Curve AUC Research Articles

A novel prognostic model for predicting the risk of first variceal hemorrhage in patients with HBV-related cirrhosis.

Variceal hemorrhage (VH) is a life-threatening complication of cirrhosis. An accurate VH risk evaluation is critical to determine appropriate prevention strategies. We aimed to develop an individualized prediction model to predict the risk of first VH in hepatitis B virus (HBV)-related cirrhotic patients. A nomogram was developed based on a retrospective analysis of 527 consecutive HBV-related cirrhotic patients with gastroesophageal varices (GEVs). The nomogram evaluation was performed using the area under the receiver operating characteristic curve (AUC), concordance index (C-index), calibration plot, and decision curve analysis (DCA). The results were verified using an external cohort (n = 187). We developed a nomogram based on clinical and endoscopic features, including the size of varices, red wale marks, ascites, spleen thickness, γ-glutamyltransferase, and hematocrit. The C-index of the nomogram in the derivation and validation cohort was 0.806 and 0.820, respectively, and the calibration plot fitted well. Compared with those of the North Italian Endoscopic Club (NIEC) and revised NIEC indexes, the AUC (derivation cohort: 0.822 vs. 0.653 vs. 0.713; validation cohort: 0.846 vs. 0.685 vs. 0.747) and DCA curves of this nomogram were better. Further, based on the risk scores, patients were classified into low-, medium-, and high-risk groups, and significant differences were noted in VH incidence among the three risk groups (P <0.001 for each cohort). An effective individualized nomogram to predict the risk of first VH in HBV-related GEV patients was established, which can assist clinicians in developing more appropriate prevention strategies.

GAPDH Is a Novel Ferroptosis-Related Marker and Correlates with Immune Microenvironment in Lung Adenocarcinoma

Lung adenocarcinoma (LUAD) is a prevalent form of lung cancer with high morbidity and fatality rates. Ferroptosis is a type of programmed cell death that is iron-dependent. Recent findings have suggested that ferroptosis inducers have promising prospects for the therapy of LUAD. However, ferroptosis-related gene expression in LUAD and its relationship with the tumor prognosis and tumor immune microenvironment remain unknown. We identified a total of 638 ferroptosis-related genes, built a LUAD ferroptosis-related risk model (FRRM) with the help of Least Absolute Shrinkage Selection Operator (LASSO) regression analysis based on The Cancer Genome Atlas (TCGA) database, split LUAD patients into high- and low-risk clusters, and verified the model utilizing the Gene Expression Omnibus (GEO) database. The results of the FRRM's principal component analysis (PCA) demonstrated its strong predictive power. Further, univariate and multivariate Cox and AUC curve analyses demonstrated that the model was independent of other clinical traits and served as an independent prognostic factor. The nomogram demonstrated strong predictive power for overall survival, according to calibration plots. We also explored variations in clinical characteristics, immune cell infiltration, immune-related function, and functional pathways between the high- and low-risk groups. Additionally, we used a protein-protein interaction (PPI) network of various genes in the two groups to search for potential target genes. GAPDH was then chosen for a follow-up investigation. An analysis was performed on the relationship between GAPDH and variations in survival prognosis, clinical traits, immune cell infiltration, immune checkpoints, and immunotherapy. In vitro tests further supported the probable functions of GAPDH as a ferroptosis marker in LUAD. In conclusion, a novel ferroptosis-related prognostic gene, GAPDH, was discovered, whose expression was connected to the tumor immune microenvironment. The combination of immunotherapy and the targeting of GAPDH to induce ferroptosis in LUAD may provide a novel therapeutical option.

Development of Clinical Decision Models for the Prediction of Systemic Lupus Erythematosus and Sjogren's Syndrome Overlap.

To explore the clinical features of patients with systemic lupus erythematosus and Sjögren's syndrome overlap (SLE-SS) compared to concurrent SLE or primary SS (pSS) patients, we utilized a predictive machine learning-based tool to study SLE-SS. This study included SLE, pSS, and SLE-SS patients hospitalized at Nanjing Drum Hospital from December 2018 to December 2020. To compare SLE versus SLE-SS patients, the patients were randomly assigned to discovery cohorts or validation cohorts by a computer program at a ratio of 7:3. To compare SS versus SLE-SS patients, computer programs were used to randomly assign patients to the discovery cohort or the validation cohort at a ratio of 7:3. In the discovery cohort, the best predictive features were determined using a least absolute shrinkage and selection operator (LASSO) logistic regression model among the candidate clinical and laboratory parameters. Based on these factors, the SLE-SS prediction tools were constructed and visualized as a nomogram. The results were validated in a validation cohort, and AUC, calibration plots, and decision curve analysis were used to assess the discrimination, calibration, and clinical utility of the predictive models. This study of SLE versus SLE-SS included 290 patients, divided into a discovery cohort (n = 203) and a validation cohort (n = 87). The five best characteristics were selected by LASSO logistic regression in the discovery cohort of SLE versus SLE-SS and were used to construct the predictive tool, including dry mouth, dry eye, anti-Ro52 positive, anti-SSB positive, and RF positive. This study of SS versus SLE-SS included 266 patients, divided into a discovery cohort (n = 187) and a validation cohort (n = 79). In the discovery cohort of SS versus SLE-SS, by using LASSO logistic regression, the eleven best features were selected to build the predictive tool, which included age at diagnosis (years), fever, dry mouth, photosensitivity, skin lesions, arthritis, proteinuria, hematuria, hypoalbuminemia, anti-dsDNA positive, and anti-Sm positive. The prediction model showed good discrimination, good calibration, and fair clinical usefulness in the discovery cohort. The results were validated in a validation cohort of patients. The models are simple and accessible predictors, with good discrimination and calibration, and can be used as a routine tool to screen for SLE-SS.

An intelligent approach for detection and grading of diabetic retinopathy and diabetic macular edema using retinal images

ABSTRACT Early detection of diabetic retinopathy (DR) and diabetic macular edema (DME) is difficult due to the presence of abnormalities like micro-aneurysms in the fundus images which causes vision loss for diabetic patients. In this paper, we detect DR and DME in an earlier stage by performing preprocessing to enhance the quality of the input image, which includes three steps: noise filtering, artefact removal and contrast enhancement. Second, blood vessel segmentation is performed, using Improved Mask-Regional Convolutional Neural Networks (Mask RCNN), which increases the accuracy and precision rate of DR and DME detection. Finally, feature extraction and classification using VGG-16 is performed, which extracts structural features, colour and orientation features. Based on the extracted features, the proposed VGG-16 classifies the image into three classes: normal, DR and DME. After detecting DR and DME, we calculate the severity level of the disease using conditional entropy which classifies the severity into mild, moderate and severe. The proposed work is evaluated by two different datasets: IDRiD and MESSIDOR. The performance of the DR and DME detection is evaluated in terms of accuracy (80.7%), sensitivity (93.67%), specificity (93.67%), F1 score (94.61%), ROC curve and AUC curve compared to existing works.

Chromogranin B in blood serum of patients with neuroendocrine tumors.

Neuroendocrine tumors (NETs) are a heterogeneous group of neoplasms from cells of the diffuse neuroendocrine system. Chromogranin B (CgB) is an acidic protein of the granin family, which can be used to detect the tumours of neuroendocrine nature. Analysis of levels and evaluation of the diagnostic efficiency of CgB in the blood serum of patients with NETs of various localizations. Patients with NETs (n=121) without specific treatment were examined. In the study were presented next localizations: 74 - pancreas, 20 - stomach, 12 - large intestine, 15 - other localizations (lungs, mammary gland, prostate gland, NETs with unidentified primary). 54 practically healthy donors were examined as control group. The determination of CgB in blood serum was performed with ELISA method on BEP 2000 analyzer using a standardized test system Human Chromogranin B (USCN, China). CgB levels in common NET group (median 18.9 ng/mL) were statistically significantly higher than in the control group (8.8 ng/mL). The highest median was obtained in group of intestinal NETs (21.2 ng/ml), which exceeded the median of the control group by more than 2.4 times. According to ROC analysis in the common NET group relative to the control group, the area under the curve AUC was 0.88 (95% CI 0.83-0.929). According to cut-off level of CgB - 15.8 ng/ml, the diagnostic sensitivity was 69.4%, with a specificity of 96.3%. The highest diagnostic sensitivity was in the group of the intestinal NETs (75.0%) and pancreas (71.2%). The study showed the significance of CgB as a potential biochemical marker of NETs with various localizations, alternative to CgA.

Machine Learning Methods for Predicting Adverse Drug Reactions in Hospitalized Patients

Adverse Drug Reaction (ADR) in hospitalized patients have significant impacts on costs as well as on morbidity and mortality. A subset of ADR is preventable at the moment of prescribing. The use of supervised learning tools has been shown to be a good alternative to predict ADR, but more research is needed to make it a standard. A study was conducted to demonstrate how the use of machine learning methods can strengthen the ability to predict ADR in hospitalized patients. Binary classification models were constructed using three class balancing, feature selection and supervised learning approaches known as deep learning, random forest, and gradient boosting trees. Five-fold cross-validation and various evaluation metrics such as AUC, recall and F-measure were used to evaluate the models’ performance. Gradient boosting trees models was the approach that performed best, exhibiting a recall level of 78.3% and an ROC curve AUC of 0.81. Findings on attribute importance were also generated, which should thus prove useful in developing tools for clinical applications. Our predictive models’ results outperform those reported in the recent literature using traditional statistical methods. In this work we have shown the enormous potential of ML tools in this field and their great contribution to the development of information systems that use them.

Investigation of electrical and viscoelastic parameters of erythrocytes in patients with autoimmune gastritis

Patients with autoimmune gastritis (AIG) often have anemia of various origins . Hematological disorders usually portend severe atrophy , and in many cases , are the only indicators of the disease . Aim of the study was to investigate the electrical and viscoelastic parameters of erythrocytes in patients with AIG for their possible use in diagnostics . Material and methods . 73 patients with AIG ( mean age 55.3 ± 12.54 years ) and 38 people of the control group were examined . Electrical and viscoelastic parameters of erythrocytes were studied by dielectrophoresis . Results . Statistically significant decrease in the average cell diameter , the proportion of discocytes cells and an increase in the proportion of spherocytes , deformed forms were found in the group of patients with AIG in combination with Helicobaсter рylori (H. pylori , H.p .) infection compared with healthy individuals . Patients with AIG had significantly lower levels of amplitude of deformation , membrane capacity , dipole moment , speed of cell movement to the electrodes , polarizability at high frequencies of the electric field ( 106 , 0.5×106 Hz ), relative polarizability , and , conversely , higher values of membrane conductivity , aggregation , destruction indexes , summarized viscosity , rigidity , than those in the comparison group . Between groups of patients with and without H. pylori infection , differences were found in indicators reflecting the surface charge of erythrocytes – the speed of movement to the electrodes (p = 0.019), the dipole moment (p &lt; 0.001 ) and the state of the membranes – its capacity (p = 0.004). The diagnostic model , which includes three parameters of erythrocytes – the dipole moment , the speed of movement to the electrodes , the capacity of the cell membrane , provided high diagnostic accuracy of distinguishing AIG H.p . (+) and H.p . (–) – area under ROC curve AUC 0.925, sensitivity 92.4 %, specificity 89.7 %. Conclusions : Electrical and viscoelastic parameters of erythrocytes are promising in the diagnosis of AIG, including on the background of H. pylori infection .

Relationship between occupational noise exposure and hypertension: Cross-sectional evidence from real-world.

Occupational noise is one of the most common and prevalent occupational hazards worldwide and may induce adverse auditory and/or non-auditory health effects. However, the relationship between occupational noise exposure and hypertension is controversial and has long been debated. Based on large sample cross-sectional data from all registered occupational health examination units from 2021 to 2022 (N = 101,605), this study aimed to analyze the prevalence of hearing loss and hypertension and to explore the influencing factors of hypertension of workers in Wuhan. Descriptive statistics, univariate analyses and multivariate analyses were used. Forest plot and nomograms were constructed for the visualization of predictive results. The ROC curve, AUC, C-index and calibration curves were used to assess the predictive accuracy and validity. DCA was performed to evaluate the net benefit that workers could receive. Higher rate of high-frequency hearing loss (25.3%), speech frequency hearing loss (8.8%), ECG abnormalities (31.9%) and hypertension (21.0%) were found in workers exposed to occupational noise in Wuhan. Occupational noise exposure (OR = 1.09, 95% CI: 1.01-1.18, p = 0.04), growth of age (OR: 1.07, 95% CI: 1.07-1.07, p < 0.001), overweight (OR: 1.82, 95% CI: 1.73-1.92, p < 0.001), obesity (OR: 3.62, 95% CI: 3.42-3.83, p < 0.001), hyperglycemia (OR: 1.84, 95% CI: 1.73-1.96, p < 0.001), hypercholesterolemia (OR = 1.34; 95% CI 1.22-1.48; p < 0.001), ECG abnormalities (OR = 1.11; 95% CI 1.07-1.15; p < 0.001) and family history of hypertension (OR = 1.69; 95% CI 1.58-1.81; p < 0.001) were risk factors of hypertension for workers. Male workers had a relatively higher hypertension risk than female workers (OR = 1.61; 95% CI 1.54-1.69; p < 0.001). Ear protective measures could not reduce the risk of hypertension in workers. Our nomogram has good predictive accuracy and validity. A dynamic nomogram to predict the workers' risk of hypertension was established publicly available online. Occupational noise exposure may elevate workers' hypertension risk. More effective and relevant prevention measures should be taken. Our nomogram may help identify high-risk workers and facilitate timely interventions.

Development and validation of a cuproptosis-related lncRNA model correlated to the cancer-associated fibroblasts enable the prediction prognosis of patients with osteosarcoma

BackgroundOsteosarcoma is the most common primary pediatric and adolescent bone malignancy. An imbalance in copper homeostasis caused by copper ion accumulation could increase intracellular toxicity and regulate cancer cell growth. This study aimed to identify long non-coding RNAs (lncRNAs) associated with cuproptosis to predict prognosis and target drug use to improve patient survival. MethodsRNA sequencing and relevant clinical information of ninety-three osteosarcoma patients were obtained from the TARGET database. We then identified thirteen prognostic cuproptosis-related lncRNAs(CRLncs) using coexpression and univariate Cox analyses. The prognostic risk model with three CRLncs was constructed using the least absolute shrinkage and selection operator (LASSO) and multivariate Cox regression analysis. Patients were divided into low-risk and high-risk subgroups using the median risk score. The tumor microenvironment (TME) and immune status of identified subgroups were analyzed using ESTIMATE, CIBERSORT, MCP-counter, xCELL, EPIC, and ssGSEA analyses. Functional analyses were conducted to elucidate the underlying mechanisms, including GO, KEGG, GSVA, and GSEA analyses. Also, the relationships between the model, tumor immunity, and drug sensitivity were explored. Lastly, the expression level of ZNF37BP, AL353759.1, and AC005034.5 was validated in vitro. ResultsWe constructed a model containing three CRLncs (ZNF37BP, AL353759.1, and AC005034.5) and validated its excellent prognostic and predictive power. The AUC curves for 1-year, 3-year, and 5-year survival probabilities were 0.76, 0.84, and 0.89, respectively. Patients in the high-risk group had a shorter overall survival (OS) time than those in the low-risk. The stroma score and cancer-associated fibroblasts (CAFs) were significantly higher in the low-risk group. Immune cells such as T cells CD4 naive, T cells gamma delta, NK cells resting, dendritic cells resting, and mast cells activated were significantly upregulated in the high-risk group. Based on functional analyses, the PI3K-Akt pathway was identified as a critical metabolic pathway in osteosarcoma. Additionally, we obtained three potentially effective drugs for OS: erlotinib, MP470, and WH-4-023 targeting the PI3K-Akt pathway. The expression level of ZNF37BP was significantly elevated in OS cell lines than in normal osteoblast hFOB1.19 cells, and that of ATP7A, LIPT1, AL353759.1, and AC005034.5 were decreased considerably in OS cell lines. ConclusionCuproptosis-related lncRNAs are correlated with the CAFs of osteosarcoma, and this could serve as a foundation for OS survival prediction and treatment.

Fibroblast growth factor 23 as a predictor of left atrial appendage thrombosis in patients with nonvalvular atrial fibrillation and various risk of stroke

Aim. To study the role of fibroblast growth factor 23 (FGF-23) as a predictor of left atrial appendage (LAA) thrombosisin patients (pts) with non-valvular atrial fibrillation (AF) and different risk levels of thromboembolic complications (TEC) according to the CHA2DS2-VASc score.Methods. The study included 250 pts with non-valvular AF hospitalized for radiofrequency ablation, divided into 2 groups: group (gr.) 1 - with 0-1 points (n=79) and gr.2 - with ≥2 points on the CHA2DS2-VASc scale (n=171). According to the results of transesophageal echocardiography, pts of each group were divided into subgroups depending on the presence of LAA thrombosis, which was found in 19 pts of gr.1 and 72 of gr.2. Patients underwent transthoracic echocardiography, determination of blood biomarkers: NT-proBNP (pg/ml), GDF-15 (pg/ml), FGF-23 (pmol/l), highly sensitive C-reactive protein (mg/l), cystatin C (mg/l).Results. Pts with LAA thrombosis in both groups had in common: a longer history of arterial hypertension, a greater proportion of coronary artery disease and persistent AF, lower GFR, higher values of atrial volume indices, left ventricle myocardial mass, systolic pressure in the pulmonary artery, median NT-proBNP, FGF-23 and GDF-15. Logistic regression analysis performed separately in the groups revealed the same independent predictors of LAA thrombosis in both groups: LA volume index (LAVI) and the level of FGF-23 in the blood. At the same time, the threshold value of FGF-23, separating pts by the presence of LAA thrombosis, was higher in gr.1 (≥0.85 and ≥0.5 pmol/l, respectively), and the area under the AUC curve for FGF-23 in gr.1 was greater: 0.964 vs 0.865 in gr.2 (p=0.026).Conclusion. The serum level of FGF-23, along with the LAVI, is an independent predictor of LAA thrombosis in pts with AF and various risk of TEC according to CHA2DS2-VASc; the threshold level of FGF-23 for pts at low risk of TEC is significantly higher than for high risk pts, and allows more accurate prediction of the presence of LAA thrombosis.

Machine learning-supported interpretation of kidney graft elementary lesions in combination with clinical data

Interpretation of kidney graft biopsies using the Banff classification is still heterogeneous. In this study, extreme gradient boosting classifiers learned from two large training datasets (n=631 and 304 cases) where the "reference diagnoses" were not strictly defined following the Banff rules but from central reading by expert pathologists and further interpreted consensually by experienced transplant nephrologists, in light of the clinical context. In three external validation datasets (n=3744, 589, and 360), the classifiers yielded a mean ROC curve AUC (95%CI) of: 0.97 (0.92-1.00), 0.97 (0.96-0.97), and 0.95 (0.93-0.97) for antibody-mediated rejection (ABMR); 0.94 (0.91-0.96), 0.94 (0.92-0.95), and 0.91 (0.88-0.95) for T cell-mediated rejection; >0.96 (0.90-1.00) with all three for interstitial fibrosis-tubular atrophy. We also developed a classifier to discriminate active and chronic active ABMR with 95% accuracy. In conclusion, we built highly sensitive and specific artificial intelligence classifiers able to interpret kidney graft scoring together with a few clinical data and automatically diagnose rejection, with excellent concordance with the Banff rules and reference diagnoses made by a group of experts. Some discrepancies may point toward possible improvements that could be made to the Banff classification.

Near infrared spectroscopy detection of hemispheric cerebral ischemia following middle cerebral artery occlusion in rats

Timely and sensitive in vivo estimation of ischemic stroke-induced brain infarction are necessary to guide diagnosis and evaluation of treatments’ efficacy. The gold standard for estimation of the cerebral infarction volume is magnetic resonance imaging (MRI), which is expensive and not readily accessible. Measuring regional cerebral blood flow (rCBF) with Laser Doppler flowmetry (LDF) is the status quo for confirming reduced blood flow in experimental ischemic stroke models. However, rCBF reduction following cerebral artery occlusion often does not correlate with subsequent infarct volume. In the present study, we employed the continuous-wave near infrared spectroscopy (NIRS) technique to monitor cerebral oxygenation during 90 min of the intraluminal middle cerebral artery occlusion (MCAO) in Sprague-Dawley rats (n = 8, male). The NIRS device consisted of a controller module and an optical sensor with two LED light sources and two photodiodes making up two parallel channels for monitoring left and right cerebral hemispheres. Optical intensity measurements were converted to deoxyhemoglobin (Hb) and oxyhemoglobin (HbO2) changes relative to a 2-min window prior to MCAO. Area under the curve (auc) for Hb and HbO2 was calculated for the 90-min occlusion period for each hemisphere (ipsilateral and contralateral). To obtain a measure of total ischemia, auc of the contralateral side was subtracted from the ipsilateral side resulting in ΔHb and ΔHbO2 parameters. Infarct volume (IV) was calculated by triphenyl tetrazolium chloride (TTC) staining at 24h reperfusion. Results showed a significant negative correlation (r = −0.81, p = 0.03) between ΔHb and infarct volume. In conclusion, our results show feasibility of using a noninvasive optical imaging instrument, namely NIRS, in monitoring cerebral ischemia in a rodent stroke model. This cost-effective, non-invasive technique may improve the rigor of experimental models of ischemic stroke by enabling in vivo longitudinal assessment of cerebral oxygenation and ischemic injury.

Stacking Ensemble and ECA-EfficientNetV2 Convolutional Neural Networks on Classification of Multiple Chest Diseases Including COVID-19

Early detection and treatment of COVID-19 patients is crucial. Convolutional neural networks have been proven to accurately extract features in medical images, which accelerates time required for testing and increases the effectiveness of COVID-19 diagnosis. This study proposes two classification models for multiple chest diseases including COVID-19. The first is Stacking-ensemble model, which stacks six pretrained models including EfficientNetV2-B0, EfficientNetV2-B1, EfficientNetV2-B2, EfficientNetV2-B3, EfficientNetV2-S and EfficientNetV2-M. The second model is self-designed model ECA-EfficientNetV2 based on ECA-Net and EfficientNetV2. Ten-fold cross validation was performed for each model on chest X-ray and CT images. One more dataset, COVID-CT dataset, was tested to verify the performance of the proposed Stacking-ensemble and ECA-EfficientNetV2 models. The best performance comes from the proposed ECA-EfficientNetV2 model with the highest Accuracy of 99.21%, Precision of 99.23%, Recall of 99.25%, F1-score of 99.20%, and (area under the curve) AUC of 99.51% on chest X-ray dataset; the best performance comes from the proposed ECA-EfficientNetV2 model with the highest Accuracy of 99.81%, Precision of 99.80%, Recall of 99.80%, F1-score of 99.81%, and AUC of 99.87% on chest CT dataset. The differences for five metrics between Stacking-ensemble and ECA-EfficientNetV2 models are not significant. Ensemble model achieves better performance than single pretrained models. Compared to the SOTA, Stacking-ensemble and ECA-EfficientNetV2 models proposed in this study demonstrate promising performance on classification of multiple chest diseases including COVID-19.

Detecting Anomalies in Network Communities Based on Structural and Attribute Deviation

Anomaly detection in online social networks (OSNs) is an important data mining task that aims to detect unexpected and suspicious users. To enhance anomaly exploration, anomaly ranking is used to assess the degree of user anomaly rather than applying binary detection methods, which depend on identifying users as either anomalous users or normal users. In this paper, we propose a community-based anomaly detection approach called Community ANOMaly detection (CAnom). Our approach aims to detect anomalous users in an OSN community and rank them based on their degree of deviation from normal users. Our approach measures the level of deviation in both the network structure and a subset of the attributes, which is defined by the context selection. The approach consists of two phases. First, we partition the network into communities. Then, we compute the anomaly ranking score, which is composed of a community-structure-based score and an attribute-based score. Experiments on real-world benchmark datasets show that CAnom detects ground-truth groups and outperforms baseline algorithms on accuracy. On synthetic datasets, the results show that CAnom has high AUC and ROC curves even when the attribute number increases; therefore, our model is suitable for today’s applications, where the number of attributes is rising.

Driver Identification Using Optimized Deep Learning Model in Smart Transportation

The Intelligent Transportation System (ITS) is said to revolutionize the travel experience by making it safe, secure, and comfortable for the people. Although vehicles have been automated up to a certain extent, it still has critical security issues that require thorough study and advanced solutions. The security vulnerabilities of ITS allows the attacker to steal the vehicle. Therefore, the identification of drivers is required in order to develop a safe and secure system so that the vehicles can be protected from theft. There are two ways in which a driver can be identified: 1) face recognition of the driver, and 2) based on driving behavior. Face recognition includes image processing of 2-D images and learning of the features, which require high computational power. Drivers are known to have unique driving styles, whose data can be captured by the sensors. Therefore, the second method identifies drivers based on the analysis of the sensor data and it requires comparatively lesser computational power. In this paper, an optimized deep learning model is trained on the sensor data to correctly identify the drivers. The Long Short-Term Memory (LSTM) deep learning model is optimized for better performance. The novelty of the approach in this work is the inclusion of hyperparameter tuning using a nature-inspired optimization algorithm, which is an important and essential step in discovering the optimal hyperparameters for training the model which in turn increases the accuracy. The CAN-BUS dataset is used for experimentation and evaluation of the training model. Evaluation parameters such as accuracy, precision score, F1 score, and ROC AUC curve are considered to evaluate the performance of the model.

Linked Open Government Data to Predict and Explain House Prices: The Case of Scottish Statistics Portal

Accurately estimating the prices of houses is important for various stakeholders including house owners, real estate agencies, government agencies, and policy-makers. Towards this end, traditional statistics and, only recently, advanced machine learning and artificial intelligence models are used. Open Government Data (OGD) have a huge potential especially when combined with AI technologies. OGD are often published as linked data to facilitate data integration and re-usability. EXplainable Artificial Intelligence (XAI) can be used by stakeholders to understand the decisions of a predictive model. This work creates a model that predicts house prices by applying machine learning on linked OGD. We present a case study that uses XGBoost, a powerful machine learning algorithm, and linked OGD from the official Scottish data portal to predict the probability the mean prices of houses in the various data zones of Scotland to be higher than the average price in Scotland. XAI is also used to globally and locally explain the decisions of the model. The created model has Receiver Operating Characteristic (ROC) AUC score 0.923 and Precision Recall Curve (PRC) AUC score 0.891. According to XAI, the variable that mostly affects the decisions of the model is Comparative Illness Factor, an indicator of health conditions. However, local explainability shows that the decisions made in some data zones may be mostly affected by other variables such as the percent of detached dwellings and employment deprived population.

Deep Learning Classification of Colorectal Lesions Based on Whole Slide Images

Microscopic tissue analysis is the key diagnostic method needed for disease identification and choosing the best treatment regimen. According to the Global Cancer Observatory, approximately two million people are diagnosed with colorectal cancer each year, and an accurate diagnosis requires a significant amount of time and a highly qualified pathologist to decrease the high mortality rate. Recent development of artificial intelligence technologies and scanning microscopy introduced digital pathology into the field of cancer diagnosis by means of the whole-slide image (WSI). In this work, we applied deep learning methods to diagnose six types of colon mucosal lesions using convolutional neural networks (CNNs). As a result, an algorithm for the automatic segmentation of WSIs of colon biopsies was developed, implementing pre-trained, deep convolutional neural networks of the ResNet and EfficientNet architectures. We compared the classical method and one-cycle policy for CNN training and applied both multi-class and multi-label approaches to solve the classification problem. The multi-label approach was superior because some WSI patches may belong to several classes at once or to none of them. Using the standard one-vs-rest approach, we trained multiple binary classifiers. They achieved the receiver operator curve AUC in the range of 0.80–0.96. Other metrics were also calculated, such as accuracy, precision, sensitivity, specificity, negative predictive value, and F1-score. Obtained CNNs can support human pathologists in the diagnostic process and can be extended to other cancers after adding a sufficient amount of labeled data.

Dosiomics and Radiomics Features-Based Model to Predict the Neutrophil-Lymphocyte Ratio in Locally Advanced Non-Small Cell Lung Cancer Treated with Definitive Radiotherapy

<h3>Purpose/Objective(s)</h3> The neutrophil-to-lymphocyte (NLR) is a marker of systemic inflammatory that may predict clinical outcomes of many malignancies. This study therefore developed and validated a dosiomics and radiomics model based on three-dimensional (3D) dose distribution map and planning computed tomography (CT) images to predict the post-radiotherapy (post-RT) NLR. <h3>Materials/Methods</h3> We retrospectively collected 242 locally advanced non-small cell lung cancer (LA-NSCLC) patients. These patients were treated with definitive radiotherapy from 2012 to 2016. The model was developed and validated to predict the NLR collected one month after the completion of RT. Patients were randomly assigned (2:1) into development set and independent testing set. We extracted dosiomics features from 3D dose maps and radiomics features from CT images within different region of interest (ROIs). We developed the prediction model with gradient boosting tree (GBDT) using the features with good discriminability and low redundancy. We tuned the hyperparameters of GBDT model by using five-fold cross-validation. At last, we developed a fusion mode which combined dosiomics, radiomics and clinical features with weighted sum of different model's prediction probability. The model's discriminability was evaluated by receiver operating characteristic curve (ROC) and area under the curve AUC). Kaplan-Meier (KM) curves were used to explore the association between the predicted NLR and patient's survival. <h3>Results</h3> The clinical-based model and DVH-based model achieved an AUC of 0.632 and 0.621 in the testing set, respectively. The 8 features-based dosiomics and 4 features-based radiomics model achieved an AUC of 0.730 (95% CI: 0.599 – 0.860) and 0.701 (0.576 – 0.827). The dosiomics and radiomics and clinical fusion model improved the model's generalization ability with an AUC of 0.751 (95% CI: 0.623 – 0.880), sensitivity of 60.9% and specificity of 80.7% in the testing set. Moreover, the fusion model can significantly divide patients into high-risk and low-risk groups of progression free survival (PFS) and overall survival (OS) (log-rank test, p = 0.0056 and 0.0013). <h3>Conclusion</h3> In patients with LA-NSCLC treated with definitive radiotherapy, the dosiomics and radiomics can improve the prediction of the post-RT NLR.

Neuroeconomic predictors of smoking cessation outcomes: A preliminary study of delay discounting in treatment-seeking adult smokers

Large proportions of smokers are unsuccessful in evidence-based smoking cessation treatment and identifying prognostic predictors may inform improvements in treatment. Steep discounting of delayed rewards (delay discounting) is a robust predictor of poor smoking cessation outcome, but the underlying neural predictors have not been investigated. Forty-one treatment-seeking adult smokers completed a functional magnetic resonance imaging (fMRI) delay discounting paradigm prior to initiating a 9-week smoking cessation treatment protocol. Behavioral performance significantly predicted treatment outcomes (verified 7-day abstinence, n = 18; relapse, n = 23). Participants in the relapse group exhibited smaller area under the curve (d = 1.10) and smaller AUC was correlated with fewer days to smoking relapse (r = 0.56, p < 0.001) Neural correlates of discounting included medial and dorsolateral prefrontal cortex, posterior cingulate, precuneus and anterior insula, and interactions between choice type and relapse status were present for the dorsolateral prefrontal cortex, precuneus and the striatum. This initial investigation implicates differential neural activity in regions associated with frontal executive and default mode activity, as well as motivational circuits. Larger samples are needed to improve the resolution in identifying the neural underpinnings linking steep delay discounting to smoking cessation.

Predictive and prognostic value of different cardiac troponin assays: a nationwide register-based cohort study.

Guidelines do not differentiate between the available assays of cardiac troponin (cTn). We compared the prognostic and predictive ability of cTn assays. This was a nationwide cohort study of patients with acute coronary syndrome (ACS) and≥2 cTn measurements of one of four assays: Roche high-sensitivity cTnT (hs-cTnT), Abbott high sensitivity cTnI (hs-cTnI), Siemens Vista cTnI, and Siemens cTnI Ultra. Data were collected from Danish registries from 2009-18. Peak cTn concentration normalized to the 99th percentile was used. Outcomes were myocardial infarction (MI) during admission, one-year all-cause-, cardiovascular-, and non-cardiovascular mortality. Receiver operating characteristics and logistic regression calculating odds ratios (OR) were used. A total of 90 705 patients were included, of which 20 550 (23%) had MI. Siemens Vista cTnI was the strongest predictor of MI, Area under the curve (auc) 0.93 (95% CI 0.93-0.93). In 1 year 9012 (9.9%) of patients had died. An inverted U-shape relationship was observed between concentration of cTn and all-cause mortality. Hs-cTnT OR 21.3 (95% CI 18.4-24.8) at 2-5 times the 99th percentile and 12.1 (95% CI 10.3-14.1) for concentrations>100 times the 99th percentile. The inverted U-shape relationship was only present for non-cardiovascular mortality. The strongest predictor of cardiovascular mortality was hs-cTnT, OR 11.3 (95% CI 6.4-21.8) at 1-2 times the 99th percentile and 88.8 (95% CI 53.2-163.0) for concentrations>100 times the 99th percentile. Siemens Vista cTnI was the strongest predictor of MI and hs-cTnT was the strongest predictor of mortality. An inverted U-shape relationship was observed between cTn concentration and non-cardiovascular mortality.