selects the optimal knot point as one-knot point based on the lowest CV and GCV values. Simultaneous and partial parameter test results show that there are 10 area classifications with different effects on each group of classification results. Some of the highlights of the proposed approach are:•The method is the development of a nonparametric regression model with geographic weighting, which combines nonparametric and spatial regression in modeling the national food security index.•There are three-knot points tested in nonparametric truncated spline regression and three geographic weightings in spatial regression. Then the optimal knot point and best bandwidth are determined using Cross Validation and Generalized Cross Validation.•This article will determine regional groupings in Indonesia in 2022 based on significant predictors in modeling the national food security index numbers.