ROC AUC Research Articles

Abstract PS05-08: Restriction Spectrum Imaging MRI for automated evaluation of response to neoadjuvant therapy in breast cancer

Abstract Introduction: Dynamic contrast-enhanced MRI (DCE) is currently used to evaluate neoadjuvant therapy response of breast cancer (1). However, DCE requires expert radiologist readers to assess the change in longest tumor dimension during therapy, as well as administration of Gadolinium contrast agents. One MRI modality that does not require contrast agents is diffusion-weighted MRI (DWI), a method that detects the microscopic diffusion of water molecules. However, the commonly used DWI method apparent diffusion coefficient (ADC) is not fully optimised in the breast (2). The purpose of the current study was to evaluate the recent DWI method Restriction Spectrum Imaging (RSI) (2) to automatically monitor breast tumor size during neoadjuvant therapy. Methods: Twenty-seven women underwent 3T MRI at four time points during therapy at University of California San Diego; 17 received all four scans (see Table 1 for patient details). Inclusion criteria included biopsy-proven unilateral invasive breast cancer ≥2.5 cm (defined on clinical examination/imaging) with indication for neoadjuvant therapy. The therapy used was primarily paclitaxel (+/-experimental agent) followed by anthracycline. The MRI protocol included Gadolinium DCE and DWI (b-values 0, 500, 1500, 4000 s/mm2); TE/TR = 82/9000 ms. ADC was calculated using b-values < 1000 s/mm2 while signal from all available b-values were fitted to the previously-developed three-component RSI model (2). The tumor size by RSI was assessed against manual DCE tumor size and mean ADC values. Prediction of therapy response during therapy and residual tumor post-therapy were assessed using non-pathological complete response (non-pCR) as endpoint. pCR was defined as ypT0/is or ypN0. Results: Ten patients experienced pCR. Prediction of non-pCR by ROC AUC at the early-therapy time point was 0.65 for RSI, 0.64 for DCE and 0.45 for ADC (Table 2). Prediction of post-therapy residual tumor is given in Table 3. Discussion: The novel RSI cancer tissue classifier predicted response to neoadjuvant therapy after only 19 days. RSI could also identify 71% of cases with residual tumor at surgery with 90% specificity post-therapy. RSI performance was similar to performance by standard MRI by manual tumor measurement on DCE. In contrast to standard-of-care MRI by using DCE and ADC that requires manual user input, the RSI classifier is automatic. This suggests that RSI may aid to cost-efficiently evaluate neoadjuvant therapy of breast cancer, with the aim to help guide clinical decision-making and enable tailored therapy regimens.

Abstract 1020: Comparative Assessment Of Hscrp And Apolipoprotein B As Ascvd Risk Biomarkers

According to the American Heart Association, the accumulation of plaque in the walls of arteries is identified as the primary cause of atherosclerotic cardiovascular disease (ASCVD). Currently, ASCVD-related conditions remain the leading cause of morbidity and mortality globally. Apolipoprotein B has been identified as a more precise cardiovascular risk marker than LDL-C, while hsCRP has shown potential as a cardiovascular disease indicator. This study aims to investigate the diagnostic performance and routine screening cut-off of hsCRP for early atherosclerosis vascular diseases (ASCVD) risk in adult patients, comparing it with Apo B. Objective: To compare the diagnostic performance of high-sensitivity C-reactive protein (hsCRP) with apolipoprotein B in assessing ASCVD risk. Method: A sample of 494 individuals from the NHANES 2015-2016 laboratory dataset, with a mean age greater than 17 years, was used for this study. ASCVD risk was measured by non-HDL-C, categorized into low and high risk based on the Mayo Clinic reference range. Predictors included Apo B, and hs-CRP. Binomial logistic regression and ROC curve analyses were conducted using the generalised linear models and pROC packages in RStudio IDE. Hypotheses were validated at p≤0.05, and diagnostic performance metrics such as ROC AUC, sensitivity, and specificity were measured on a scale of 0-1. Result: The findings revealed that for every 1g/L increase in apo B concentration, the odds of high ASCVD risk were approximately 3.8 х1011 times higher. Additionally, the model indicated that the odds of high ASCVD risk were 1.03 times higher for every 1mg/L increase in hsCRP concentration. However, this indicate that hsCRP level was not associated with odds of ASCVD risk. The ROC AUC for apo B and hsCRP were approximately 0.9739 and 0.6165, respectively, with cut-off values (sensitivity, specificity) of approximately 0.9g/L (0.927, 0.897) and 2.4 mg/L (0.596, 0.601), respectively. Thus, levels above these thresholds for both apo B and hsCRP are associated with high ASCVD risk. Conclusion: The study demonstrates that apo B exhibits high discriminatory and diagnostic accuracy, making it a suitable ASCVD risk biomarker compared to hsCRP. While hsCRP shows moderate diagnostic accuracy, it is not sufficient as a standalone ASCVD risk diagnostic marker. Therefore, apo B could serve as a replacement for LDL-C, while hsCRP could possibly serve as an add-on test in ASCVD risk assessment.

Identifying bias in models that detect vocal fold paralysis from audio recordings using explainable machine learning and clinician ratings.

Detecting voice disorders from voice recordings could allow for frequent, remote, and low-cost screening before costly clinical visits and a more invasive laryngoscopy examination. Our goals were to detect unilateral vocal fold paralysis (UVFP) from voice recordings using machine learning, to identify which acoustic variables were important for prediction to increase trust, and to determine model performance relative to clinician performance. Patients with confirmed UVFP through endoscopic examination (N = 77) and controls with normal voices matched for age and sex (N = 77) were included. Voice samples were elicited by reading the Rainbow Passage and sustaining phonation of the vowel "a". Four machine learning models of differing complexity were used. SHapley Additive exPlanations (SHAP) was used to identify important features. The highest median bootstrapped ROC AUC score was 0.87 and beat clinician's performance (range: 0.74-0.81) based on the recordings. Recording durations were different between UVFP recordings and controls due to how that data was originally processed when storing, which we can show can classify both groups. And counterintuitively, many UVFP recordings had higher intensity than controls, when UVFP patients tend to have weaker voices, revealing a dataset-specific bias which we mitigate in an additional analysis. We demonstrate that recording biases in audio duration and intensity created dataset-specific differences between patients and controls, which models used to improve classification. Furthermore, clinician's ratings provide further evidence that patients were over-projecting their voices and being recorded at a higher amplitude signal than controls. Interestingly, after matching audio duration and removing variables associated with intensity in order to mitigate the biases, the models were able to achieve a similar high performance. We provide a set of recommendations to avoid bias when building and evaluating machine learning models for screening in laryngology.

Application of BukaGini algorithm for enhanced feature interaction analysis in intrusion detection systems

This article presents an evaluation of BukaGini, a stability-aware Gini index feature selection algorithm designed to enhance model performance in machine learning applications. Specifically, the study focuses on assessing BukaGini’s effectiveness within the domain of intrusion detection systems (IDS). Recognizing the need for improved feature interaction analysis methodologies in IDS, this research aims to investigate the performance of BukaGini in this context. BukaGini’s performance is evaluated across four diverse datasets commonly used in IDS research: NSLKDD (22,544 samples), WUSTL EHMS (16,318 samples), WSN-DS (374,661 samples), and UNSWNB15 (175,341 samples), amounting to a total of 588,864 data samples. The evaluation encompasses key metrics such as stability score, accuracy, F1-score, recall, precision, and ROC AUC. Results indicate significant advancements in IDS performance, with BukaGini achieving remarkable accuracy rates of up to 99% and stability scores consistently surpassing 99% across all datasets. Additionally, BukaGini demonstrates an average reduction in dimensionality of 25%, selecting 10 features for each dataset using the Gini index. Through rigorous comparative analysis with existing methodologies, BukaGini emerges as a promising solution for feature interaction analysis within cybersecurity applications, particularly in the context of IDS. These findings highlight the potential of BukaGini to contribute to robust model performance and propel intrusion detection capabilities to new heights in real-world scenarios.

Identification of prognostic signatures in remnant gastric cancer through an interpretable risk model based on machine learning: a multicenter cohort study

ObjectiveThe purpose of this study was to develop an individual survival prediction model based on multiple machine learning (ML) algorithms to predict survival probability for remnant gastric cancer (RGC).MethodsClinicopathologic data of 286 patients with RGC undergoing operation (radical resection and palliative resection) from a multi-institution database were enrolled and analyzed retrospectively. These individuals were split into training (80%) and test cohort (20%) by using random allocation. Nine commonly used ML methods were employed to construct survival prediction models. Algorithm performance was estimated by analyzing accuracy, precision, recall, F1-score, area under the receiver operating characteristic curve (AUC), confusion matrices, five-fold cross-validation, decision curve analysis (DCA), and calibration curve. The best model was selected through appropriate verification and validation and was suitably explained by the SHapley Additive exPlanations (SHAP) approach.ResultsCompared with the traditional methods, the RGC survival prediction models employing ML exhibited good performance. Except for the decision tree model, all other models performed well, with a mean ROC AUC above 0.7. The DCA findings suggest that the developed models have the potential to enhance clinical decision-making processes, thereby improving patient outcomes. The calibration curve reveals that all models except the decision tree model displayed commendable predictive performance. Through CatBoost-based modeling and SHAP analysis, the five-year survival probability is significantly influenced by several factors: the lymph node ratio (LNR), T stage, tumor size, resection margins, perineural invasion, and distant metastasis.ConclusionsThis study established predictive models for survival probability at five years in RGC patients based on ML algorithms which showed high accuracy and applicative value.

Application of Decision Tree Algorithm in Teaching and Learning Management in Colleges and Universities

In modern educational settings, effective decision-making plays a crucial role in ensuring student success and fostering academic excellence. The decision tree algorithm is a valuable tool in teaching and learning management in colleges and universities, offering insights that inform strategic decision-making processes. By analyzing student data, including academic performance, engagement metrics, and demographic information, decision tree algorithms can identify patterns and trends that help optimize educational practices. These algorithms aid in course planning, student advising, and resource allocation by predicting student outcomes, identifying at-risk students, and recommending personalized interventions. This paper explores the application and impact of the Decision Tree Wrapper Sampling Class Imbalance Knowledge Assimilation (DT-WSCIKA) method in enhancing educational decision-making processes. Through a comprehensive analysis of teaching and learning aspects, classification performance metrics, and student performance outcomes across various scenarios, we investigate the efficacy of DT-WSCIKA in addressing class imbalances, optimizing decision-making, and improving educational outcomes. The findings reveal significant improvements in key metrics following the implementation of DT-WSCIKA. For example, classification accuracy increased by an average of 8%, precision improved by 10%, recall by 6%, F1 Score by 7%, and ROC AUC by 9% over multiple epochs. Additionally, student performance outcomes showed an average increase of 7.5% in Scenario 1, 6.2% in Scenario 3, and 8.3% in Scenario 7 when DT-WSCIKA was employed.

An exploration on the machine-learning-based stroke prediction model.

With the rapid development of artificial intelligence technology, machine learning algorithms have been widely applied at various stages of stroke diagnosis, treatment, and prognosis, demonstrating significant potential. A correlation between stroke and cytokine levels in the human body has recently been reported. Our study aimed to establish machine-learning models based on cytokine features to enhance the decision-making capabilities of clinical physicians. This study recruited 2346 stroke patients and 2128 healthy control subjects from Chongqing University Central Hospital. A predictive model was established through clinical experiments and collection of clinical laboratory tests and demographic variables at admission. Three classification algorithms, namely Random Forest, Gradient Boosting, and Support Vector Machine, were employed. The models were evaluated using methods such as ROC curves, AUC values, and calibration curves. Through univariate feature selection, we selected 14 features and constructed three machine-learning models: Support Vector Machine (SVM), Random Forest (RF), and Gradient Boosting Machine (GBM). Our results indicated that in the training set, the RF model outperformed the GBM and SVM models in terms of both the AUC value and sensitivity. We ranked the features using the RF algorithm, and the results showed that IL-6, IL-5, IL-10, and IL-2 had high importance scores and ranked at the top. In the test set, the stroke model demonstrated a good generalization ability, as evidenced by the ROC curve, confusion matrix, and calibration curve, confirming its reliability as a predictive model for stroke. We focused on utilizing cytokines as features to establish stroke prediction models. Analyses of the ROC curve, confusion matrix, and calibration curve of the test set demonstrated that our models exhibited a strong generalization ability, which could be applied in stroke prediction.

Predictive modeling of initiation and delayed mental health contact for depression

BackgroundDepression is prevalent among Operation Enduring Freedom and Operation Iraqi Freedom (OEF/OIF) Veterans, yet rates of Veteran mental health care utilization remain modest. The current study examined: factors in electronic health records (EHR) associated with lack of treatment initiation and treatment delay; the accuracy of regression and machine learning models to predict initiation of treatment.MethodsWe obtained data from the VA Corporate Data Warehouse (CDW). EHR data were extracted for 127,423 Veterans who deployed to Iraq/Afghanistan after 9/11 with a positive depression screen and a first depression diagnosis between 2001 and 2021. We also obtained 12-month pre-diagnosis and post-diagnosis patient data. Retrospective cohort analysis was employed to test if predictors can reliably differentiate patients who initiated, delayed, or received no mental health treatment associated with their depression diagnosis.Results108,457 Veterans with depression, initiated depression-related care (55,492 Veterans delayed treatment beyond one month). Those who were male, without VA disability benefits, with a mild depression diagnosis, and had a history of psychotherapy were less likely to initiate treatment. Among those who initiated care, those with single and mild depression episodes at baseline, with either PTSD or who lacked comorbidities were more likely to delay treatment for depression. A history of mental health treatment, of an anxiety disorder, and a positive depression screen were each related to faster treatment initiation. Classification of patients was modest (ROC AUC = 0.59 95%CI = 0.586–0.602; machine learning F-measure = 0.46).ConclusionsHaving VA disability benefits was the strongest predictor of treatment initiation after a depression diagnosis and a history of mental health treatment was the strongest predictor of delayed initiation of treatment. The complexity of the relationship between VA benefits and history of mental health care with treatment initiation after a depression diagnosis is further discussed. Modest classification accuracy with currently known predictors suggests the need to identify additional predictors of successful depression management.

Enhancing Diabetes Prediction Accuracy through Hybrid Machine Learning Models: A Comparative Study

This study investigates the effectiveness of various machine learning (ML) models in predicting the onset of diabetes, emphasizing the superior performance of hybrid models over single learner models. Employing a dataset comprising 10,000 individuals with features like Glucose level, BMI, Insulin, and more, we meticulously processed and engineered the data to optimize it for ML applications. We developed several models, including Decision Trees, Random Forest, KNN, and XGBoost, and then advanced to hybrid models using ensemble techniques like stacking and soft voting classifiers. Our findings indicate that hybrid models significantly outperform single learner models. These models achieved remarkable accuracy (98.11%), precision (97.31%), and ROC AUC (99.82%), highlighting their potential in clinical settings. The study underscores the value of hybrid ML models in enhancing predictive accuracy and reliability in diabetes diagnostics.

Subsidence vulnerability indexing using convolutional neural networks based on clustering and regression modeling strategies

Land subsidence triggered by groundwater over-abstraction is a topical research activity, and this study investigates the Subsidence Vulnerability Index (SVI) using Convolutional Neural Networks (CNNs). The study predicts SVI using the clustering and regression modeling strategies, which have identical aims of decreasing the inherent subjectivity within the basic ALPRIFT but distinct formulations. Both strategies incorporated ALPRIFT data layers, referring to hydrogeological and physical settings of a study area. Also, subsidence maps were incorporated in CNN formulations using 15 SAR (ALOS-2 PALSAR-2 Synthetic Aperture Radar) images. The formulations were implemented in the Tabriz aquifer, northwest of Iran, suffering a severe decline in the water table. The results provide evidence that both CNNs increase the accuracy compared to the basic ALPRIFT as per the ROC curve and AUC values. The Clustering-CNN has slightly better performance (AUC = 0.87) compared to the Regression-CNN (AUC = 0.83). Also, both CNNs have approximately identical spatial patterns and hotspots. The trained and tested CNNs on pixels with high-reliability of InSAR (Interferometric Synthetic Aperture Radar) results show their capability in predicting SVI on pixels with low-reliability.

Comparative Analysis of Machine Learning Models for Predicting Diabetes: Unveiling the Superiority of Advanced Ensemble Methods

This study presents a comprehensive analysis of various machine learning models to predict diabetes. The research evaluates and compares the predictive performance of advanced ensemble techniques Extra Trees Classifier and LightGBM with traditional machine learning algorithms and simpler deep neural network (DNN) architectures. The dataset comprises numerous features pertinent to diabetes diagnosis, such as glucose concentration, BMI, and insulin levels, among others. A methodology, including polynomial feature transformation and ten-fold cross-validation, was employed to ensure the study's reliability and the models' capability to generalize. The advanced ensemble models, Extra Trees and LightGBM, achieved stellar predictive metrics, with the former attaining a near-perfect ROC AUC, accuracy, precision, and an F1 score close to 1. LightGBM followed closely, demonstrating the high efficacy of ensemble methods in complex data settings. These results were contrasted with significantly lower performance metrics from DNNs and respectable, albeit lower, scores from traditional models like Decision Trees, Random Forest, KNN, and XGBoost.

Meta-Feature-Based Traffic Accident Risk Prediction: A Novel Approach to Forecasting Severity and Incidence

This study aims to improve the accuracy of predicting the severity of traffic accidents by developing an innovative traffic accident risk prediction model—StackTrafficRiskPrediction. The model combines multidimensional data analysis including environmental factors, human factors, roadway characteristics, and accident-related meta-features. In the model comparison, the StackTrafficRiskPrediction model achieves an accuracy of 0.9613, 0.9069, and 0.7508 in predicting fatal, serious, and minor accidents, respectively, which significantly outperforms the traditional logistic regression model. In the experimental part, we analyzed the severity of traffic accidents under different age groups of drivers, driving experience, road conditions, light and weather conditions. The results showed that drivers between 31 and 50 years of age with 2 to 5 years of driving experience were more likely to be involved in serious crashes. In addition, it was found that drivers tend to adopt a more cautious driving style in poor road and weather conditions, which increases the margin of safety. In terms of model evaluation, the StackTrafficRiskPrediction model performs best in terms of accuracy, recall, and ROC–AUC values, but performs poorly in predicting small-sample categories. Our study also revealed limitations of the current methodology, such as the sample imbalance problem and the limitations of environmental and human factors in the study. Future research can overcome these limitations by collecting more diverse data, exploring a wider range of influencing factors, and applying more advanced data analysis techniques.

OA03 Developing a multivariable prediction model to identify anti-CCP positive individuals among those with non-specific musculoskeletal symptoms in primary care

Abstract Background/Aims A phase of immunologically imminent rheumatoid arthritis (RA) precedes the onset of clinical RA, characterised by increased circulating anti-cyclic citrullinated peptide (anti-CCP) antibodies. Early identification in the pre-clinical phase of RA in primary care (PC), through targeted anti-CCP testing, could prompt secondary care referral for earlier diagnosis, achieving better long-term outcomes. This study aimed to develop a model identifying those likely to be anti-CCP positive amongst people presenting to PC with non-specific musculoskeletal symptoms. Methods Participants were recruited across the UK between 2014 and 2020 to the Leeds ‘Co-ordinated Programme to Prevent Arthritis’, a prospective, observational cohort study. Participants were aged ≥18 years who presented to PC with non-specific musculoskeletal symptoms and no history of clinical synovitis. Fifteen baseline predictors were considered: age; gender; smoking status; first degree relative (FDR) with RA; and patient reported pain in: neck; back; shoulders; elbows; hips; wrists; thumbs; hand and/or fingers; knees; ankles; foot and/or toes. Participants were followed up at 12 months to determine if they had developed RA. Analysis was performed in R. Variable selection was carried out via LASSO logistic regression. Model performance was assessed using area under the receiver operating characteristic curve (ROC AUC); decision curve analysis was utilised to assess clinical utility. Internal validation was carried out via 200 bootstrap re-samples with replacement to estimate corrected model fit estimates. Results Analysis included 6,879 participants; 203 (2.95%) anti-CCP positive. Thirteen predictors were retained (age and ankles omitted); scores for each predictor were: sex (male) [+3], FDR with RA [+3], smoking (ever) [+2], pain in back [-3], neck [-2], knee [-1], wrist [+4], foot/toes [+3], hand/fingers [+3], shoulder [+3], thumb [+1]. With a score of ≥ 11 (≥4% risk) for anti-CCP testing, the ROC AUC was 0.65 (95% CI [0.61-0.69], corrected AUC=0.49). Choosing a threshold of 4% suggested that the benefit of testing one anti-CCP positive patient is 24 times larger than the harm of testing one anti-CCP negative patient. Sensitivity and specificity were 38.42 and 81.88 and positive- and negative-predicted values were 6.06 and 97.76. The net benefit was 0.0040 (corrected=0.0026). This threshold resulted in 19% being tested, corresponding to 39% of all anti-CCP positive participants. 12 month follow-up data were available for 2,480 participants. Out of those who would be tested using our model, 6.68% developed RA, compared with 1.94% if everyone was tested. Out of those tested who were anti-CCP positive, 56.85% developed RA (amongst those tested who were anti-CCP negative, 1.23% developed RA). Conclusion Targeted use of anti-CCP testing in PC may prompt earlier identification of people at risk of RA. Interviews with PC clinicians are currently being undertaken to understand the key behaviours required to support clinicians using this prediction model, alongside heath economic modelling to explore potential cost-effectiveness. Disclosure H.J. Siddle: Grants/research support; HJS has received funding from HEE/NIHR for this research [ICA-SCL-2018-04-ST2-004]. M. Wilson: None. J.L. Nam: None. L. Garcia-Montoya: None. L. Duquenne: None. K. Mankia: Consultancies; Serac Lifesciences. Honoraria; Galapagos, Lilly, Abbvie. Grants/research support; Eli Lilly, Gilead. Other; Attending meetings/travel: Abbvie. P. Emery: Consultancies; Lilly, Galapagos. Honoraria; Novartis, Lilly, Abbvie, Gilead, Galapagos. Grants/research support; Lilly, Samsung. Other; Advisory Board: Lilly, Astra Zeneca, BMS. E.M.A. Hensor: Grants/research support; Institution level grant for the NIHR Leeds Biomedical Research Centre.

Discovery and technical validation of high-performance methylated DNA markers for the detection of cervical lesions at risk of malignant progression in low- and middle-income countries

BackgroundCervical cancer remains a leading cause of death, particularly in developing countries. WHO screening guidelines recommend human papilloma virus (HPV) detection as a means to identify women at risk of developing cervical cancer. While HPV testing identifies those at risk, it does not specifically distinguish individuals with neoplasia. We investigated whether a quantitative molecular test that measures methylated DNA markers could identify high-risk lesions in the cervix with accuracy.ResultsMarker discovery was performed in TCGA-CESC Infinium Methylation 450 K Array database and verified in three other public datasets. The panel was technically validated using Quantitative Multiplex-Methylation-Specific PCR in tissue sections (N = 252) and cervical smears (N = 244) from the USA, South Africa, and Vietnam. The gene panel consisted of FMN2, EDNRB, ZNF671, TBXT, and MOS. Cervical tissue samples from all three countries showed highly significant differential methylation in squamous cell carcinoma (SCC) with a sensitivity of 100% [95% CI 74.12–100.00], and specificity of 91% [95% CI 62.26–99.53] to 96% [95% CI 79.01–99.78], and receiver operating characteristic area under the curve (ROC AUC) = 1.000 [95% CI 1.00–1.00] compared to benign cervical tissue, and cervical intraepithelial neoplasia 2/3 with sensitivity of 55% [95% CI 37.77–70.84] to 89% [95% CI 67.20–98.03], specificity of 93% [95% CI 84.07–97.38] to 96% [95% CI 79.01–99.78], and a ROC AUC ranging from 0.793 [95% CI 0.68–0.89] to 0.99 [95% CI 0.97–1.00] compared to CIN1. In cervical smears, the marker panel detected SCC with a sensitivity of 87% [95% CI 77.45–92.69], specificity 95% [95% CI 88.64–98.18], and ROC AUC = 0.925 [95% CI 0.878–0.974] compared to normal, and high-grade squamous intraepithelial lesion (HSIL) at a sensitivity of 70% (95% CI 58.11–80.44), specificity of 94% (95% CI 88.30–97.40), and ROC AUC = 0.884 (95% CI 0.822–0.945) compared to low-grade intraepithelial lesion (LSIL)/normal in an analysis of pooled data from the three countries. Similar to HPV-positive, HPV-negative cervical carcinomas were frequently hypermethylated for these markers.ConclusionsThis 5-marker panel detected SCC and HSIL in cervical smears with a high level of sensitivity and specificity. Molecular tests with the ability to rapidly detect high-risk HSIL will lead to timely treatment for those in need and prevent unnecessary procedures in women with low-risk lesions throughout the world. Validation of these markers in prospectively collected cervical smear cells followed by the development of a hypermethylated marker-based cervical cancer detection test is warranted.

Identification of Perceived Challenges in the Green Energy Transition by Turkish Society through Sentiment Analysis

Green energy refers to energy derived from renewable sources such as solar, wind, hydro, and biomass, which are environmentally sustainable. It aims to reduce reliance on fossil fuels and mitigate environmental impacts. In the Turkish context, alongside positive sentiments regarding the establishment of energy plants, there are also prevalent negative perspectives. Societal responses to the transition towards green energy can be effectively gauged through the analysis of individual comments. However, manually examining thousands of comments is both time-consuming and impractical. To address this challenge, this study proposes the integration of the Transformer method, a Natural Language Processing (NLP) technique. This study presents a defined NLP procedure that utilizes a multi-labeled NLP model, with a particular emphasis on the analysis of comments on social media classified as “dirty text”. The primary objective of this investigation is to ascertain the evolving perception of Turkish society regarding the transition to green energy over time and to conduct a comprehensive analysis utilizing NLP. The study utilizes a dataset that is multi-labeled, wherein emotions are not equally represented and each dataset may contain multiple emotions. Consequently, the measured accuracy rates for the risk, environment, and cost labels are, respectively, 0.950, 0.924, and 0.913, whereas the ROC AUC scores are 0.896, 0.902, and 0.923. The obtained results indicate that the developed model yielded successful outcomes. This study aims to develop a forecasting model tailored to green energy to analyze the current situation and monitor societal behavior dynamically. The central focus is on determining the reactions of Turkish society during the transition to green energy. The insights derived from the study aim to guide decision-makers in formulating policies for the transition. The research concludes with policy recommendations based on the model outputs, providing valuable insights for decision-makers in the context of the green energy transition.

User tendency-based rating scaling in online trading networks.

Social networks often involve the users rating each other based on their beliefs, abilities, and other characteristics. This is particularly common in e-commerce platforms where buyers rate sellers based on their trustworthiness. However, the rating tends to vary between users due to differences in their individual scoring criteria. For example, in a transaction network, a positive user may give a high rating unless the transaction was unsatisfactory while a neutral user may give a mid-rating, consequently giving the same numeric score to different levels of satisfaction. In this paper, we propose a novel method called user tendency-based rating scaling, which adjusts the current rating (its score) based on the pattern of past ratings. We investigate whether this rating scaling method can classify between "good users" and "bad users" in online trade social networks better when compared with using the original rating scores without scaling. Classifying between good users and bad users is especially important for anonymous rating networks like Bitcoin transaction networks, where users' reputations must be recorded to preclude fraudulent and risky users. We evaluate the proposed rating scaling method by performing user classification, link prediction, and clustering tasks, using three real-world online rating network datasets. We use both the original ratings and the scaled ratings as weights of graphs and use a weighted graph embedding method to find node representations that reflect users' positive and negative information. The experimental results showed that using the proposed rating scaling method outperformed using the original (i.e., unscaled) ratings by up to 17% in classification accuracy, and by up to 2.5% in link prediction based on the AUC ROC measure, and by up to 21% in the clustering tasks based on the Dunn-index.

Serum Levels of miR-122-5p and miR-125a-5p Predict Hepatotoxicity Occurrence in Patients Undergoing Autologous Hematopoietic Stem Cell Transplantation.

Hepatic complications are an acknowledged cause of mortality and morbidity among patients undergoing hematopoietic stem cell transplantation. In this study, we aimed to evaluate the potential role in the prediction of liver injury of five selected microRNAs (miRNAs)-miR-122-5p, miR-122-3p, miR-15b-5p, miR-99b-5p, and miR-125a-5p-in the setting of autologous hematopoietic stem cell transplantation (ASCT). A total of 66 patients were included in the study: 50 patients (75.8%) with multiple myeloma (MM) and 16 (24.2%) with lymphoma. Blood samples were collected after the administration of the conditioning regimen, on the day of transplant (day 0). The expression levels of selected miRNAs were quantified by reverse transcription-quantitative polymerase chain reaction (RT-qPCR) using the miRCURY LNA miRNA Custom PCR Panels (QIAGEN). In a multivariate logistic regression analysis adjusted for age, sex, and the administered conditioning regimen, two miRNAs, hsa-miR-122-5p (odds ratio, OR 2.10, 95% confidence interval, CI: 1.29-3.42, p = 0.0029) and hsa-miR-125a-5p (OR 0.27, 95% CI: 0.11-0.71, p = 0.0079), were independent for hepatic toxicity occurrence during the 14 days after transplant. Our model in 10-fold cross-validation preserved its diagnostic potential with a receiver operating characteristics area under the curve (ROC AUC) of 0.75, 95% CI: 0.63-0.88 and at optimal cut-off reached 72.0% sensitivity and 74.4% specificity. An elevated serum level of miR-122-5p and decreased level of miR-125a-5p on day 0 are independent risk factors for hepatotoxicity in ASCT recipients, showing promise in accurately predicting post-ASCT complications. Identifying patients susceptible to complications has the potential to reduce procedure costs and optimize the selection of inpatient or outpatient procedures.

Optimization and Performance Evaluation of Deep Learning Algorithm in Medical Image Processing

In this paper, the optimization and performance evaluation of deep learning algorithm in medical image processing are studied. Firstly, the paper introduces the importance and challenges of medical image processing, and expounds the application prospect of deep learning in this field. Subsequently, this paper discusses the optimization methods of deep learning algorithm in detail, including model structure design, data preprocessing, super parameter adjustment and so on. In terms of performance evaluation, this study selected classic models such as U-Net, DeepLab and DenseNet, and compared them with ROC curve and AUC value to evaluate their predictive ability in medical image classification. The results show that the DenseNet model shows high performance in prediction accuracy, while the performance of U-Net and DeepLab models is slightly average. Finally, the advantages and disadvantages of each model are analyzed, and the future research direction is prospected. This study is of great significance to promote the development and application of medical image processing technology, and provides important theoretical and technical support for medical diagnosis and treatment.

Applying Machine Learning for Enhanced MicroRNA Analysis: A Companion Risk Tool for Oral Squamous Cell Carcinoma in Standard Care Incisional Biopsy.

Machine learning analyses within the realm of oral cancer outcomes are relatively underexplored compared to other cancer types. This study aimed to assess the performance of machine learning algorithms in identifying oral cancer patients, utilizing microRNA expression data. In this study, we implemented this approach using a panel of oral cancer-associated microRNAs sourced from standard incisional biopsy specimens to identify cases of oral squamous cell carcinomas (OSCC). For the model development process, we used a dataset comprising 30 OSCC and 30 histologically normal epithelium (HNE) cases. We initially trained a logistic regression prediction model using 70 percent of the dataset, while reserving the remaining 30 percent for testing. Subsequently, the model underwent hyperparameter tuning resulting in enhanced performance metrics. The hyperparameter-tuned model exhibited high accuracy (0.894) and ROC AUC (0.898) in predicting OSCC. Testing the model on cases of potentially malignant disorders (OPMDs) revealed that leukoplakia with mild dysplasia was predicted as having a high risk of progressing to OSCC, emphasizing machine learning's advantage over histopathology in detecting early molecular changes. These findings underscore the necessity for further refinement, incorporating a broader set of variables to enhance the model's predictive capabilities in assessing the risk of oral potentially malignant disorders.

Adipokine (adiponectin-rs1501299) Gene Variant and Patient Characteristics in Relation to Metabolic-associated Fatty Liver Disease

BackgroundSeveral genetic and metabolic variables, most notably the variation in the adipokine gene rs1501298, have been linked to metabolic-associated fatty liver disease etiopathogenesis (MAFLD). Liver biopsy, the gold standard for diagnosing MAFLD, is an invasive procedure; therefore, alternative diagnostic methods are required. Consequently, the integration of these metabolic variables with some of the patients’ characteristics may facilitate the development of non-invasive diagnostic methods that aid in the early detection of MAFLD, identification of at-risk individuals, and planning of management strategies. MethodsThis study included 224 Egyptians (107 healthy individuals and 117 MAFLD patients). Age, sex, BMI, clinical and laboratory characteristics, and rs1501299 adipokine gene polymorphisms were examined. The Rs1501299 variant, insulin resistance, hypertension, obesity, blood pressure, lipid profile, hemoglobin A1C level, and hepatic fibrosis predictors were evaluated for MAFLD risk. The feasibility and effectiveness of developing non-invasive MAFLD diagnostic models will be investigated. ResultsThe +276G/T (rs1501299) polymorphism (GG vs GT/TT) was linked with MAFLD (OR: 0.43, CI: 0.26-0.69, p=0.002). The GG variants had lower MAFLD rates than those of the GT and TT variants. In addition to altered lipid profiles, patients with MAFLD showed increased gamma-glutamyl transferase levels (GGT: 56 IU/L vs. 36 IU/L). Genetic diversity also affects the accuracy of hepatic fibrosis and steatosis prediction. Hepatic fibrosis and steatosis predictors had ROC AUCs of 0.529%, 0.846%, and 0.700–0.825%, respectively. We examined a diagnostic model based on these variables, and demonstrated its effectiveness. ConclusionThe Adipokine variant rs1501299 increased the risk of MAFLD. Identifying and genotyping this variation and other metabolic variables allows for a non-invasive diagnostic model for early MAFLD diagnosis and identification of those at risk. This study illuminates the prevention and management of MAFLD. Further research with more participants is needed to verify these models and to prove their MAFLD diagnostic efficacy.