Area Under ROC Curve Research Articles

A New Formula Based on Simple Blood Indices to Differentiate Beta Thalassemia Trait from Iron Deficiency Anemia.

Thalassemia is an inherited blood disorder with a defect in the sufficient production of a protein called hemoglobin. We aimed to investigate the simple blood indices of patients with Beta Thalassemia Trait (BTT) and Iron Deficiency Anemia (IDA) to propose a new formula using logistic regression for differentiate two characteristics from each other. Among the 702 records of the BTT Counseling Center (Khoy, Iran-2022), 292 cases (219 iron deficiency anemia (IDA) and 73 BTT) were eligible for the study. Blood indices such as RBC, HGB, HbA2 described and used to diagnose two types of participants. Blood indices had high multicollinearity that was modified. Logistic regression for blood indices fitted and goodness of fit indices with Area Under ROC curve (AUC) estimated. The average age of the participants was 24.56 yr. The status of Multicollinearity between independent variables was modified. The HGB, MCV, HbA2, and HbA variables were used in the model and only HbA2 status was significant (P<0.001). According to the output of the model, for each unit increase in HbA2, the chance of having BTT was about 8.5 times higher than IDA. The sensitivity, specificity, AUC curve, and accuracy of the final model were estimated to be 97, 72, 84, and 93%, respectively. A regression formula to differentiate BTT from IDA proposed. In studies related to the differentiation of the BTT from IDA, the presence of the HbA2 index in the model and prediction is very necessary.

Risk Factors and Clinical Significance of Ultra-Long-Term Microischemia After Intracranial Aneurysm Embolization.

This study aimed to explore influencing factors and clinical significance of ultra-long-term microischemia following intracranial aneurysm (IA) embolization and establish a theoretical foundation for reducing both the incidence of ultra-long-term microischemia and cognitive dysfunction in patients post embolization. A retrospective analysis was conducted on data from 147 patients who received endovascular treatment for IAs. Patients were categorized into microischemic and control (non-microischemic) groups on the based on the findings of high-resolution magnetic resonance vessel wall imaging (HR-VWI) examinations performed 3days postoperatively and 6months postoperatively. Risk factors for the occurrence of ultra-long-term microischemia were determined by univariate analysis and multivariate logistic regression analysis. Out of 147 patients included in the study, 51 (34.69%) developed microischemia while the remaining 96 (65.31%) did not experience this condition. Analysis revealed that factors such as sex, age, history of underlying diseases (hypertension, diabetes mellitus), aneurysmal site characteristics, the presence or absence of stenosis in the aneurysm-bearing artery, modified Fisher score at admission, Barthel's index at discharge, immunoinflammatory index at 3days postoperatively and at the 6-month follow-up, the presence or absence of aneurysmal wall enhancement, and the presence or absence of aneurysmal lumen showed no statistically significant differences between the two groups (all P > 0.05). By contrast, variables like in operative time, rupture status of the aneurysm before surgery according to World Federation of Neurologic Surgeons (WFNS) grade, aneurysm size, number of stents used, number of guidewires and catheters used, and Evans index between the two groups were found to have statistically significant disparities between those who developed microischemia and those who did not (P < 0.05). A subsequent multivariate analysis revealed that aneurysm size, Evans index, and the number of stents used were independent risk factors for the occurrence of ultra-long-term microischemia after surgical intervention of aneurysms (P < 0.05). The receiver operating characteristic (ROC) curves of the patients were constructed on the basis of risk factors determined through multivariate logistic regression analysis. Results indicated that aneurysm size (area under ROC curve (AUC) 0.619, sensitivity 94.7%, specificity 17.1%, P = 0.049), Evans index (AUC 0.670, sensitivity 96.4%, specificity 26.8%, P = 0.004), and number of stents (AUC 0.639, sensitivity 44.6%, specificity 90.2%, P < 0.001) effectively predicted the occurrence of microischemia. The incidence of cognitive dysfunction was higher in the microischemic group than in the control group (P < 0.05), and a greater number of microischemic foci was associated with a higher incidence of cognitive dysfunction. The proportion of microschemia foci in the thalamus and basal ganglia in patients with cognitive dysfunction (60.87%) was significantly higher than that in patients without cognitive dysfunction (34.55%) (P < 0.05). Aneurysm size, Evans index > 0.3, and the quantity of stents were independent risk factors for the occurrence of ultra-long-term microischemia after aneurysm embolization and provided good predictive performance. Cognitive dysfunction was closely associated with microischemia, with its severity increasing with an increase in the number of ischemic foci.

Machine Learning Identifies Key Proteins in Primary Sclerosing Cholangitis Progression and Links High CCL24 to Cirrhosis.

Primary sclerosing cholangitis (PSC) is a rare, progressive disease, characterized by inflammation and fibrosis of the bile ducts, lacking reliable prognostic biomarkers for disease activity. Machine learning applied to broad proteomic profiling of sera allowed for the discovery of markers of disease presence, severity, and cirrhosis and the exploration of the involvement of CCL24, a chemokine with fibro-inflammatory activity. Sera from 30 healthy controls and 45 PSC patients were profiled with proximity extension assay, quantifying the expression of 2870 proteins, and used to train an elastic net model. Proteins that contributed most to the model were tested for correlation to enhanced liver fibrosis (ELF) score and used to perform pathway analysis. Statistical modeling for the presence of cirrhosis was performed with principal component analysis (PCA), and receiver operating characteristics (ROC) curves were used to assess the useability of potential biomarkers. The model successfully predicted the presence of PSC, where the top-ranked proteins were associated with cell adhesion, immune response, and inflammation, and each had an area under receiver operator characteristic (AUROC) curve greater than 0.9 for disease presence and greater than 0.8 for ELF score. Pathway analysis showed enrichment for functions associated with PSC, overlapping with pathways enriched in patients with high levels of CCL24. Patients with cirrhosis showed higher levels of CCL24. This data-driven approach to characterize PSC and its severity highlights potential serum protein biomarkers and the importance of CCL24 in the disease, implying its therapeutic potential in PSC.

Connectome-based schizophrenia prediction using structural connectivity - Deep Graph Neural Network(sc-DGNN).

Connectome is understanding the complex organization of the human brain's structural and functional connectivity is essential for gaining insights into cognitive processes and disorders. To improve the prediction accuracy of brain disorder issues, the current study investigates dysconnected subnetworks and graph structures associated with schizophrenia. By using the proposed structural connectivity-deep graph neural network (sc-DGNN) model and compared with machine learning (ML) and deep learning (DL) models.This work attempts to focus on eighty-eight subjects of diffusion magnetic resonance imaging (dMRI), three classical ML, and five DL models. The structural connectivity-deep graph neural network (sc-DGNN) model is proposed to effectively predict dysconnectedness associated with schizophrenia and exhibits superior performance compared to traditional ML and DL (GNNs) methods in terms of accuracy, sensitivity, specificity, precision, F1-score, and Area under receiver operating characteristic (AUC). The classification task on schizophrenia using structural connectivity matrices and experimental results showed that linear discriminant analysis (LDA) performed 72% accuracy rate in ML models and sc-DGNN performed at a 93% accuracy rate in DL models to distinguish between schizophrenia and healthy patients.

Development and validation of a new prognostic model for patients with acute-on-chronic liver failure in intensive care unit.

Cirrhotic patients with acute-on-chronic liver failure (ACLF) in the intensive care unit (ICU) have a poor but variable prognoses. Accurate prognosis evaluation can guide the rational management of patients with ACLF. However, existing prognostic scores for ACLF in the ICU environment lack sufficient accuracy. To develop a new prognostic model for patients with ACLF in ICU. Data from 938 ACLF patients in the Medical Information Mart for Intensive Care (MIMIC) database were used to develop a new prognostic model (MIMIC ACLF) for ACLF. Discrimination, calibration and clinical utility of MIMIC ACLF were assessed by area under receiver operating characteristic curve (AUROC), calibration curve and decision curve analysis (DCA), respectively. MIMIC ACLF was then externally validated in a multiple-center cohort, the Electronic Intensive Care Collaborative Research Database and a single-center cohort from the Second Hospital of Hebei Medical University in China. The MIMIC ACLF score was determined using nine variables: ln (age) × 2.2 + ln (white blood cell count) × 0.22 - ln (mean arterial pressure) × 2.7 + respiratory failure × 0.6 + renal failure × 0.51 + cerebral failure × 0.31 + ln (total bilirubin) × 0.44 + ln (internationalized normal ratio) × 0.59 + ln (serum potassium) × 0.59. In MIMIC cohort, the AUROC (0.81/0.79) for MIMIC ACLF for 28/90-day ACLF mortality were significantly greater than those of Chronic Liver Failure Consortium ACLF (0.76/0.74), Model for End-stage Liver Disease (MELD; 0.73/0.71) and MELD-Na (0.72/0.70) (all P < 0.001). The consistency between actual and predicted 28/90-day survival rates of patients according to MIMIC ACLF score was excellent and superior to that of existing scores. The net benefit of MIMIC ACLF was greater than that achieved using existing scores within the 50% threshold probability. The superior predictive accuracy and clinical utility of MIMIC ACLF were validated in the external cohorts. We developed and validated a new prognostic model with satisfactory accuracy for cirrhotic patients with ACLF hospitalized in the ICU. The model-based risk stratification and online calculator might facilitate the rational management of patients with ACLF.

Preoperative rectus femoris muscle ultrasound, its relationship with frailty scores, and the ability to predict recovery after cardiac surgery: a prospective cohort study

BackgroundFrailty is common in patients undergoing cardiac surgery and is associated with poorer postoperative outcomes. Ultrasound examination of skeletal muscle morphology may serve as an objective assessment tool as lean muscle mass reduction is a key feature of frailty.MethodsThis study investigated the association of ultrasound-derived muscle thickness, cross-sectional area, and echogenicity of the rectus femoris muscle (RFM) with preoperative frailty and predicted subsequent poor recovery after surgery. Eighty-five patients received preoperative RFM ultrasound examination and frailty-related assessments: Clinical Frailty Scale (CFS) and 5-m gait speed test (GST5m). Association of each ultrasound measurement with frailty assessments was examined. Area under receiver-operating characteristic curve (AUROC) was used to assess the discriminative ability of each ultrasound measurement to predict days at home within 30 days of surgery (DAH30).ResultsBy CFS and GST5m criteria, 13% and 34% respectively of participants were frail. RFM cross-sectional area alone demonstrated moderate predictive association for frailty by CFS criterion (AUROC: 0.76, 95% CI: 0.66–0.85). Specificity improved to 98.7% (95% CI: 93.6%-100.0%) by utilising RFM cross-sectional area as an ‘add-on’ test to a positive gait speed test, and thus a combined muscle size and function test demonstrated higher predictive performance (positive likelihood ratio: 40.4, 95% CI: 5.3–304.3) for frailty by CFS criterion than either test alone (p < 0.001). The combined ‘add-on’ test predictive performance for DAH30 (AUROC: 0.90, 95% CI: 0.81–0.95) may also be superior to either CFS or gait speed test alone.ConclusionsPreoperative RFM ultrasound examination, especially when integrated with the gait speed test, may be useful to identify patients at high risk of frailty and those with poor outcomes after cardiac surgery.Trial registrationThe study was registered on the Chinese Clinical Trials Registry (ChiCTR2000031098) on 22 March 2020.

#635 Effect of multi-biomarker panels based on urinary N-terminal osteopontin on prediction of diabetic kidney disease in patients with diabetes mellitus

Abstract Background and Aims Diabetic kidney disease (DKD), occurs in 20–40% of patients with diabetes mellitus (DM), is the leading cause of end-stage renal disease (ESRD). DKD is a clinical diagnosis mainly based on the persistent albuminuria and reduced estimated glomerular filtration rate (eGFR) [1]. However, increased UACR and reduced eGFR are the final consequences due to DKD, novel biomarkers are critical for predicting DKD development. Osteopontin (OPN) is a profibrotic adhesion phosphoprotein that participates in cell chemotaxis, adhesion, migration, and proliferation, as well as extracellular matrix (ECM) hyperplasia. Hyperglycemia could enhance OPN gene expression through the activation of the renin-angiotensin system (RAS), mTOR pathway, NF-κB, and TGF-β pathway and then cause podocyte injury and ECM hyperplasia. Therefore, OPN upregulation is not only the result of various pathophysiological processes in DKD, but also results in kidney injury. Osteopontin (OPN) could predict incident DKD in DM patients [2], N-terminal OPN (ntOPN) has a stronger profibrotic adhesion effect than full-length OPN. This study aims to reveal the clinical benefit of ntOPN as a potential marker to identify DM patients at high risk of DKD, and establish ntOPN-based diagnostic and forecast models for renal outcomes in DM patients. Method We performed a cross-sectional study of 316 adults with Type 1 DM≥ 5 years or Type 2 DM, then followed by a prospective observational cohort study of 143 adult DM patients without renal involvement at baseline and follow-up for at least one year. During the follow-up period, the primary endpoint was “DKD occurrence”, defined as the presence of one of the following conditions in DM patients [3]: (1) repeat UACR ≥30 mg/g at least 2 of 3 measurements within 3 to 6 months; (2) eGFR &amp;lt;60 mL/min/1.73 m2 for more than three months; (3) renal pathological findings were consistent with DKD. The secondary endpoint was “DKD progression”, which included: (1) eGFR sustained decreased by at least 25%; (2) development of ESRD, and/or need for renal replacement therapy; (3) death from the renal cause. Logistic regression analysis was performed to analyze the relationship between parameters and the events of DKD occurrence and progression. Receiver operator characteristic (ROC) analysis was used to assess the predictive ability of established models for clinical endpoints. Results The median value of urinary ntOPN (UntOPN) was 44.15 ng/ml in the cross-sectional cohort, DKD prevalence was significantly higher in the high UntOPN group than in the low UntOPN group. The ROC curves of UntOPN, urinary neutrophil gelatinase-associated lipocalin (UNGAL), and their combination indicated that both combination and ntOPN alone perform better than UNGAL for DKD diagnosis (Fig. 1A). In the prospective cohort, UntOPN was an independent risk factor and further improved the predictive ability for DKD occurrence and DKD progression than UNGAL (Figs. 1B and 1C). Based on the parameters detected as risk factors for DKD occurrence and progression, we set up a series of multi-biomarker panels for DKD prediction using UntOPN, UNGAL, serum cystatin C, serum creatinine (Scr), UACR, and TCH/HDL-C ratio. Compared with the model of Scr + UACR, the area under ROC curve (AUC) of the six-biomarker model was higher, and also ranked the highest among the six ROC curves in predicting 1-year risk of DKD occurrence and DKD progression (Figs. 1B and 1C). Conclusion Our results showed that urinary ntOPN is associated with DKD development, and elevated urinary ntOPN is an independent predictor for DKD occurrence and progression. Compared with the traditional biomarkers of Scr + UACR, our multi-biomarker models based on urinary ntOPN performed better in predicting DKD development, which could provide more accurate tools for DKD risk prediction, thereby improving the renal prognosis in DM patients.

Analysis of the effects of early screening combined with blood lactate on the severity of patients with sepsis

This work aimed to investigate the adoption value of blood lactic acid (BLA) combined with the National Early Warning Score (NEWS) in the early screening of sepsis patients and assessing their severity. The data and materials utilized in this work were obtained from the electronic medical record system of 537 anonymized sepsis patients who received emergency rescue in the emergency rescue area of Liuzhou People's Hospital, Guangxi, from July 1, 2020, to December 26, 2020. Based on the 28-day outcomes of sepsis patients, the medical records were rolled into Group S (407 survival cases) and Group D (130 dead cases). Basic information such as the mode of hospital admission, initial management, use of emergency ventilator within 24 h of admission, NEWS score, arterial oxygen pressure/alveolar oxygen pressure ratio (PaO2/PAO2), alveolar-arterial oxygen difference (A-aDO2), serum creatinine (SCr), blood urea nitrogen (BUN), oxygenation index (OI), Glasgow Coma Scale (GCS), D-dimer, use of vasoactive drugs within 24 h of admission, C-reactive protein (CRP), procalcitonin (PCT), interleukin-6 (IL-6), N-terminal pro-B-type natriuretic peptide (NT-proBNP), quick Sequential Organ Failure Assessment (qSOFA) score, SOFA score, BLA level, NEWS with lactate (NEWS-L) score, SOFA score including lactate level (SOFA-L) score, Intensive Care Unit (ICU) length of stay, total hospital stay, ICU stay/total hospital stay, and septic shock condition were compared between groups. Logistic regression analysis was performed to assess the impact of various predictive factors on prognosis and to plot the receiver operating characteristic (ROC) curve. The results suggested marked differences between Group S and Group D in terms of mean age (t = −5.620; OR = −9.96, 95 % CI: −13.44∼-6.47; P < 0.001). Group S showed drastic differences in terms of mode of hospital admission (χ2 = 9.618, P < 0.01), method of initial management (χ2 = 51.766, P < 0.001), use of emergency ventilator within 24 h of admission (χ2 = 98.564, P < 0.001), incidence of septic shock (χ2 = 77.545, P < 0.001), use of vasoactive drugs within 24 h of admission (χ2 = 102.453, P < 0.001), heart rate (t = −4.063, P < 0.001), respiratory rate (t = −4.758, P < 0.001), oxygenation status (χ2 = 20.547, P < 0.001), NEWS score (t = −6.120, P < 0.001), PaO2/PAO2 ratio (t = 2.625, P < 0.01), A-aDO2 value (Z = −3.581, P < 0.001), OI value (Z = −3.106, P < 0.01), PLT value (Z = −2.305, P < 0.05), SCr value (Z = −3.510, P < 0.001), BUN value (Z = −3.170, P < 0.01), D-dimer (Z = −4.621, P < 0.001), CRP level (Z = −4.057, P < 0.001), PCT value (Z = −2.783, P < 0.01), IL-6 level (Z = −2.904, P < 0.001), length of hospital stay (Z = −4.138, P < 0.001), total hospital stay (Z = −8.488, P < 0.001), CCU/total hospital stay (Z = −9.118, P < 0.001), NEWS score (t = −6.120, P < 0.001), SOFA score (t = −6.961, P < 0.001), SOFA-L score (Z = −4.609, P < 0.001), NEWS-L score (Z = −5.845, P < 0.001), BLA level (Z = −6.557, P < 0.001), and GCS score (Z = 6.909, P < 0.001) when compared to Group D. The use of ventilators, septic shock, PCT, NEWS score, GCS score, SOFA score, SOFA-L score, NEWS-L score, and BLA level were identified as independent risk factors for predicting the prognosis of sepsis patients (P < 0.001). The areas under ROC curve (AUC) of blood lactic acid, PCT, NEWS, NEWS-L, GCS, SOFA, and SOFA-L were 0.695, 0.665, 0.692, 0.698, 0.477, 0.700, and 0.653, respectively. These findings indicate that the combination of BLA with NEWS (NEWS-L) score and SOFA score has certain advantages in assessing the prognosis of sepsis.

Deep learning driven diagnosis of malignant soft tissue tumors based on dual-modal ultrasound images and clinical indexes

BackgroundSoft tissue tumors (STTs) are benign or malignant superficial neoplasms arising from soft tissues throughout the body with versatile pathological types. Although Ultrasonography (US) is one of the most common imaging tools to diagnose malignant STTs, it still has several drawbacks in STT diagnosis that need improving.ObjectivesThe study aims to establish this deep learning (DL) driven Artificial intelligence (AI) system for predicting malignant STTs based on US images and clinical indexes of the patients.MethodsWe retrospectively enrolled 271 malignant and 462 benign masses to build the AI system using 5-fold validation. A prospective dataset of 44 malignant masses and 101 benign masses was used to validate the accuracy of system. A multi-data fusion convolutional neural network, named ultrasound clinical soft tissue tumor net (UC-STTNet), was developed to combine gray scale and color Doppler US images and clinic features for malignant STTs diagnosis. Six radiologists (R1-R6) with three experience levels were invited for reader study.ResultsThe AI system achieved an area under receiver operating curve (AUC) value of 0.89 in the retrospective dataset. The diagnostic performance of the AI system was higher than that of one of the senior radiologists (AUC of AI vs R2: 0.89 vs. 0.84, p=0.022) and all of the intermediate and junior radiologists (AUC of AI vs R3, R4, R5, R6: 0.89 vs 0.75, 0.81, 0.80, 0.63; p &amp;lt;0.01). The AI system also achieved an AUC of 0.85 in the prospective dataset. With the assistance of the system, the diagnostic performances and inter-observer agreement of the radiologists was improved (AUC of R3, R5, R6: 0.75 to 0.83, 0.80 to 0.85, 0.63 to 0.69; p&amp;lt;0.01).ConclusionThe AI system could be a useful tool in diagnosing malignant STTs, and could also help radiologists improve diagnostic performance.

#357 Sex differences in the aldosterone-renin system must be taken into account during screening for primary aldosteronism

Abstract Background and Aims Primary aldosteronism (PA) is one of the most frequent causes of secondary hypertension. Patients with PA are at higher risk for cardiovascular and renal complications than those with essential hypertension. Screening for PA is rarely considered in the assessment of a hypertensive patient, in part because of difficulties in the diagnostic process. Nevertheless, identification of PA is crucial to guide patients towards optimal management, either by mineralocorticoid receptor blockade or by unilateral adrenalectomy, with the latter offering potential cure for patients with aldosterone-producing adenomas (APAs). Numerous studies show inadequate diagnostic accuracy of the aldosterone:renin ratio (ARR) for screening for PA, particularly in females, who have higher ARR variability and lower renin compared to males. Women have a higher risk of false-positive screening results leading to unnecessary confirmatory tests. Use of oral contraception (OC) or hormone replacement therapy (HRT) may also confound the ARR. The aim of this study was to develop improved screening tools for PA, including adjustments for sex-specific differences. Method PROSALDO is a multicenter prospective cohort study designed to test new diagnostic procedures for patients with suspected PA. Medication that interfere with the renin-angiotensin-aldosterone system are withdrawn before laboratory measurements. For screening, the routine ARR is calculated along with additional probability scores based on a peripheral venous steroid hormone profile (PVSP). The latter allows for identification of potentially false-negative ARR results. After a positive screening result of the ARR and/or a high probability score based on the PVSP, PA is confirmed or excluded with a saline infusion test (SIT). PA confirmation is followed by adrenal venous sampling (AVS) to identify patients with unilateral PA. We compared aldosterone, renin and ARR levels between females and males with and without PA. Renin- and sex dependent formulas for aldosterone screening cut-offs were derived from the study population. Logistic regression and the corresponding receiver operating characteristic (ROC) curves were used to determine and compare diagnostic accuracies. Results PA was confirmed in 159 (24.5%) of 649 (328 (50.5%) females and 321 (49.5%) males). Lateralization was shown in 58 out of 117 patients who underwent AVS. In hypertensive patients without PA, ARR was higher in females vs. males (26.4 vs. 11.7 pmol/mU, P &amp;lt; 0.0001), mainly due to a lower renin (9.2 vs. 16.2 mU/L, P &amp;lt; 0.0001). Furthermore, ARRs in patients with vs. without PA showed more overlap in females compared to males. Areas under ROC curves (AUROCs) for the ARR were 0.857 (CI 95%: 0.797-0.902) vs. 0.912 (CI 95%: 0.876-0.938) for females and males respectively. Using liquid chromatography mass spectrometry measurements (LC-MS/MS) of aldosterone, an ARR cut-off of 21.75 pmol/L yielded a sensitivity of 93% with only 45% specificity in females, while for males, a specificity of 66% was obtained for a sensitivity of 97.7%. This meant for females having a higher risk of false-positive screening results (OR 2.36 (CI 95%: 1.65-3.41)), even after excluding patients on OC or HRT (OR 2.17 (CI 95%: 1.5-3.17)). We derived a renin- and sex-dependent formula for the calculation of aldosterone cut-offs, which improved diagnostic accuracy for both males and females (ΔAUROC = 0.05 and 0.045, P &amp;lt; 0.0001 and 0.02 respectively): for a diagnostic sensitivity of 94-95%, specificity was improved to 61% in females and 82% in males. Conclusion Lower renin in females without PA leads to higher ARRs and thus a higher risk of false-positive screening results. Higher overlap of the ARR between females with and without PA compared to their male counterparts weakens utility of the ARR for discrimination of patients with and without PA. Sex- and renin-specific aldosterone cut-offs for screening for PA offer improved diagnostic accuracy. Diagnostic accuracy remains however superior in males, which highlights the need for further optimization of diagnostic procedures specifically in women.

Risk factors and prediction model of delirium in elderly patients after hip arthroplasty.

To analyze the risk factors of delirium in elderly patients after hip arthroplasty and to construct a prediction model. Clinical data of 248 elderly patients who underwent hip arthroplasty in the Department of Traumatology and Orthopedics at Wuhan Fourth Hospital were retrospectively collected from November 2021 to February 2023. Logistic regression analysis was used to identify the risk factors of delirium after hip arthroplasty, and a nomogram prediction model was constructed using the RMS package of R4.1.2 software. The accuracy and stability of the model was evaluated based on the Hosmer-Lemeshow goodness-of-fit test and the receiver operating characteristic (ROC) curve. Age, nighttime sleep, anesthesia method, intraoperative blood loss, hypoxemia, and C-reactive protein (CRP) level were all risk factors of delirium after the hip arthroplasty (P<0.05). These factors were used to construct a nomogram prediction model that was internally validated using the Bootstrap method. The prediction model had the area under ROC curve (AUC) of 0.980 (95% CI: 0.964-0.996), indicating that it has certain predictive value for postoperative delirium. When the optimal cut off value was selected, the sensitivity and specificity were 92.7% and 92.3%, respectively, indicating that the prediction model is effective. Age, short nighttime sleep, general anesthesia, high intraoperative blood loss, hypoxemia, and high CRP levels are independent risk factors for delirium after hip arthroplasty. The nomogram prediction model constructed based on these risk factors can effectively predict delirium in elderly patients after hip arthroplasty.

Detecting anomalies in blockchain transactions using machine learning classifiers and explainability analysis

As the use of blockchain for digital payments continues to rise, it becomes susceptible to various malicious attacks. Successfully detecting anomalies within blockchain transactions is essential for bolstering trust in digital payments. However, the task of anomaly detection in blockchain transaction data is challenging due to the infrequent occurrence of illicit transactions. Although several studies have been conducted in the field, a limitation persists: the lack of explanations for the model's predictions. This study seeks to overcome this limitation by integrating explainable artificial intelligence (XAI) techniques and anomaly rules into tree-based ensemble classifiers for detecting anomalous Bitcoin transactions. The shapley additive explanation (SHAP) method is employed to measure the contribution of each feature, and it is compatible with ensemble models. Moreover, we present rules for interpreting whether a Bitcoin transaction is anomalous or not. Additionally, we introduce an under-sampling algorithm named XGBCLUS, designed to balance anomalous and non-anomalous transaction data. This algorithm is compared against other commonly used under-sampling and over-sampling techniques. Finally, the outcomes of various tree-based single classifiers are compared with those of stacking and voting ensemble classifiers. Our experimental results demonstrate that: (i) XGBCLUS enhances true positive rate (TPR) and receiver operating characteristic-area under curve (ROC-AUC) scores compared to state-of-the-art under-sampling and over-sampling techniques, and (ii) our proposed ensemble classifiers outperform traditional single tree-based machine learning classifiers in terms of accuracy, TPR, and false positive rate (FPR) scores.

Landslide susceptibility assessment through multi-model stacking and meta-learning in Poyang County, China

This study aims to evaluate the effectiveness of various individual machine learning and their ensemble techniques such as Stacking, Voting and Meta-learning in landslide susceptibility assessment taking Poyang, Jiangxi, China as an example. Multi-source geo-environmental data including field surveys, Sentinel-2A/B satellite images, Digital Elevation Models (DEM), and geological and hydrological data were utilized to construct and validate landslide susceptibility models. Results show that the Stacking Classifier outperformed other models, achieving the highest F1 Score of 0.846 and AUC (Area Under ROC Curve) of 0.923, demonstrating its strong predictivity, followed by the Voting Classifier with the F1 Score of 0.829 and AUC of 0.922. Among the individual models, the Multi-Layer Perceptron (MLP) performed best with the F1 Score of 0.828 and AUC of 0.904. Furthermore, the explainable Artificial Intelligence (XAI) technique was applied to better understand the mechanism of classifiers in predicting landslide susceptibility and it suggests a significant correlation between land use, distance to fault, and landslide occurrences. In conclusion, Stacking and Voting hybrid learning models show clear advantages over the individual ones for landslide risk zoning. The results of study may provide technical support for disaster mitigation efforts and future urban planning in areas prone to landslides in Poyang.

PREDICTION OF FLUID RESPONSIVENESS WITH CRYSTALLOID MINI-FLUID CHALLENGE IN CRITICALLY ILL PATIENTS

Background: The mini-fluid challenge test is a good predictor and has minimal limitations in determining fluid responsiveness in the critically ill. However, it primarily relies on colloid infusion. The availability of colloids may vary among hospitals in resource-limited countries, and they can be more expensive than crystalloid fluids. Objective: The study aimed to use crystalloids instead of colloids to predict fluid responsiveness. Methods: This study was conducted from August 2022 to February 2023 among patients with shock in the medical ICU to assess fluid responsiveness. Arterial and central venous catheters were used for the intravenous infusion and monitoring. Data were collected during two consecutive crystalloid mini-fluid challenges of 50 mL each in 1 minute, 100 mL in 2 minutes, and a standard fluid challenge of 300 mL over 15 minutes. The objective was to predict a stroke volume index (SVI) increase of &gt; 10%. Diagnostic accuracy was evaluated using the Receiver Operating Characteristic (ROC) curve, and hemodynamic variables in the positive fluid challenge group were compared for different volume expansions. Results: Twenty-nine patients (62% males; median age, 75 years) were included, with 42 fluid challenge test events. Septic shock was the primary condition in 83% of the cases. The 50 mL crystalloid mini-fluid challenge showed an accuracy of 80% with a sensitivity of 69% and specificity of 100% at the cutoff level of SVI &gt; 5%, Area Under ROC (AUROC)=0.79. In comparison, the 100 mL crystalloid mini-fluid challenge demonstrated an accuracy of 85% (sensitivity of 79% and specificity of 100%) at the cutoff level of SVI &gt; 10% (AUROC= 0.89). In the positive fluid challenge group, administration of crystalloids led to a significant increase in mean arterial pressure and SVI, while other parameters remained comparable. Conclusions: A mini-fluid challenge test with 100 mL crystalloids can predict fluid responsiveness in critically ill patients. The best cutoff level was a change in SVI &gt; 10% from baseline.

Impact of an artificial intelligence based model to predict non-transplantable recurrence among patients with hepatocellular carcinoma

ObjectiveWe sought to develop Artificial Intelligence (AI) based models to predict non-transplantable recurrence (NTR) of hepatocellular carcinoma (HCC) following hepatic resection (HR). MethodsHCC patients who underwent HR between 2000-2020 were identified from a multi-institutional database. NTR was defined as recurrence beyond Milan Criteria. Different machine learning (ML) and deep learning (DL) techniques were used to develop and validate two prediction models for NTR, one using only preoperative factors and a second using both preoperative and postoperative factors. ResultsOverall, 1763 HCC patients were included. Among 877 patients with recurrence, 364 (41.5%) patients developed NTR. An ensemble AI model demonstrated the highest area under ROC curves (AUC) of 0.751 (95% CI: 0.719–0.782) and 0.717 (95% CI:0.653–0.782) in the training and testing cohorts, respectively which improved to 0.858 (95% CI: 0.835–0.884) and 0.764 (95% CI: 0.704–0.826), respectively after incorporation of postoperative pathologic factors. Radiologic tumor burden score and pathological microvascular invasion were the most important preoperative and postoperative factors, respectively to predict NTR. Patients predicted to develop NTR had overall 1- and 5-year survival of 75.6% and 28.2%, versus 93.4% and 55.9%, respectively, among patients predicted to not develop NTR (p < 0.0001). ConclusionThe AI preoperative model may help inform decision of HR versus LT for HCC, while the combined AI model can frame individualized postoperative care (https://altaf-pawlik-hcc-ntr-calculator.streamlit.app/).

Identifying potential (re)hemorrhage among sporadic cerebral cavernous malformations using machine learning

The (re)hemorrhage in patients with sporadic cerebral cavernous malformations (CCM) was the primary aim for CCM management. However, accurately identifying the potential (re)hemorrhage among sporadic CCM patients in advance remains a challenge. This study aims to develop machine learning models to detect potential (re)hemorrhage in sporadic CCM patients. This study was based on a dataset of 731 sporadic CCM patients in open data platform Dryad. Sporadic CCM patients were followed up 5 years from January 2003 to December 2018. Support vector machine (SVM), stacked generalization, and extreme gradient boosting (XGBoost) were used to construct models. The performance of models was evaluated by area under receiver operating characteristic curves (AUROC), area under the precision-recall curve (PR-AUC) and other metrics. A total of 517 patients with sporadic CCM were included (330 female [63.8%], mean [SD] age at diagnosis, 42.1 [15.5] years). 76 (re)hemorrhage (14.7%) occurred during follow-up. Among 3 machine learning models, XGBoost model yielded the highest mean (SD) AUROC (0.87 [0.06]) in cross-validation. The top 4 features of XGBoost model were ranked with SHAP (SHapley Additive exPlanations). All-Elements XGBoost model achieved an AUROCs of 0.84 and PR-AUC of 0.49 in testing set, with a sensitivity of 0.86 and a specificity of 0.76. Importantly, 4-Elements XGBoost model developed using top 4 features got a AUROCs of 0.83 and PR-AUC of 0.40, a sensitivity of 0.79, and a specificity of 0.72 in testing set. Two machine learning-based models achieved accurate performance in identifying potential (re)hemorrhages within 5 years in sporadic CCM patients. These models may provide insights for clinical decision-making.

Urban explosion and hotspots of forest loss in western Himalaya: Mapping land use/cover change trends since 1975

The Western Himalayan Region (WHR) is under constant threat from Land Use and Land Cover Change (LULCC) brought on by various socioeconomic activities such as urbanization, agricultural expansion, and population growth. Detailed research into the scope of these changes is necessary to comprehend the challenges they represent. This study aimed at (1) examining and quantifying spatiotemporal patterns of LULCC from 1975 to 2015, (2) mapping the hotspots of forest loss, and (3) modelling future LULCC for the year 2055 using medium spatial resolution multispectral satellite imagery (MSS, TM, LISS-I, LISS-III, AWiFS), field data, visual interpretation and logistic regression model in the WHR. Logistic regression integrated various drivers of LULCC, including climatic, topographic, and socio-economic factors and past LULCC trends to forecast potential LULC in the WHR by 2055. The results highlight an outrageous 184% increase in regional urbanization. Expansion of urban and cropland (12%) led to deforestation which reduced the area of natural habitats, specifically forests (−11%), water bodies (−8%) and scrubland (−6%). A significant rise in barren land by 30% was noted, while there was a decrease of 20% in glacier/snow melt. The future projection for 2055 anticipates continuing current trends, emphasizing an intensified urban expansion of 63% and an approximate 9% reduction in overall forest coverage. The accuracy of the predicted 2015 LULC map was assessed using ROC/AUC (Receiver Operating Characteristic/Area under Curve) which was found to be 0.79.

A comprehensive evaluation of serum circCSPP1 as a novel diagnostic and prognostic biomarker for gastric cancer

PurposeGastric cancer (GC) has high incidence and mortality due to its low early screening efficiency. Circular RNAs (CircRNAs) are a new class of non-coding RNAs which is closely related to GC. Nevertheless, the clinical application value of circRNAs in GC are largely unknown. Therefore, we studied the role of a novel circRNA named circCSPP1 in patients with GC. MethodsCircRNA sequencing was performed to screen out the target molecule. Real-time fluorescent quantitative PCR (RT-qPCR) was utilized to detect the expression level of circCSPP1 in GC tissues, cells, and serum. Gel and Sanger sequencing were utilized to verify the ring structure of circCSPP1. RNase R enzyme digestion experiment and actinomycin D experiment were verifed the advantage of circCSPP1 as a diagnostic biomarker in patients with GC when that compared with linear RNA. The correlation between the expression level of serum circCSPP1 and clinicopathological data of GC patients was further analyzed. Receiver operating characteristic curve (ROC) and the area under ROC curve (AUC) were utilied to evaluate the diagnostic performance. ResultsCircCSPP1 has a circular structure which with resistance to RNA exonuclease digestion and long half-life compared with linear RNA. In our study, circCSPP1 was first found up-regulated in patients with GC. Serum circCSPP1 level was decreased significantly after surgical resection whereas increased after recurrence. High expression of circCSPP1 was associated with poor survival rates. The expression level of circCSPP1 was significantly correlated to tumor size, T stage, lymph node metastasis, and TNM stage. The AUC of serum circCSPP1 was 0.834, with high sensitivity and specificity in discriminating patients with GC from healthy donors. More importantly, the combined diagnosis of circCSPP1, CEA, and CA19–9 achieved the superior AUC of 0.882, with the highest specificity. ConclusionSerum circCSPP1 may prove to be a potential non-invasive auxiliary diagnostic biomarker for patients with GC.

Circulating long noncoding RNA, Zfpm2-As1, and XIST based on medical data analysis are potential plasma biomarkers for gastric cancer diagnosis.

Long noncoding RNAs (lncRNAs) participate in diseases, especially tumorigenesis, including gastric cancer (GC). Although lncRNAs in GC tissues have been extensively studied in previous research, the possible significance of circulating lncRNAs in diagnosing GC is still unknown. The present work investigated lncRNAs ZFPM2-AS1 and XIST with high expression in GC tissues proved as potential plasma biomarkers from 20 early GC cases, 100 GC cases, and 90 normal subjects. The possible correlation between ZFPM2-AS1 and XIST expression levels was analyzed with general characteristics and clinicopathological features. The performance in diagnosis was assessed according to receiver operating characteristic (ROC) analysis. According to the results, XIST and ZFPM2-AS1 expression remarkably increased within GC plasma relative to normal subjects (P< 0.01); besides, lncRNA XIST expression after surgery had a tendency of downregulation compared with preoperative levels (P< 0.05). Moreover, the area under ROC curve (AUC) values were 0.62 for ZFPM2-AS1 and 0.68 for XIST, while the pooled AUC value of CA-724 and two lncRNAs was 0.751. Circulating lncRNAs ZFPM2-AS1 and XIST can serve as the candidate plasma biomarkers used to diagnose GC.

Age-adjusted Charlson comorbidity index is associated with the risk of osteoporosis in older fall-prone men: a retrospective cohort study

BackgroundThere is growing evidence linking the age-adjusted Charlson comorbidity index (aCCI), an assessment tool for multimorbidity, to fragility fracture and fracture-related postoperative complications. However, the role of multimorbidity in osteoporosis has not yet been thoroughly evaluated. We aimed to investigate the association between aCCI and the risk of osteoporosis in older adults at moderate to high risk of falling.MethodsA total of 947 men were included from January 2015 to August 2022 in a hospital in Beijing, China. The aCCI was calculated by counting age and each comorbidity according to their weighted scores, and the participants were stratified into two groups by aCCI: low (aCCI < 5), and high (aCCI ≥5). The Kaplan Meier method was used to assess the cumulative incidence of osteoporosis by different levels of aCCI. The Cox proportional hazards regression model was used to estimate the association of aCCI with the risk of osteoporosis. Receiver operating characteristic (ROC) curve was adapted to assess the performance for aCCI in osteoporosis screening.ResultsAt baseline, the mean age of all patients was 75.7 years, the mean BMI was 24.8 kg/m2, and 531 (56.1%) patients had high aCCI while 416 (43.9%) were having low aCCI. During a median follow-up of 6.6 years, 296 participants developed osteoporosis. Kaplan–Meier survival curves showed that participants with high aCCI had significantly higher cumulative incidence of osteoporosis compared with those had low aCCI (log-rank test: P < 0.001). When aCCI was examined as a continuous variable, the multivariable-adjusted model showed that the osteoporosis risk increased by 12.1% (HR = 1.121, 95% CI 1.041–1.206, P = 0.002) as aCCI increased by one unit. When aCCI was changed to a categorical variable, the multivariable-adjusted hazard ratios associated with different levels of aCCI [low (reference group) and high] were 1.00 and 1.557 (95% CI 1.223–1.983) for osteoporosis (P < 0.001), respectively. The aCCI (cutoff ≥5) revealed an area under ROC curve (AUC) of 0.566 (95%CI 0.527–0.605, P = 0.001) in identifying osteoporosis in older fall-prone men, with sensitivity of 64.9% and specificity of 47.9%.ConclusionsThe current study indicated an association of higher aCCI with an increased risk of osteoporosis among older fall-prone men, supporting the possibility of aCCI as a marker of long-term skeletal-related adverse clinical outcomes.