Assess Model Performance Research Articles

A Machine Learning Pipeline for Cervical Cancer Prediction

Cervical cancer ranks as the fourth leading cause of cancer-related deaths among women and often remains asymptomatic during its initial stages. Therefore, it’s very important to predict it. The study's objective is to explore cervical cancer risk variables and develop an ensemble prediction model to forecast cervical cancer more precisely. The cervical cancer data were represented by 11 risk factors and a dependent variable consisted of Hinselmann, Schiller, Cytology, and Biopsy. SVM, decision tree, and random forest methods are compared for prediction, with models trained on 75% of the data and evaluated using confusion matrices. Precision and recall metrics assess model performance, with recall being vital in critical situations like identifying cervical cancer patients. The comparison was made among these three methods and compared the ranking result of risk factors with the ground truth, random forest method offers high recall (0.8821) and accuracy (0.8744). Recognizing risk factors can enhance cure rates and female patient survival. And the cervical cancer prediction system enhances medical care and reduces the cost. Data mining techniques in the medical sector facilitate diagnostic and prognostic apps for early treatment of life-threatening diseases.

MESED: A Multi-Modal Entity Set Expansion Dataset with Fine-Grained Semantic Classes and Hard Negative Entities

The Entity Set Expansion (ESE) task aims to expand a handful of seed entities with new entities belonging to the same semantic class. Conventional ESE methods are based on mono-modality (i.e., literal modality), which struggle to deal with complex entities in the real world such as (1) Negative entities with fine-grained semantic differences. (2) Synonymous entities. (3) Polysemous entities. (4) Long-tailed entities. These challenges prompt us to propose novel Multi-modal Entity Set Expansion (MESE), where models integrate information from multiple modalities to represent entities. Intuitively, the benefits of multi-modal information for ESE are threefold: (1) Different modalities can provide complementary information. (2) Multi-modal information provides a unified signal via common visual properties for the same semantic class or entity. (3) Multi-modal information offers robust alignment signals for synonymous entities. To assess model performance in MESE, we constructed the MESED dataset which is the first multi-modal dataset for ESE with large-scale and elaborate manual calibration. A powerful multi-modal model MultiExpan is proposed which is pre-trained on four multimodal pre-training tasks. The extensive experiments and analyses on MESED demonstrate the high quality of the dataset and the effectiveness of our MultiExpan, as well as pointing the direction for future research. The benchmark and code are public at https://github.com/THUKElab/MESED.

Abstract 7392: Machine learning-based identification of H&E-derived morphologic features associated with CD8+ T cell immune exclusion

Abstract Background: Immune exclusion is characterized by a predominance of CD8+ T cells in the stroma of the tumor microenvironment (TME) which are not in contact with tumor cells. Immune exclusion is a significant obstacle to effective cancer immunotherapy and reversing it has emerged as a new therapeutic opportunity. However, the objective assessment of immune exclusion remains a challenge due to the complexity of the TME and the need for immunostaining to quantify CD8+ cells in tumor and stroma. While artificial intelligence (AI) models applied to H&E images can quantify spatial tissue and cell features at scale, they are unable to identify the molecular subtypes of cells. Here we describe the development of a machine learning model that classifies tumors by their CD8-defined immune phenotypes using features extracted from H&E whole slide images alone. Methods: Al-powered TME models (PathAI PathExplore™) were deployed on H&E images from 39 non-small lung cancer (NSCLC) and triple-negative breast cancer (TNBC) samples. CD8-based assessment of immune phenotypes was done using an adjacent section. 115 H&E-derived features from each of the 25 tumor samples (12 NSCLC, 13 TNBC) classified as immune excluded and 7 tumor samples (3 NSCLC, 4 TNBC) classified as immune infiltrated were used as input into a feature selection process to identify the subset of H&E features predictive of immune phenotype. The feature selection process was driven by an ensemble L1 norm support vector machine (Ens-L1-SVM) model containing 500 L1-SVM base models each trained by 80% of the data. The degree of immune exclusion was determined in the feature space by the distance to the separation hyperplane defined by the model. Out-of-bag error estimation was employed on the remaining 20% of data from each base model to assess model performance. Results: The Ens-L1-SVM model identified 6 H&E features predictive of CD8-defined immune phenotypes, including the density ratio of lymphocytes to macrophages in the cancer epithelium, the density of fibroblasts, and the spatial distributions of cancer cells, macrophages, and lymphocytes. Immune phenotypes were classified by the Ens-L1-SVM model with an estimated accuracy of 97%. Additionally, all selected features and their direction of association aligned with prior knowledge of the mechanisms or manifestations of immune exclusion or infiltration in the TME. Conclusion: This work illustrated that morphologic features derived from H&E images using AI-powered models can be effective predictors of CD8-defined immune exclusion, providing an option for patient stratification by immune phenotype using widely-available H&E images. While the Ens-L1-SVM model showed robust performance on a small number of samples and is expected to have good generalizability due to the linear nature of the model, additional validation across diverse datasets and various tumor indications is needed. Citation Format: Yanchao Wang, Fredrick D. Gootkind, Florent Peyraud, Jean-Philippe Guégan, Antoine Italiano, G. Travis Clifton, Laura A. Dillon, Xinwei Sher. Machine learning-based identification of H&E-derived morphologic features associated with CD8+ T cell immune exclusion [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 7392.

A computational model of motion sickness dynamics during passive self-motion in the dark.

Predicting the time course of motion sickness symptoms enables the evaluation of provocative stimuli and the development of countermeasures for reducing symptom severity. In pursuit of this goal, we present an Observer-driven model of motion sickness for passive motions in the dark. Constructed in two stages, this model predicts motion sickness symptoms by bridging sensory conflict (i.e., differences between actual and expected sensory signals) arising from the Observer model of spatial orientation perception (stage 1) to Oman's model of motion sickness symptom dynamics (stage 2; presented in 1982 and 1990) through a proposed "Normalized Innovation Squared" statistic. The model outputs the expected temporal development of human motion sickness symptom magnitudes (mapped to the Misery Scale) at a population level, due to arbitrary, 6-degree-of-freedom, self-motion stimuli. We trained model parameters using individual subject responses collected during fore-aft translations and off-vertical axis of rotation motions. Improving on prior efforts, we only used datasets with experimental conditions congruent with the perceptual stage (i.e., adequately provided passive motions without visual cues) to inform the model. We assessed model performance by predicting an unseen validation dataset, producing a Q2 value of 0.91. Demonstrating this model's broad applicability, we formulate predictions for a host of stimuli, including translations, earth-vertical rotations, and altered gravity, and we provide our implementation for other users. Finally, to guide future research efforts, we suggest how to rigorously advance this model (e.g., incorporating visual cues, active motion, responses to motion of different frequency, etc.).

‘Drivin' with your eyes closed’: Results from an international, blinded simulation experiment to evaluate spatial stock assessments

AbstractSpatial models enable understanding potential redistribution of marine resources associated with ecosystem drivers and climate change. Stock assessment platforms can incorporate spatial processes, but have not been widely implemented or simulation tested. To address this research gap, an international simulation experiment was organized. The study design was blinded to replicate uncertainty similar to a real‐world stock assessment process, and a data‐conditioned, high‐resolution operating model (OM) was used to emulate the spatial dynamics and data for Indian Ocean yellowfin tuna (Thunnus albacares). Six analyst groups developed both single‐region and spatial stock assessment models using an assessment platform of their choice, and then applied each model to the simulated data. Results indicated that across all spatial structures and platforms, assessments were able to adequately recreate the population trends from the OM. Additionally, spatial models were able to estimate regional population trends that generally reflected the true dynamics from the OM, particularly for the regions with higher biomass and fishing pressure. However, a consistent population biomass scaling pattern emerged, where spatial models estimated higher population scale than single‐region models within a given assessment platform. Balancing parsimony and complexity trade‐offs were difficult, but adequate complexity in spatial parametrizations (e.g., allowing time‐ and age‐variation in movement and appropriate tag mixing periods) was critical to model performance. We recommend expanded use of high‐resolution OMs and blinded studies, given their ability to portray realistic performance of assessment models. Moreover, increased support for international simulation experiments is warranted to facilitate dissemination of methodology across organizations.

Unravelling the role of tropical cyclones in shaping present species distributions.

Driven by climate change, tropical cyclones (TCs) are predicted to change in intensity and frequency through time. Given these forecasted changes, developing an understanding of how TCs impact insular wildlife is of heightened importance. Previous work has shown that extreme weather events may shape species distributions more strongly than climatic averages; however, given the coarse spatial and temporal scales at which TC data are often reported, the influence of TCs on species distributions has yet to be explored. Using TC data from the National Hurricane Center, we developed spatially and temporally explicit species distribution models (SDMs) to examine the role of TCs in shaping present-day distributions of Puerto Rico's 10 Anolis lizard species. We created six predictor variables to represent the intensity and frequency of TCs. For each occurrence of a species, we calculated these variables for TCs that came within 500 km of the center of Puerto Rico and occurred within the 1-year window prior to when that occurrence was recorded. We also included predictor variables related to landcover, climate, topography, canopy cover and geology. We used random forests to assess model performance and variable importance in models with and without TC variables. We found that the inclusion of TC variables improved model performance for the majority of Puerto Rico's 10 anole species. The magnitude of the improvement varied by species, with generalist species that occur throughout the island experiencing the greatest improvements in model performance. Range-restricted species experienced small, almost negligible, improvements but also had more predictive models both with and without the inclusion of TC variables compared to generalist species. Our findings suggest that incorporating data on TCs into SDMs may be important for modeling insular species that are prone to experiencing these types of extreme weather events.

Performance of logistic regression analyses in the identification of inner ear synapse pathology

Currently, a diagnostic tool to identify cochlear synaptopathy in humans is unavailable. However, several features of electrophysiological responses evoked by auditory stimuli have been evaluated that may serve as useful indicators of synapse pathology. This study is part of a larger project whose goal is to develop a statistical model designed to accurately and reliably detect cochlear synaptopathy in humans. Univariate logistic regression analyses were employed to identify electrophysiological outcomes that predict synapse pathology in a guinea pig model and the relative performance of each model was evaluated. Previous efforts to assess model performance included area under the Receiver Operating Characteristic curve. In this report, metrics, such as F1-score and Matthews Correlation Coefficient (MCC), were included. Expectations are that analyses that incorporate true negatives, as does MCC, will more completely describe the performance of the binary classification of interest here: synapse pathology or synapse normalcy. Findings will be presented in the context of a non-human mammalian model with the ultimate purpose of developing a statistical model that can be used to optimize the diagnosis of synapse pathology in humans. [Work supported by the Department of Defense Award #W81XWH-19-1-0862.]

Prediction of 30-day, 90-day, and 1-year mortality after colorectal cancer surgery using a data-driven approach

PurposeTo develop prediction models for short-term mortality risk assessment following colorectal cancer surgery.MethodsData was harmonized from four Danish observational health databases into the Observational Medical Outcomes Partnership Common Data Model. With a data-driven approach using the Least Absolute Shrinkage and Selection Operator logistic regression on preoperative data, we developed 30-day, 90-day, and 1-year mortality prediction models. We assessed discriminative performance using the area under the receiver operating characteristic and precision-recall curve and calibration using calibration slope, intercept, and calibration-in-the-large. We additionally assessed model performance in subgroups of curative, palliative, elective, and emergency surgery.ResultsA total of 57,521 patients were included in the study population, 51.1% male and with a median age of 72 years. The model showed good discrimination with an area under the receiver operating characteristic curve of 0.88, 0.878, and 0.861 for 30-day, 90-day, and 1-year mortality, respectively, and a calibration-in-the-large of 1.01, 0.99, and 0.99. The overall incidence of mortality were 4.48% for 30-day mortality, 6.64% for 90-day mortality, and 12.8% for 1-year mortality, respectively. Subgroup analysis showed no improvement of discrimination or calibration when separating the cohort into cohorts of elective surgery, emergency surgery, curative surgery, and palliative surgery.ConclusionWe were able to train prediction models for the risk of short-term mortality on a data set of four combined national health databases with good discrimination and calibration. We found that one cohort including all operated patients resulted in better performing models than cohorts based on several subgroups.

Transfer Learning-Based B-Line Assessment of Lung Ultrasound for Acute Heart Failure

ObjectiveB-lines assessed by lung ultrasound (LUS) outperform physical exam, chest radiograph, and biomarkers for the associated diagnosis of acute heart failure (AHF) in the emergent setting. The use of LUS is however limited to trained professionals and suffers from interpretation variability. The objective was to utilize transfer learning to create an AI-enabled software that can aid novice users to automate LUS B-line interpretation. MethodsData from an observational AHF LUS study provided standardized cine clips for AI model development and evaluation. A total of 49,952 LUS frames from 30 patients were hand scored and trained on a convolutional neural network (CNN) to interpret B-lines at the frame level. A random independent evaluation set of 476 LUS clips from 60 unique patients assessed model performance. The AI models scored the clips on both a binary and ordinal 0–4 multiclass assessment. ResultsA multiclassification AI algorithm had the best performance at the binary level when applied to the independent evaluation set, AUC of 0.967 (95% CI 0.965–0.970) for detecting pathologic conditions. When compared to expert blinded reviewer, the 0–4 multiclassification AI algorithm scale had a reported linear weighted kappa of 0.839 (95% CI 0.804–0.871). ConclusionsThe multiclassification AI algorithm is a robust and well performing model at both binary and ordinal multiclass B-line evaluation. This algorithm has the potential to be integrated into clinical workflows to assist users with quantitative and objective B-line assessment for evaluation of AHF.

Deep learning segmentation of the choroid plexus from structural magnetic resonance imaging (MRI): validation and normative ranges across the adult lifespan

BackgroundThe choroid plexus functions as the blood-cerebrospinal fluid (CSF) barrier, plays an important role in CSF production and circulation, and has gained increased attention in light of the recent elucidation of CSF circulation dysfunction in neurodegenerative conditions. However, methods for routinely quantifying choroid plexus volume are suboptimal and require technical improvements and validation. Here, we propose three deep learning models that can segment the choroid plexus from commonly-acquired anatomical MRI data and report performance metrics and changes across the adult lifespan.MethodsFully convolutional neural networks were trained from 3D T1-weighted, 3D T2-weighted, and 2D T2-weighted FLAIR MRI using gold-standard manual segmentations in control and neurodegenerative participants across the lifespan (n = 50; age = 21–85 years). Dice coefficients, 95% Hausdorff distances, and area-under-curve (AUCs) were calculated for each model and compared to segmentations from FreeSurfer using two-tailed Wilcoxon tests (significance criteria: p < 0.05 after false discovery rate multiple comparisons correction). Metrics were regressed against lateral ventricular volume using generalized linear models to assess model performance for varying levels of atrophy. Finally, models were applied to an expanded cohort of adult controls (n = 98; age = 21–89 years) to provide an exemplar of choroid plexus volumetry values across the lifespan.ResultsDeep learning results yielded Dice coefficient = 0.72, Hausdorff distance = 1.97 mm, AUC = 0.87 for T1-weighted MRI, Dice coefficient = 0.72, Hausdorff distance = 2.22 mm, AUC = 0.87 for T2-weighted MRI, and Dice coefficient = 0.74, Hausdorff distance = 1.69 mm, AUC = 0.87 for T2-weighted FLAIR MRI; values did not differ significantly between MRI sequences and were statistically improved compared to current commercially-available algorithms (p < 0.001). The intraclass coefficients were 0.95, 0.95, and 0.96 between T1-weighted and T2-weighted FLAIR, T1-weighted and T2-weighted, and T2-weighted and T2-weighted FLAIR models, respectively. Mean lateral ventricle choroid plexus volume across all participants was 3.20 ± 1.4 cm3; a significant, positive relationship (R2 = 0.54-0.60) was observed between participant age and choroid plexus volume for all MRI sequences (p < 0.001).ConclusionsFindings support comparable performance in choroid plexus delineation between standard, clinically available, non-contrasted anatomical MRI sequences. The software embedding the evaluated models is freely available online and should provide a useful tool for the growing number of studies that desire to quantitatively evaluate choroid plexus structure and function (https://github.com/hettk/chp_seg).

Deep learning for automatic bowel-obstruction identification on abdominal CT.

Automated evaluation of abdominal computed tomography (CT) scans should help radiologists manage their massive workloads, thereby leading to earlier diagnoses and better patient outcomes. Our objective was to develop a machine-learning model capable of reliably identifying suspected bowel obstruction (BO) on abdominal CT. The internal dataset comprised 1345 abdominal CTs obtained in 2015-2022 from 1273 patients with suspected BO; among them, 670 were annotated as BO yes/no by an experienced abdominal radiologist. The external dataset consisted of 88 radiologist-annotated CTs. We developed a full preprocessing pipeline for abdominal CT comprising a model to locate the abdominal-pelvic region and another model to crop the 3D scan around the body. We built, trained, and tested several neural-network architectures for the binary classification (BO, yes/no) of each CT. F1 and balanced accuracy scores were computed to assess model performance. The mixed convolutional network pretrained on a Kinetics 400 dataset achieved the best results: with the internal dataset, the F1 score was 0.92, balanced accuracy 0.86, and sensitivity 0.93; with the external dataset, the corresponding values were 0.89, 0.89, and 0.89. When calibrated on sensitivity, this model produced 1.00 sensitivity, 0.84 specificity, and an F1 score of 0.88 with the internal dataset; corresponding values were 0.98, 0.76, and 0.87 with the external dataset. The 3D mixed convolutional neural network developed here shows great potential for the automated binary classification (BO yes/no) of abdominal CT scans from patients with suspected BO. The 3D mixed CNN automates bowel obstruction classification, potentially automating patient selection and CT prioritization, leading to an enhanced radiologist workflow. • Bowel obstruction's rising incidence strains radiologists. AI can aid urgent CT readings. • Employed 1345 CT scans, neural networks for bowel obstruction detection, achieving high accuracy and sensitivity on external testing. • 3D mixed CNN automates CT reading prioritization effectively and speeds up bowel obstruction diagnosis.

Ultrasound radiomics signature for predicting central lymph node metastasis in clinically node-negative papillary thyroid microcarcinoma

BackgroundWhether prophylactic central lymph node dissection is necessary for patients with clinically node-negative (cN0) papillary thyroid microcarcinoma (PTMC) remains controversial. Herein, we aimed to establish an ultrasound (US) radiomics (Rad) score for assessing the probability of central lymph node metastasis (CLNM) in such patients.Methods480 patients (327 in the training cohort, 153 in the validation cohort) who underwent thyroid surgery for cN0 PTMC at two institutions between January 2018 and December 2020 were included. Radiomics features were extracted from the US images. Least absolute shrinkage and selection operator logistic regression were utilized to generate a Rad score. A nomogram consisting of the Rad score and clinical factors was then constructed for the training cohort. Both cohorts assessed model performance using discrimination, calibration, and clinical usefulness.ResultsBased on the six most valuable radiomics features, the Rad score was calculated for each patient. A multivariate analysis revealed that a higher Rad score (P < 0.001), younger age (P = 0.006), and presence of capsule invasion (P = 0.030) were independently associated with CLNM. A nomogram integrating these three factors demonstrated good calibration and promising clinical utility in the training and validation cohorts. The nomogram yielded areas under the curve of 0.795 (95% confidence interval [CI], 0.745–0.846) and 0.774 (95% CI, 0.696–0.852) in the training and validation cohorts, respectively.ConclusionsThe radiomics nomogram may be a clinically useful tool for the individual prediction of CLNM in patients with cN0 PTMC.

A Standardized Benchmarking Framework to Assess Downscaled Precipitation Simulations

Abstract Presently, there is no standardized framework or metrics identified to assess regional climate model precipitation output. Because of this, it can be difficult to make a one-to-one comparison of their performance between regions or studies, or against coarser-resolution global climate models. To address this, we introduce the first steps toward establishing a dynamic, yet standardized, benchmarking framework that can be used to assess model skill in simulating various characteristics of rainfall. Benchmarking differs from typical model evaluation in that it requires that performance expectations are set a priori. This framework has innumerable applications to underpin scientific studies that assess model performance, inform model development priorities, and aid stakeholder decision-making by providing a structured methodology to identify fit-for-purpose model simulations for climate risk assessments and adaptation strategies. While this framework can be applied to regional climate model simulations at any spatial domain, we demonstrate its effectiveness over Australia using high-resolution, 0.5° × 0.5° simulations from the CORDEX-Australasia ensemble. We provide recommendations for selecting metrics and pragmatic benchmarking thresholds depending on the application of the framework. This includes a top tier of minimum standard metrics to establish a minimum benchmarking standard for ongoing climate model assessment. We present multiple applications of the framework using feedback received from potential user communities and encourage the scientific and user community to build on this framework by tailoring benchmarks and incorporating additional metrics specific to their application. Significance Statement We introduce a standardized benchmarking framework for assessing the skill of regional climate models in simulating precipitation. This framework addresses the lack of a uniform approach in the scientific community and has diverse applications in scientific research, model development, and societal decision-making. We define a set of minimum standard metrics to underpin ongoing climate model assessments that quantify model skill in simulating fundamental characteristics of rainfall. We provide guidance for selecting metrics and defining benchmarking thresholds, demonstrated using multiple case studies over Australia. This framework has broad applications for numerous user communities and provides a structured methodology for the assessment of model performance.

Sample Size Requirements of a Pharmaceutical Material Library: A Case in Predicting Direct Compression Tablet Tensile Strength by Latent Variable Modeling.

The material library is an emerging, new data-driven approach for developing pharmaceutical process models. How many materials or samples should be involved in a particular application scenario is unclear, and the impact of sample size on process modeling is worth discussing. In this work, the direct compression process was taken as the research object, and the effects of different sample sizes of material libraries on partial least squares (PLS) modeling in the prediction of tablet tensile strength were investigated. A primary material library comprising 45 materials was built. Then, material subsets containing 5 × i (i = 1, 2, 3, …, 8) materials were sampled from the primary material library. Each subset underwent sampling 1000 times to analyze variations in model fitting performance. Both hierarchical sampling and random sampling were employed and compared, with hierarchical sampling implemented with the help of the tabletability classification index d. For each subset, modeling data were organized, incorporating 18 physical properties and tableting pressure as the independent variables and tablet tensile strength as the dependent variable. A series of chemometric indicators was used to assess model performance and find important materials for model training. It was found that the minimum R2 and RMSE values reached their maximum, and the corresponding values were kept almost unchanged when the sample sizes varied from 20 to 45. When the sample size was smaller than 15, the hierarchical sampling method was more reliable in avoiding low-quality few-shot PLS models than the random sampling method. Two important materials were identified as useful for building an initial material library. Overall, this work demonstrated that as the number of materials increased, the model's reliability improved. It also highlighted the potential for effective few-shot modeling on a small material library by controlling its information richness.

Fractal Analysis of Left Ventricular Trabeculae in Patients with End-Stage Renal Disease: A Random Survival Tree Analysis.

The complexity of left ventricular (LV) trabeculae is related to the prognosis of several cardiovascular diseases. To evaluate the prognostic value of LV trabecular complexity in patients with end-stage renal disease (ESRD). Prospective outcome study. 207 participants on maintenance dialysis, divided into development (160 patients from 2 centers) and external validation (47 patients from a third center) cohorts, and 72 healthy controls. 3.0T, steady-state free precession (SSFP) and modified Look-Locker imaging sequences. All participants had their trabecular complexity quantified by fractal analysis using cine SSFP images. Patients were followed up every 2 weeks until April 2023, or endpoint events happened. Random Forest (RF) and Cox regression models including age, diabetes, LV mass index, mean basal fractal dimension (FD), and left atrial volume index, were developed to predict major adverse cardiac events (MACE). Patients were divided into low- and high-risk groups based on scores derived from the RF model and survival compared. Receiver operating characteristic curve analysis; Kaplan-Meier survival analysis with log rank tests; Harrel's C-index to assess model performance. A P value <0.05 was considered statistically significant. Fifty-five patients (26.57%) experienced MACE during a median follow-up time of 21.83 months. An increased mean basal FD (≥1.324) was associated with a significantly higher risk of MACE. The RF model (C-index: 0.81) had significantly better discrimination than the Cox regression model (C-index: 0.74). Participants of the external validation dataset classified into the high-risk group had a hazard of experiencing MACE increased by 12.29 times compared to those in the low-risk group. LV basal FD was an independent predictor for MACE in patients with ESRD. Reliable risk stratification models could be generated based on LV basal FD and other MRI variables using RF analysis. 2 TECHNICAL EFFICACY: Stage 2.

Prediction of atmospheric PM2.5 level by machine learning techniques in Isfahan, Iran

With increasing levels of air pollution, air quality prediction has attracted more attention. Mathematical models are being developed by researchers to achieve precise predictions. Monitoring and prediction of atmospheric PM2.5 levels, as a predominant pollutant, is essential in emission mitigation programs. In this study, meteorological datasets from 9 years in Isfahan city, a large metropolis of Iran, were applied to predict the PM2.5 levels, using four machine learning algorithms including Artificial Neural |Networks (ANNs), K-Nearest-Neighbors (KNN), Support Vector |Machines (SVMs) and ensembles of classification trees Random Forest (RF). The data from 7 air quality monitoring stations located in Isfahan City were taken into consideration. The Confusion Matrix and Cross-Entropy Loss were used to analyze the performance of classification models. Several parameters, including sensitivity, specificity, accuracy, F1 score, precision, and the area under the curve (AUC), are computed to assess model performance. Finally, by introducing the predicted data for 2020 into ArcGIS software and using the IDW (Inverse Distance Weighting) method, interpolation was conducted for the area of Isfahan city and the pollution map was illustrated for each month of the year. The results showed that, based on the accuracy percentage, the ANN model has a better performance (90.1%) in predicting PM2.5 grades compared to the other models for the applied meteorological dataset, followed by RF (86.1%), SVM (84.6%) and KNN (82.2%) models, respectively. Therefore, ANN modelling provides a feasible procedure for the managerial planning of air pollution control.

Music Genre Classification Using Machine Learning

Music genre classification is a subfield of audio and music analysis in which machine learning and data analysis techniques are used to automatically categorize music tracks into predefined genre categories. In this study, we explore music genre classification using three machine learning algorithms: Support Vector Machine (SVM), Naive Bayes, and k-Nearest Neighbors (k-NN). Our dataset spans diverse music genres, from mainstream to niche, and we employ feature extraction techniques like rhythm-based features. Evaluation metrics, including accuracy, precision, recall, and F1-score, assess model performance. Cross-validation ensures robustness, while addressing imbalanced data is considered. Our findings offer insights into the suitability of SVM, Naive Bayes, and k-NN for music genre classification, providing valuable guidance for audio analysis practitioners. This research sets the stage for further exploration of advanced modeling techniques and real-world challenges in audio classification.

Establishment and evaluation of a nomogram prediction model for the risk of vascular calcification in stage 5 chronic kidney disease patients

Vascular calcification (VC) is a common complication of chronic kidney disease (CKD) that has a detrimental effect on patients' survival and prognosis. The aim of this study was to develop and validate a practical and reliable prediction model for VC in CKD5 patients. The medical records of 544 CKD5 patients were reviewed retrospectively. Multivariate logistic regression analysis was used to identify the independent risk factors for vascular calcification in patients with CKD5 and then created a nomogram prediction model. The area under the receiver operating characteristic curve (AUC), Hosmer–Lemeshow test, and decision curve analysis (DCA) were used to assess model performance. The patients were split into groups with normal and high serum uric acid levels, and the factors influencing these levels were investigated. Age, BUN, SUA, P and TG were independent risk factors for vascular calcification in CKD5 patients in the modeling group (P < 0.05). In the internal validation, the results of model showed that the AUC was 0.917. No significant divergence between the predicted probability of the nomogram and the actual incidence rate (x2 = 5.406, P = 0.753) was revealed by the calibration plot and HL test, thus confirming that the calibration was satisfactory. The external validation also showed good discrimination (AUC = 0.973). The calibration chart and HL test also demonstrated good consistency. Besides, the correlation analysis of serum uric acid levels in all CKD5 patients revealed that elevated uric acid levels may be related to gender, BUN, P, and TG.

Performance of ERA5 wind speed and significant wave height within Extratropical cyclones using collocated satellite radar altimeter measurements

ABSTRACT Similar in strength to hurricanes, Extratropical Cyclones (ECs) are responsible for innavigable sea states, coastal inundation and erosion, and subsequent destruction to coastal infrastructure. Across modern operational wave models, there exists a known systematic underestimation of wave heights during these extreme events. Using a global database of EC storm tracks and 36 years of satellite altimeter data, we examine EC structure and assess model performance through storm centered composite analyses of significant wave height ( H s ) and U10 wind speed ( | U 10 | ). Through the collocation of satellite altimeter observations with a state-of-the-art reanalysis product (ERA5), we investigate model performance with respect to | U 10 | and H s within ECs of varying intensities. By rotating our data reference frames, we align all storm directions to account for asymmetry in EC wind fields and subsequent increased wave growth This rotation results in the organization of the strongest wind speed and H s , as well as trends in the ERA5 performance. A characteristic EC radii is calculated and used to normalize data coordinates in the storm centered reference frame, which highlights the organization of EC structures. Performance metrics are then compared within different EC quadrants to explore the relationship between wind forcing accuracy and underestimation of H s .

A prediction model for coronary artery abnormalities in children with Kawasaki disease older than 5 years

ObjectiveReliably prediction models for coronary artery abnormalities (CAA) in children aged >5 years with Kawasaki disease (KD) are still lacking. This study aimed to develop a nomogram model for predicting CAA at 4 to 8 weeks of illness in children with KD older than 5 years. MethodsA total of 644 eligible children were randomly assigned to a training cohort (n = 450) and a validation cohort (n = 194). The least absolute shrinkage and selection operator (LASSO) analysis was used for optimal predictors selection, and multivariate logistic regression was used to develop a nomogram model based on the selected predictors. Area under the receiver operating characteristic curve (AUC), calibration curves, Hosmer-Lemeshow test, Brier score, and decision curve analysis (DCA) were used to assess model performance. ResultsNeutrophil to lymphocyte ratio, intravenous immunoglobulin resistance, and maximum baseline z-score ≥ 2.5 were identified by LASSO as significant predictors. The model incorporating these variables showed good discrimination and calibration capacities in both training and validation cohorts. The AUC of the training cohort and validation cohort were 0.854 and 0.850, respectively. The DCA confirmed the clinical usefulness of the nomogram model. ConclusionsA novel nomogram model was established to accurately assess the risk of CAA at 4–8 weeks of onset among KD children older than 5 years, which may aid clinical decision-making.