Cross Validation Research Articles

Differential diagnosis between low-risk and high-risk thymoma: Comparison of diagnostic performance of radiologists with and without deep learning model.

There are few CT-based deep learning (DL) studies on thymoma according to the World Health Organization classification. To develop a CT-based DL model to distinguish between low-risk and high-risk thymoma and to compare the diagnostic performance of radiologists with and without the DL model. 159 patients with 160 thymomas were included. A fine-tuning VGG16 network model with Adam optimizer was used, followed by k-fold cross validation. The dataset consisted of three axial slices, including the maximum tumor size from the CT volume data. The data were augmented 50 times by rotation, zoom, shear, and horizontal/vertical flip. Three independent networks for the CT dataset were considered, and the result was determined by voting. Three radiologists independently diagnosed thymomas with and without the model. The area under the curve (AUC) of the diagnostic performance was compared using receiver operating characteristic analysis. Accuracy of the DL model was 71.3%. Diagnostic performance of the radiologists was as follows: AUC and accuracy without the DL model, 0.61-0.68 and 61.9%-69.3%; and with the DL model, 0.66-0.69 and 68.1%-70.0%, respectively. AUC of the diagnostic performance showed no significant differences between radiologists with and without the DL model. The DL model tended to increase the diagnostic accuracy, but AUC was not significantly improved. Diagnostic performance of the DL was comparable to that of radiologists. The DL model assistance tended to increase diagnostic accuracy.

Deep Learning-Based Synthetic Computed Tomography for Low-Field Brain Magnetic Resonance-Guided Radiation Therapy

PurposeMR-guided Radiation Therapy (MRgRT) enables online adaptation to address intra- and inter-fractional changes. To address the need of high-fidelity synthetic CT (synCT) required for dose calculation, we developed a conditional generative adversarial network (cGAN) for synCT generation from low-field MRI in the brain. Methods and MaterialsSimulation MR-CT pairs from twelve glioma patients imaged with a head and neck surface coil and treated on a 0.35T MR-linac were prospectively included to train the model consisting of a 9-block residual network generator and a PatchGAN discriminator. Four-fold cross validation was implemented. SynCT was quantitatively evaluated against real CT using mean absolute error (MAE), Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity (SSIM). Dose was calculated on synCT applying original treatment plan. Dosimetric performance was evaluated by dose-volume histogram (DVH) metric comparison and local three-dimensional gamma analysis. To demonstrate utilization in treatment adaptation, longitudinal synCTs were generated for qualitative evaluation, and one offline adaptation case underwent two comparative plan evaluations. Secondary validation was conducted with 9 patients on a different MR-linac using a high-resolution brain coil. ResultsOur model generated high-quality synCTs with MAE, PSNR and SSIM of 70.9±10.4 HU, 28.4±1.5 d.B. and 0.87±0.02 within the field-of-view, respectively. Underrepresented post-surgical anomalies challenged model performance. Nevertheless, excellent dosimetric agreement was observed with the mean difference between real and synCT DVH metrics of -0.07±0.29 Gy for target D95 and within [-0.14, 0.02] Gy for organs at risk. Significant differences were only observed in the right lens D0.01cc with negligible overall difference (<0.13 Gy). Mean gamma analysis pass rates were 92.2%±3.0%, 99.2%±0.7% and 99.9%±0.1% at 1%/1mm, 2%/2mm and 3%/3mm, respectively. Secondary validation yielded no significant differences in synCT performance for whole brain MAE, PSNR, and SSIM with comparable dosimetric results. ConclusionsOur cGAN model generated high-fidelity brain synCTs from low-field MRI with excellent dosimetric performance. Secondary validation suggests great promise of implementing synCTs to facilitate robust dose calculation for online adaptive brain MRgRT.

Allometric Models for Estimating Above-Ground, Below-Ground and Total Biomass of Tea (&lt;em&gt;Camellia sinensis&lt;/em&gt; (L.) O. Kuntze) Plants Grown in Uva and Up Country, Sri Lanka

This study was conducted to identify the dendrometric variables highly correlating with above-ground (AGB), below-ground (BGB) and total (TB) biomass of tea grown in Uva and Up Country Sri Lanka, develop allometric models to estimate AGB, BGB and TB of tea plants, and to investigate the applicability of already developed allometric models in other countries to Sri Lanka. Twelve experimental plots from TRIORCON field experiment at Tea Research Institute of Sri Lanka, and 18 farmer fields in Uva region, Sri Lanka were selected as the study sites. Representative tea plants were uprooted from the selected sites and seven dendrometric variables were measured. Fresh and dry AGB and BGB of the tea plants were determined. The best allometric models to predict AGB, BGB and TB using dendrometric variables were developed using multiple linear and non-linear regression analyses. Non-linear allometric models fitted best when predicting AGB and TB. Over 64% of total variability of AGB and TB was explained by these models using collar diameter at the soil level. Linear allometric fitted best when predicting BGB. Over 92% of total variability of BGB was explained by linear regression models using collar circumference, main-stem circumference, mean length and circumference at end of the primary branches. Leave-one-out cross validation showed strong positive (AGB and TB) and moderate positive (BGB) correlations between observed and predicted biomass values. Developed linear allometric models in comparison to those in literature showed better fit to data collected under Sri Lankan conditions and precisely estimated AGB, BGB and TB.

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.

Development and validation of a risk prediction model for acute care use among patients with advanced cancer on clinical trials.

277 Background: Patients with advanced cancer on clinical trials are at high risk for unplanned Emergency Room visits (ER_visits) and hospital stays (HS). Identifying patients at highest risk could inform personalized intervention strategies. Methods: The data were from the SWOG Cancer Research Network. Medicare claims data were linked to data from advanced cancer clinical trials. The occurrence of any HS or ER visit within 1 year after enrollment was the primary endpoint. Patients must have had 6 months of continuous Medicare claims prior to registration to determine comorbid conditions prior to treatment, and 12 months of claims after registration (or until death) to detect outcomes. We examined 23 sociodemographic (age, sex, race, ethnicity, insurance status), geographic (rural or urban, area-level deprivation), clinical (disease severity, performance status), treatment (line of therapy, therapy type), and individual comorbidity factors to establish a predictive risk model. Model derivation was conducted in a 60% random sample and tested in the remaining 40%. Best subset selection with logistic regression was used with iterative application of 3-, 4-, and 5-fold cross validation repeated 10 times. Results: Among N=1397 total patients from 6 trials (lung, 3; prostate, 2; pancreas, 1), 33.9% were 75+ years, 20.3% were female, and 8.2% were Black. The overall proportion of patients with ≥1 HS/ER visit was 68.8%. In the training set (n=839), a best 5-variable risk was identified with adverse risk factors including cancer type (prostate vs other), performance status (PS; 0 vs. &gt;=1), coronary artery disease (CAD; yes vs. no), hypertension (yes vs. no), and liver disease (yes vs. no). As cancer type was a study-level variable, we derived a risk model using PS, CAD, hypertension, and liver disease and adjusted for cancer type. There were no differences in adverse risk between those in with and without prostate cancer (interaction p-value=.44), justifying combining across cancer types. In the training set, patients with ≥2 risk factors (high risk; n=303, 36.1%) vs. 0/1 risk factor (low risk; n=536, 63.9%) had nearly a 3-fold increased odds of experiencing HS/ER visit (82.2% vs. 59.1%, OR=2.85, 95% CI, 2.01-4.03, p&lt;.001). Results were successfully validated in the test set (OR=1.93, p=.002). In all patients, those in the highest risk quartile (3/4 factors) vs the lowest risk (0 factors) had nearly a fourfold increased odds of HS/ER visits (OR=3.92, 95% CI, 2.25-6.83, p&lt;.001). Conclusions: Rates of HS_ER visits were high within the first year after enrolling on a trial for advanced cancer. We derived and validated a risk model that delineated strong differences in risk. With expansion of eligibility to include patients with more comorbid conditions, personalized interventions aimed at preventing acute care use could decrease the cost and improve the quality of cancer care.

Visualization analysis of predictive model of acute kidney injury in patients with sepsis by online dynamic nomogram: research on development and validation of application

To explore the risk factors of septic acute kidney injury (sAKI) in patients with sepsis, construct a predictive model for sAKI, verify the predictive value of the model, and develop a dynamic nomogram to help clinical doctors identify patients with high-risk sAKI earlier and more easily. A cross-sectional study was conducted. A total of 245 patients with sepsis admitted to intensive care unit (ICU) of the Affiliated Hospital of Jining Medical University from May 2013 to November 2023 were enrolled as the research subjects. The patients were divided into sAKI group and non-sAKI group based on whether they suffered from sAKI during ICU hospitalization. The differences of the demographic, clinical and laboratory indicators of patients between the two groups were compared. Logistic ordinal regression analysis was performed to analyze the imbalanced variables between the two groups, and to construct a sAKI predictive model. The predictive value of the sAKI predictive model was evaluated through 5-fold cross validation, calibration curve, and decision curve analysis (DCA), and to develop an online dynamic nomogram for the predictive model. A total of 245 patients were enrolled in the final analysis. 110 (44.9%) patients developed sAKI during ICU hospitalization and 135 (55.1%) patients did not develop sAKI. Compared with the non-sAKI group, the patients in the sAKI group had higher ratios of female, hypertension, invasive mechanical ventilation (IMV), renal replacement therapy (RRT), vasopressin usage, and neutrophil count (NEU), aspartate aminotransferase (AST), blood urea nitrogen (BUN), serum creatinine (SCr), uric acid (UA), Na+, K+, procalcitonin (PCT), acute physiology and chronic health evaluation II (APACHE II) score, and sequential organ failure assessment (SOFA) score. Multivariate Logistic ordinal regression analysis showed that female [odd ratio (OR) = 2.208, 95% confidence interval (95%CI) was 1.073-4.323, P = 0.020], hypertension (OR = 2.422, 95%CI was 1.255-5.073, P = 0.012), vasopressin usage (OR = 2.888, 95%CI was 1.380-6.679, P = 0.002), and SCr (OR = 1.015, 95%CI was 1.009-1.024, P < 0.001) were independent risk factors for sAKI in septic patients, and a sAKI predictive model was constructed: ln[P/(1+P)] = -4.665+0.792×female+0.885×hypertension+1.060×vasopressin usage+0.015×SCr. The 5-fold cross validation showed that the average area under the receiver operator characteristic curve (AUC) was 0.860, indicating the sAKI predictive model had a good performance. The calibration curve analysis showed that the calibration degree of the sAKI predictive model was good. DCA showed that the net profit of the sAKI predictive model was relatively high. A static nomogram and an online dynamic nomogram were constructed for the sAKI predictive model. Compared with the static nomogram, the dynamic nomogram allowed for manual selection of corresponding patient characteristics and viewing the corresponding sAKI risk directly. Female, hypertension, vasopressin usage, and SCr are the main risk factors for sAKI in patients with sepsis. The sAKI predictive model constructed based on these factors can help clinical doctors identifying high-risk patients as early as possible, and intervene in a timely manner to provide preventive effects. Compared with the common static nomogram, online dynamic nomogram can make predictive models clearer, more intuitive, and easier.

A Study on the Paleoclimate and Sedimentary Environment of the Upper Kuli Formation in the Hailar Basin

Abstract The Kuli Formation in the Hailar Basin has a wide distribution range, complex and variable lithology, and oil and gas indications during field investigations, which has great research significance and value. In order to better understand the paleoclimate and sedimentary environment of the Upper Kuli Formation, this article uses three methods for the first time: heavy mineral analysis, sporopollen fossils identification, and inorganic element analysis. Based on the characteristics of heavy mineral assemblages, sporopollen fossils identification, and element analysis correspond to the five bottom layer combinations of heavy mineral analysis, analyzing their respective combination characteristics and change patterns. With a more comprehensive perspective and higher time resolution, Cross validation improves the accuracy and credibility of its inference. The results indicate that the Upper Kuli Formation can be divided into five combinations from bottom to top. The paleoclimate underwent a transition from a humid environment to a drier environment. The sedimentary environment has undergone a transition from a deep-water environment to a shallow water environment, transitioning from dynamic, rapid sedimentation, and strong terrestrial influences to a sedimentary environment dominated by more static water, chemical precipitation, and diagenesis. This study combines different geological and geochemical analysis methods to provide strong basis for inferring the paleoclimate and sedimentary environment of the Upper Kuli Formation. The research results are of great significance for interpreting geological changes such as paleoclimate change and ecosystem evolution.

Application Research of Industrial Internet Networked Operation Control Platform Based on RFID and Blockchain Traceability

With the development of Industry 4.0, the demand for data accuracy, transparency, and traceability in the manufacturing industry is increasing day by day. This paper proposes an industrial networking operation management and control platform based on radio frequency identification (RFID) and blockchain technology. Through the deep integration with the national industrial Internet identification analysis system, the platform has realized the full life cycle traceability of products from raw materials to finished products, ensuring the data tamper proof and cross validation of multi-dimensional data. The research results show that after adopting this platform, the accuracy of data collection in manufacturing enterprises is significantly improved, production efficiency is optimized, and operating costs are significantly reduced. In addition, the introduction of blockchain technology effectively prevents data tampering, enhances system transparency and security. This paper provides a new technological path for the digital transformation of the manufacturing industry, and verifies the application potential of RFID and blockchain technology in the industrial Internet environment.

Evaluation of the Impact of Artificial Intelligence on the Systems Audit Process

The paper analyzes the impact of artificial intelligence (AI) in systems auditing, fastening on process optimization through the use of advanced technologies similar as intelligent independent systems. A comprehensive literature review was conducted to understand the operation of AI in checkups, revealing that the integration of these technologies has increased inspection delicacy by over to 93. Specific ways similar as cross-validation (CV), support vector machines (SVMs), and artificial neural networks (ANNs) were employed, demonstrating their effectiveness in perfecting delicacy, receptivity, and particularly in anomaly discovery, with results of 87, 90, and 93 independently. The findings emphasize the need to address the ethical and sequestration pitfalls that accompany the use of AI in checkups, given that while these technologies ameliorate effectiveness and delicacy, they also pose significant challenges in terms of ethical and security data running. In this environment, it's recommended that associations invest in training their staff in the use of AI tools, as well as establish clear programs to insure ethics and sequestration. In addition, it emphasizes the significance of continuing to probe and develop new AI operations that will further ameliorate the effects of system checkups in an ever-changing digital terrain. The perpetration of AI not only optimizes processes but also provides a significant competitive advantage by enabling more accurate discovery of irregularities and patterns in large volumes of data. In summary, AI represents a revolution in the field of systems auditing, offering opportunities to improve accuracy and efficiency, although it is crucial to proactively manage the associated ethical and privacy challenges.

Comparison of Classification Results of SVM, KNN, Decision Tree, and Ensemble Methods in Diabetes Diagnosis

This study aims to determine which algorithms and test techniques are the most optimal in detecting diabetes mellitus and obtaining the best results based on the value of accuracy, precision, and recall. In this study, approaches were used in early diagnosis of diabetes using KNN, SVM, Decision Tree, and Ensemble Majority Voting methods in Percentage Split and K-Fold Cross Validation methods. Diabetes is a disease characterized by high blood sugar (glucose) levels and can cause a variety of disease complications and damage to the body's organs if not treated immediately. Early diagnosis of diabetes is becoming crucial so that people can take immediate action to the hospital for immediate treatment. The data used is Healthcare-Diabetes from Kaggle. The results of this study have found that the K-Fold Cross Validation method is better because it can provide an average improvement in Ensemble accuracy of 13.42% compared to the Percentage Split method which only gives an average increase in Ensamble accuracy of 9.15%. The best algorithm for classifying diabetes disease is the Ensemble Majority Voting algorithm using the K-Fold Cross Validation method with a 98.81% accuracy rate. These excellent research results may contribute to detecting early symptoms of diabetes before it become too severe.

Comparison of machine learning models for the prediction of hypertension in transgender patients undergoing gynecologic surgery.

Transgender patients face a higher burden of cardiovascular morbidity due to structural and biological stressors, particularly in low-resource settings. No studies exist comparing machine learning model development strategies for this unique patient cohort and limited literature exists comparing data/outcomes between transgender and cisgender populations. We compare machine learning models trained solely on transgender patients against models developed on a size-matched and ratio-matched cohort of cisgender patients and a 300-fold larger, ratio-matched cohort of cisgender patients undergoing obstetric/gynecologic procedures in the National Surgical Quality Improvement Program from January 1, 2005 through December 31, 2019. All models were developed to predict the outcome of hypertension. Statistical significance between models was calculated using 5-by-2 fold cross validation hypothesis testing. Among 626,102 patients having an obstetric/gynecologic surgery, there are 1959 transgender patients of which 85,405 (13.7%) have hypertension requiring medication. Saliently, the logistic regression machine learning models trained selectively on the transgender cohort have an AUC of 0.865 (95% CI: 0.83-0.90), with an accuracy of 85% (95% CI: 0.80-0.87) compared to (p < 0.05) the logistic regression model trained on the 300-fold larger combined cohort which has an AUC of 0.861 (95% CI: 0.82-0.90), with an accuracy of 83% (95% CI: 0.80-0.87). Machine learning models can be trained on smaller, selectively transgender populations and may perform similarly or better to predict cardiovascular outcomes in transgender patients, than models developed on predominantly cisgender patients; this can be useful in lower-resource settings with smaller-volume transgender patients.

Optimizing Random Forest for Predicting Thoracic Surgery Success in Lung Cancer Using Recursive Feature Elimination and GridSearchCV

Abstract. Lung cancer is one of the deadliest forms of cancer, claiming numerous lives annually. Thoracic surgery is a strategy to manage lung cancer patients; however, it poses high risks, including potential nerve damage and fatal complications leading to mortality. Predicting the success rate of thoracic surgery for lung cancer patients can be accomplished using data mining techniques based on classification principles. Medical data mining involves employing mathematical, statistical, and computational methods. In this study, the prediction of thoracic surgery success employs the random forest algorithm with recursive feature elimination for feature selection. The feature selection process yields the top 8 features. The 8 best features include 'DGN', 'PRE4', 'PRE5', 'PRE6', 'PRE10', 'PRE14', 'PRE30', and 'AGE'. Hyperparameter using GridSearchCV is then applied to enhance classification accuracy. The results of this method implementation demonstrate a predictive accuracy of 91.41%. Purpose: The study aims to develop and evaluate a Random Forest model with a Recursive Feature Elimination feature selection and applies hyperparameter GridSearchCV for predicting thoracic surgery success rate. Methods: This study uses the thoracic surgery dataset and applies various data preprocessing techniques. The dataset is then used for classification using the Random Forest algorithm and applies the Recursive Feature Elimination feature selection to obtain the best features. GridSearchCV is used in this study for hyperparameter. Result: The accuracy using the Random Forest algorithm and Recursive Feature Elimination feature selection with hyperparameters tuning GridSearchCV resulted in an accuracy of 91,41%. The accuracy was obtained from the following parameters values: bootstrap set to false, criterion set to gini, n_estimator equal to 100, max_depth set to none, min_samples_split equal to 4, min_samples_leaf equal to 1, max_features set to auto, n_jobs set to -1, and verbose set to 2 with 10-fold cross validation. Novelty: This study comparison and analysis of various dataset preprocessing methods and different model configurations are conducted to find the best model for predicting the success rate of thoracic surgery. The study also employs feature selection to choose the best feature to be used in classification process, as well as hyperparameter tuning to achieve optimal accuracy and discover the optimal values for these hyperparameters.

Klasifikasi Predikat Kelulusan Mahasiswa Menggunakan Algoritma C4.5

Determination of student graduation predicate based on GPA and additional requirements of timely study period for honors predicate. Currently, there is no in-depth classification to identify graduation predicate, so understanding of the patterns that influence the results is still limited. Accumulation of student graduation data can be used to find new information. This study aims to produce a decision tree classification model using the C4.5 algorithm and evaluate its accuracy in classifying student graduation predicates at UIN Raden Fatah Palembang. The data division technique used is k-fold cross validation to divide the data into training and testing data. The k value used is k = 3 in the first data test, this is based on previous tests with several k values, where k = 3 produces higher accuracy than the others. The rules formed are 242 and the attribute that influences student graduation predicate is GPA. The accuracy of the application of the C4.5 algorithm in classifying student graduation predicates is 83.31% which is included in the Good Classification category.

Comparing modeled predictions of coral reef diversity along a latitudinal gradient in Mozambique

IntroductionPredictive models based on environmental proxy data are being used to predict biodiversity on large and even global scales. Yet, some of the underlying assumptions about the relationship between proxy variables and predictions require investigations and testing the consequences of using model alternatives, data sources, variables choices, and scales, extent, and overlap among the predictions. Mozambican coral reefs provide a good case study to test these assumptions given the paucity of field data, its long coastline, and transitions from tropical to temperate environments.MethodsThree modelling formulations and 5 specific models were made using satellite and shipboard measurements and extensive fish and corals field data to test their performance in predicting numbers of fish species and coral taxa from field data. Model predictions were mapped for the 1180 ~6.25 km2 Mozambican coral reef cells. Predictions were made and mapped 1) based on ~1000 field sites in the Western Indian Ocean (WIO) faunal province model, 2) using environmental variable selected in the WIO model (WIOMOD) but trained only with Mozambican field data (&amp;lt;113 sites), and 3) using only Mozambican environmental and field data and standard variable redundancy and selection procedures.Results and discussionTraining and testing cross validation of models indicated modest predictive ability (R2~0.42-0.56%) and reasonable transferability. Consequently, there was unexplained variation likely due to small-scale environmental variability finer than the mapped cell scale. Differences between model predictions were caused by different variable rankings and response relationship. For example, the Mozambique-only model predicted more fish but fewer coral taxa, a larger role of water quality and sediments, habitats, and temperature variation, and a lesser role of human influence than the WIOMOD. Therefore, differences between models indicate that large scale models (i.e. provincial or global) can contribute to understanding gross patterns but miss important local environmental and human drivers in transitional environments. Nevertheless, 79% of the fish and 88% of coral taxa cell-level predictions of taxonomic diversity had standardized coefficients of variations of &amp;lt;10%.

Hyperspectral Prediction Models of Chlorophyll Content in Paulownia Leaves under Drought Stress.

This study explored the quantitative inversion of the chlorophyll content in Paulownia seedling leaves under drought stress and analyzed the factors influencing the chlorophyll content from multiple perspectives to obtain the optimal model. Paulownia seedlings were selected as the experimental materials for the potted water control experiments. Four drought stress treatments were set up to obtain four types of Paulownia seedlings: one pair of top leaves (T1), two pairs of leaves (T2), three pairs of leaves (T3), and four pairs of leaves (T4). In total, 23 spectral transformations were selected, and the following four methods were adopted to construct the prediction model, select the best spectral preprocessing method, and explore the influence of water bands: partial least squares modeling with all spectral bands (all-band partial least squares, AB-PLS), principal component analysis partial least squares (PCA-PLS), correlation analysis partial least squares (CA-PLS), correlation analysis (water band) partial least squares, ([CA(W)-PLS]), and vegetation index modeling. Based on the prediction accuracy and the uniformity of different leaf positions, the optimal model was systematically explored. The results of the analysis of spectral reflectance showed significant differences at different leaf positions. The sensitive bands of chlorophyll were located near 550 nm, whereas the sensitive bands of water were located near 1440 and 1920 nm. The results of the vegetation index models indicate that the multiple-index models outperformed the single-index models. Accuracy decreased as the number of indicators decreased. We found that different model construction methods matched different optimal spectral preprocessing methods. First derivative spectra (R') was the best preprocessing method for the AB-PLS, PCA-PLS, and CA-PLS models, whereas the inverse log spectra (log(1/R)) was the best preprocessing method for the CA(W)-PLS model. Among the 14 indices, the green normalized difference vegetation index (GNDVI) was most correlated with the chlorophyll content sensitivity indices, and the water index (WI) was most correlated with the water sensitive indices. At the same time, the water band affected the cross validation accuracy. When characteristic bands were used for modeling, the cross validation accuracy was significantly increased. In contrast, when vegetation indices were used for modeling, the accuracy of the cross validation increased slightly but its predictive ability was reduced; thus, these changes could be ignored. We found that leaf position also affected the prediction accuracy, with the first pair of top leaves exhibiting the worst predictive ability. This was a bottleneck that limited predictive capability. Finally, we found that the CA(W)-PLS model was optimal. The model was based on 23 spectral transformations, four PLS construction methods, water bands, and different leaf positions to ensure systematicity, stability, and applicability.

Risk prediction model for pneumothorax or pleural effusion after microwave ablation in patients with lung malignancy

BackgroundAlthough microwave ablation (MWA) has been shown to be an effective treatment for lung malignancies (LM), there is no effective way to predict pneumothorax or pleural effusion after MWA so that timely measures can be taken to prevent it. MethodsThis study comprised LM patients undergoing MWA at Affiliated Hospital of Nantong University from January 2013 to September 2023. Patients before May 2023 constituted the training set (n = 340), while data from May to September served as the test set (n = 58). Unformatted and formatted data extracted from electronic medical records (EMR) were utilized for model construction. Predictors for pneumothorax or pleural effusion were determined through univariate analysis and backward stepwise regression in the training set. Six ML algorithms were employed to create four models based on the research timeframe. Evaluation of the four models was performed using receiver operating characteristic (ROC) analysis, area under the ROC curve (AUC), and 10-fold cross validation. FindingsA total of 398 patients (216 aged 70 or above, 271 males) were included, with 23.37 % (93/398) experiencing pneumothorax and 33.42 % (133/398) developing pleural effusion. Across all four predictive models, Logistic Regression (LR) demonstrated optimal predictive performance in the test set, with AUC values of 0.727 for Model Ⅰ, 0.876 for Model Ⅱ, 0.895 for Model Ⅲ, and 0.807 for Model Ⅳ. InterpretationML models effectively predict post-MWA pneumothorax or pleural effusion.

Development of decision tree classification algorithms in predicting mortality of COVID-19 patients

IntroductionThe accurate prediction of COVID-19 mortality risk, considering influencing factors, is crucial in guiding effective public policies to alleviate the strain on the healthcare system. As such, this study aimed to assess the efficacy of decision tree algorithms (CART, C5.0, and CHAID) in predicting COVID-19 mortality risk and compare their performance with that of the logistic model.MethodsThis retrospective cohort study examined 5080 cases of COVID-19 in Babol, a city in northern Iran, who tested positive for the virus via PCR from March 2020 to March 2022. In order to check the validity of the findings, the data was randomly divided into an 80% training set and a 20% testing set. The prediction models, such as Logistic regression models and decision tree algorithms, were trained on the 80% training data and tested on the 20% testing data. The accuracy of these methods for the test samples was assessed using measures like ROC curve, sensitivity, specificity, and AUC.ResultsThe findings revealed that the mortality rate for COVID-19 patients who were admitted to hospitals was 7.7%. Through cross validation, it was determined that the CHAID algorithm outperformed other decision tree and logistic regression algorithms in specificity, and precision but not sensitivity in predicting the risk of COVID-19 mortality. The CHAID algorithm demonstrated a specificity, precision, accuracy, and F-score of 0.98, 0.70, 0.95, and 0.52 respectively. All models indicated that factors such as ICU hospitalization, intubation, age, kidney disease, BUN, CRP, WBC, NLR, O2 sat, and hemoglobin were among the factors that influenced the mortality rate of COVID-19 patients.ConclusionsThe CART and C5.0 models had outperformed in sensitivity but CHAID demonstrates a better performance compared to other decision tree algorithms in specificity, precision, accuracy and shows a slight improvement over the logistic regression method in predicting the risk of COVID-19 mortality in the population under study.

A Practical In Silico Method for Predicting Compound Brain Concentration-Time Profiles: Combination of PK Modeling and Machine Learning.

Given the aging populations in advanced countries globally, many pharmaceutical companies have focused on developing central nervous system (CNS) drugs. However, due to the blood-brain barrier, drugs do not easily reach the target area in the brain. Although conventional screening methods for drug discovery involve the measurement of (unbound fraction of drug) brain-to-plasma partition coefficients, it is difficult to consider nonequilibrium between plasma and brain compound concentration-time profiles. To truly understand the pharmacokinetics/pharmacodynamics of CNS drugs, compound concentration-time profiles in the brain are necessary; however, such analyses are costly and time-consuming and require a significant number of animals. Therefore, in this study, we attempted to develop an in silico prediction method that does not require a large amount of experimental data by combining modeling and simulation (M&S) with machine learning (ML). First, we constructed a hybrid model linking plasma concentration-time profile to the brain compartment that takes into account the transit time and brain distribution of each compound. Using mouse plasma and brain time experimental values for 103 compounds, we determined the brain kinetic parameters of the hybrid model for each compound; this case was defined as scenario I (a positive control experiment) and included the full brain concentration-time profile data. Next, we built an ML model using chemical structure descriptors as explanatory variables and rate parameters as the target variable, and we then input the predicted values from 5-fold cross-validation (CV) into the hybrid model; this case was defined as scenario II, in which no brain compound concentration-time profile data exist. Finally, for scenario III, assuming that the brain concentration is obtained at only one time point, we used the brain kinetic parameters from the result of the 5-fold CV in scenario II as the initial values for the hybrid model and performed parameter refitting against the observed brain concentration at that time point. As a result, the RMSE/R2-values of the brain compound concentration-time profiles over time were 0.445/0.517 in scenario II and 0.246/0.805 in scenario III, indicating the method provides high accuracy and suggesting that it is a practical method for predicting brain compound concentration-time profiles.

Conventional Machine Learning and Ensemble Learning Techniques in Cardiovascular Disease Prediction and Analysis

Cardiovascular diseases, which significantly affect the heart and blood vessels, are one of the leading causes of death worldwide. Early diagnosis and treatment of these diseases, which cause approximately 19.1 million deaths, are essential. Many problems, such as coronary artery disease, blood vessel disease, irregular heartbeat, heart muscle disease, heart valve problems, and congenital heart defects, are included in this disease definition. Today, researchers in the field of cardiovascular disease are using approaches based on diagnosis-oriented machine learning. In this study, feature extraction is performed for the detection of cardiovascular disease, and classification processes are performed with a Support Vector Machine, Naive Bayes, Decision Tree, K-Nearest Neighbor, Bagging Classifier, Random Forest, Gradient Boosting, Logistic Regression, AdaBoost, Linear Discriminant Analysis and Artificial Neural Networks methods. A total of 918 observations from Cleveland, Hungarian Institute of Cardiology, University Hospitals of Switzerland, and Zurich, VA Medical Center were included in the study. Principal Component Analysis, a dimensionality reduction method, was used to reduce the number of features in the dataset. In the experimental findings, feature increase with artificial variables was also performed and used in the classifiers in addition to feature reduction. Support Vector Machines, Decision Trees, Grid Search Cross Validation, and existing various Bagging and Boosting techniques have been used to improve algorithm performance in disease classification. Gaussian Naïve Bayes was the highest-performing algorithm among the compared methods, with 91.0% accuracy on a weighted average basis as a result of a 3.0% improvement.

RanBALL: An Ensemble Random Projection Model for Identifying Subtypes of B-cell Acute Lymphoblastic Leukemia.

As the most common pediatric malignancy, B-cell acute lymphoblastic leukemia (B-ALL) has multiple distinct subtypes characterized by recurrent and sporadic somatic and germline genetic alterations. Identification of B-ALL subtypes can facilitate risk stratification and enable tailored therapeutic approaches. Existing methods for B-ALL subtyping primarily depend on immunophenotypic, cytogenetic and genomic analyses, which would be costly, complicated, and laborious in clinical practice applications. To overcome these challenges, we present RanBALL (an Ensemble Random Projection-Based Model for Identifying B-Cell Acute Lymphoblastic Leukemia Subtypes), an accurate and cost-effective model for B-ALL subtype identification based on transcriptomic profiling only. RanBALL leverages random projection (RP) to construct an ensemble of dimension-reduced multi-class support vector machine (SVM) classifiers for B-ALL subtyping. Results based on 100 times 5-fold cross validation tests for >1700 B-ALL patients demonstrated that the proposed model achieved an accuracy of 93.35%, indicating promising prediction capabilities of RanBALL for B-ALL subtyping. The high accuracies of RanBALL suggested that our model could effectively capture underlying patterns of transcriptomic profiling for accurate B-ALL subtype identification. We believe RanBALL will facilitate the discovery of B-ALL subtype-specific marker genes and therapeutic targets, and eventually have consequential positive impacts on downstream risk stratification and tailored treatment design.