Area Under The Receiver Operating Characteristic Research Articles

T1 mapping as a quantitative imaging biomarker for diagnosing cervical cancer: a comparison with diffusion kurtosis imaging

BackgroundT1 mapping can potentially quantitatively assess the intrinsic properties of tumors. This study was conducted to explore the ability of T1 mapping in distinguishing cervical cancer type, grade, and stage and compare the diagnostic performance of T1 mapping with diffusion kurtosis imaging (DKI).MethodsOne hundred fifty-seven patients with pathologically confirmed cervical cancer were enrolled in this prospectively study. T1 mapping and DKI were performed. The native T1, difference between native and postcontrast T1 (T1diff), mean kurtosis (MK), mean diffusivity (MD), and apparent diffusion coefficient (ADC) were calculated. Cervical squamous cell carcinoma (CSCC) and adenocarcinoma (CAC), low- and high-grade carcinomas, and early- and advanced-stage groups were compared using area under the receiver operating characteristic (AUROC) curves.ResultsThe native T1 and MK were higher, and the MD and ADC were lower for CSCC than for CAC (all p < 0.05). Compared with low-grade CSCC, high-grade CSCC had decreased T1diff, MD, ADC, and increased MK (p < 0.05). Compared with low-grade CAC, high-grade CAC had decreased T1diff and increased MK (p < 0.05). Native T1 was significantly higher in the advanced-stage group than in the early-stage group (p < 0.05). The AUROC curves of native T1, MK, ADC and MD were 0,772, 0.731, 0.715, and 0.627, respectively, for distinguishing CSCC from CAC. The AUROC values were 0.762 between high- and low-grade CSCC and 0.835 between high- and low-grade CAC, with T1diff and MK showing the best discriminative values, respectively. For distinguishing between advanced-stage and early-stage cervical cancer, only the AUROC of native T1 was statistically significant (AUROC = 0.651, p = 0.002).ConclusionsCompared with DKI-derived parameters, native T1 exhibits better efficacy for identifying cervical cancer subtype and stage, and T1diff exhibits comparable discriminative value for cervical cancer grade.

Prediction of adverse events risk in patients with comorbid post-traumatic stress disorder and alcohol use disorder using electronic medical records by deep learning models

BackgroundIdentifying co-occurring mental disorders and elevated risk is vital for optimization of healthcare processes. In this study, we will use DeepBiomarker2, an updated version of our deep learning model to predict the adverse events among patients with comorbid post-traumatic stress disorder (PTSD) and alcohol use disorder (AUD), a high-risk population. MethodsWe analyzed electronic medical records of 5565 patients from University of Pittsburgh Medical Center to predict adverse events (opioid use disorder, suicide related events, depression, and death) within 3 months at any encounter after the diagnosis of PTSD+AUD by using DeepBiomarker2. We integrated multimodal information including: lab tests, medications, co-morbidities, individual and neighborhood level social determinants of health (SDoH), psychotherapy and veteran data. ResultsDeepBiomarker2 achieved an area under the receiver operator curve (AUROC) of 0.94 on the prediction of adverse events among those PTSD+AUD patients. Medications such as vilazodone, dronabinol, tenofovir, suvorexant, modafinil, and lamivudine showed potential for risk reduction. SDoH parameters such as cognitive behavioral therapy and trauma focused psychotherapy lowered risk while active veteran status, income segregation, limited access to parks and greenery, low Gini index, limited English-speaking capacity, and younger patients increased risk. ConclusionsOur improved version of DeepBiomarker2 demonstrated its capability of predicting multiple adverse event risk with high accuracy and identifying potential risk and beneficial factors.

Advanced Sampling Technique in Radiology Free-Text Data for Efficiently Building Text Mining Models by Deep Learning in Vertebral Fracture.

This study aims to establish advanced sampling methods in free-text data for efficiently building semantic text mining models using deep learning, such as identifying vertebral compression fracture (VCF) in radiology reports. We enrolled a total of 27,401 radiology free-text reports of X-ray examinations of the spine. The predictive effects were compared between text mining models built using supervised long short-term memory networks, independently derived by four sampling methods: vector sum minimization, vector sum maximization, stratified, and simple random sampling, using four fixed percentages. The drawn samples were applied to the training set, and the remaining samples were used to validate each group using different sampling methods and ratios. The predictive accuracy was measured using the area under the receiver operating characteristics (AUROC) to identify VCF. At the sampling ratios of 1/10, 1/20, 1/30, and 1/40, the highest AUROC was revealed in the sampling methods of vector sum minimization as confidence intervals of 0.981 (95%CIs: 0.980-0.983)/0.963 (95%CIs: 0.961-0.965)/0.907 (95%CIs: 0.904-0.911)/0.895 (95%CIs: 0.891-0.899), respectively. The lowest AUROC was demonstrated in the vector sum maximization. This study proposes an advanced sampling method, vector sum minimization, in free-text data that can be efficiently applied to build the text mining models by smartly drawing a small amount of critical representative samples.

Non-Negative matrix factorization combined with kernel regression for the prediction of adverse drug reaction profiles.

Post-market unexpected Adverse Drug Reactions (ADRs) are associated with significant costs, in both financial burden and human health. Due to the high cost and time required to run clinical trials, there is significant interest in accurate computational methods that can aid in the prediction of ADRs for new drugs. As a machine learning task, ADR prediction is made more challenging due to a high degree of class imbalance and existing methods do not successfully balance the requirement to detect the minority cases (true positives for ADR), as measured by the Area Under the Precision-Recall (AUPR) curve with the ability to separate true positives from true negatives [as measured by the Area Under the Receiver Operating Characteristic (AUROC) curve]. Surprisingly, the performance of most existing methods is worse than a naïve method that attributes ADRs to drugs according to the frequency with which the ADR has been observed over all other drugs. The existing advanced methods applied do not lead to substantial gains in predictive performance. We designed a rigorous evaluation to provide an unbiased estimate of the performance of ADR prediction methods: Nested Cross-Validation and a hold-out set were adopted. Among the existing methods, Kernel Regression (KR) performed best in AUPR but had a disadvantage in AUROC, relative to other methods, including the naïve method. We proposed a novel method that combines non-negative matrix factorization with kernel regression, called VKR. This novel approach matched or exceeded the performance of existing methods, overcoming the weakness of the existing methods. Code and data are available on https://github.com/YezhaoZhong/VKR.

Predictors of micronutrient deficiency among children aged 6-23 months in Ethiopia: a machine learning approach.

Micronutrient (MN) deficiencies are a major public health problem in developing countries including Ethiopia, leading to childhood morbidity and mortality. Effective implementation of programs aimed at reducing MN deficiencies requires an understanding of the important drivers of suboptimal MN intake. Therefore, this study aimed to identify important predictors of MN deficiency among children aged 6-23 months in Ethiopia using machine learning algorithms. This study employed data from the 2019 Ethiopia Mini Demographic and Health Survey (2019 EMDHS) and included a sample of 1,455 children aged 6-23 months for analysis. Machine Learning (ML) methods including, Support Vector Machine (SVM), Logistic Regression (LR), Random Forest (RF), Neural Network (NN), and Naïve Bayes (NB) were used to prioritize risk factors for MN deficiency prediction. Performance metrics including accuracy, sensitivity, specificity, and Area Under the Receiver Operating Characteristic (AUROC) curves were used to evaluate model prediction performance. The prediction performance of the RF model was the best performing ML model in predicting child MN deficiency, with an AUROC of 80.01% and accuracy of 72.41% in the test data. The RF algorithm identified the eastern region of Ethiopia, poorest wealth index, no maternal education, lack of media exposure, home delivery, and younger child age as the top prioritized risk factors in their order of importance for MN deficiency prediction. The RF algorithm outperformed other ML algorithms in predicting child MN deficiency in Ethiopia. Based on the findings of this study, improving women's education, increasing exposure to mass media, introducing MN-rich foods in early childhood, enhancing access to health services, and targeted intervention in the eastern region are strongly recommended to significantly reduce child MN deficiency.

InterDILI: interpretable prediction of drug-induced liver injury through permutation feature importance and attention mechanism

Safety is one of the important factors constraining the distribution of clinical drugs on the market. Drug-induced liver injury (DILI) is the leading cause of safety problems produced by drug side effects. Therefore, the DILI risk of approved drugs and potential drug candidates should be assessed. Currently, in vivo and in vitro methods are used to test DILI risk, but both methods are labor-intensive, time-consuming, and expensive. To overcome these problems, many in silico methods for DILI prediction have been suggested. Previous studies have shown that DILI prediction models can be utilized as prescreening tools, and they achieved a good performance. However, there are still limitations in interpreting the prediction results. Therefore, this study focused on interpreting the model prediction to analyze which features could potentially cause DILI. For this, five publicly available datasets were collected to train and test the model. Then, various machine learning methods were applied using substructure and physicochemical descriptors as inputs and the DILI label as the output. The interpretation of feature importance was analyzed by recognizing the following general-to-specific patterns: (i) identifying general important features of the overall DILI predictions, and (ii) highlighting specific molecular substructures which were highly related to the DILI prediction for each compound. The results indicated that the model not only captured the previously known properties to be related to DILI but also proposed a new DILI potential substructural of physicochemical properties. The models for the DILI prediction achieved an area under the receiver operating characteristic (AUROC) of 0.88–0.97 and an area under the Precision-Recall curve (AUPRC) of 0.81–0.95. From this, we hope the proposed models can help identify the potential DILI risk of drug candidates at an early stage and offer valuable insights for drug development.

Relationship between Neutrophil/Lymphocyte Ratio (NLR), Platelet/Lymphocyte Ratio (PLR) and Thrombosis in Patients with Portal Vein Thrombosis (PVT) without Acquired Risk Factor for Thrombosis.

Inflammation occurring after vascular endothelial damage plays a role in thrombus formation. Changes in various blood parameters that develop after the inflammatory condition can be used as a marker to predict thrombus. This study aimed to investigate the relationship between the neutrophil/lymphocyte ratio (NLR), platelet/lymphocyte ratio (PLR), and portal vein thrombosis (PVT). After applying the exclusion and inclusion criteria to the patients diagnosed with PVT and followed up in our center between January 2006 and May 2018, a total of 38 patients without acquired risk factors for the development of PVT and 52 healthy controls were included in the study. Clinical features and NLR and PLR at diagnosis were evaluated. NLR and PLR values were detected to be significantly higher in patients diagnosed with PVT compared to the control group (P < 0.001 for NLR, P < 0.001 for PLR). Findings were as follows: In acute PVT patients for NLR = 3.645 (area under the receiver operating characteristic (AUROC) 0.886, sensitivity 69.2%, specificity 96.2%, P < 0.001), for PLR = 196.24 (AUROC 0.754, sensitivity 53.2%, specificity 96.2%, P = 0.005), while in chronic PVT patients, for NLR = 3.645 (AUROC 0.744, sensitivity 40%, specificity 96.2%, P = 0.001), and for PLR = 195.93 (AUROC 0.715, sensitivity 44%, specificity 96.2%, P = 0.002). NLR and PLR were associated with the diagnosis of PVT. In PVT patients, NLR and PLR values were observed to be significantly higher than the control group. In our study, the relationship between NLR and PLR in patients with noncirrhotic, nonmalignant PVT without acquired risk factors for thrombosis was shown for the first time.

Blood glucose and lactate levels as early predictive markers in patients presenting with cardiogenic shock: A retrospective cohort study.

Lactate and glucose are widely used biochemical parameters in current predictive risk scores for cardiogenic shock. Data regarding the relationship between lactate and glucose levels in cardiogenic shock are limited. Thus, we aimed to analyze glucose and lactate as early markers for in-hospital mortality in cardiogenic shock. In this retrospective cohort study, 312 patients presenting with cardiogenic shock to a tertiary-care hospital between 2016 and 2018 were included. Apparent cardiogenic shock was defined as hypoperfusion with hemodynamic compromise and biochemical marker increase due to diminished tissue perfusion, corresponding to SCAI shock stages. In-hospital mortality was assessed as the primary endpoint. The median age of the study population was 71 (60-79) years and the etiology of cardiogenic shock was acute myocardial infarction in 45.8%. Overall in-hospital mortality was 67.6%. In the receiver operating curve analysis, the area under the receiver-operating curve (AUC) for prediction of in-hospital mortality was higher for lactate (AUC: 0.757) than for glucose (AUC: 0.652). Both values were significantly associated with outcome (groups created with best cutoff values obtained from the Youden index). Correlation analysis showed a significant non-linear association of both values. In a multivariable stepwise Cox regression analysis, lactate remained an independent predictor for in-hospital mortality, whilst glucose, despite being implicated in energy metabolism, was not independently predictive for mortality. Together, these data suggest that lactate at admission is superior for mortality prediction in patients with apparent cardiogenic shock. Glucose was not independently predictive for mortality.

Application of a transparent artificial intelligence algorithm for US adults in the obese category of weight.

Identification of associations between the obese category of weight in the general US population will continue to advance our understanding of the condition and allow clinicians, providers, communities, families, and individuals make more informed decisions. This study aims to improve the prediction of the obese category of weight and investigate its relationships with factors, ultimately contributing to healthier lifestyle choices and timely management of obesity. Questionnaires that included demographic, dietary, exercise and health information from the US National Health and Nutrition Examination Survey (NHANES 2017-2020) were utilized with BMI 30 or higher defined as obesity. A machine learning model, XGBoost predicted the obese category of weight and Shapely Additive Explanations (SHAP) visualized the various covariates and their feature importance. Model statistics including Area under the receiver operator curve (AUROC), sensitivity, specificity, positive predictive value, negative predictive value and feature properties such as gain, cover, and frequency were measured. SHAP explanations were created for transparent and interpretable analysis. There were 6,146 adults (age > 18) that were included in the study with average age 58.39 (SD = 12.94) and 3122 (51%) females. The machine learning model had an Area under the receiver operator curve of 0.8295. The top four covariates include waist circumference (gain = 0.185), GGT (gain = 0.101), platelet count (gain = 0.059), AST (gain = 0.057), weight (gain = 0.049), HDL cholesterol (gain = 0.032), and ferritin (gain = 0.034). In conclusion, the utilization of machine learning models proves to be highly effective in accurately predicting the obese category of weight. By considering various factors such as demographic information, laboratory results, physical examination findings, and lifestyle factors, these models successfully identify crucial risk factors associated with the obese category of weight.

Predicting Brain Amyloid Status Using the National Institute of Health Toolbox (NIHTB) for Assessment of Neurological and Behavioral Function.

Amyloid-beta (Aβ) plaque is a neuropathological hallmark of Alzheimer's disease (AD). As anti-amyloid monoclonal antibodies enter the market, predicting brain amyloid status is critical to determine treatment eligibility. To predict brain amyloid status utilizing machine learning approaches in the Advancing Reliable Measurement in Alzheimer's Disease and Cognitive Aging (ARMADA) study. ARMADA is a multisite study that implemented the National Institute of Health Toolbox for Assessment of Neurological and Behavioral Function (NIHTB) in older adults with different cognitive ability levels (normal, mild cognitive impairment, early-stage dementia of the AD type). Participants across various sites were involved in the ARMADA study for validating the NIHTB. 199 ARMADA participants had either PET or CSF information (mean age 76.3 ± 7.7, 51.3% women, 42.3% some or complete college education, 50.3% graduate education, 88.9% White, 33.2% with positive AD biomarkers). We used cognition, emotion, motor, sensation scores from NIHTB, and demographics to predict amyloid status measured by PET or CSF. We applied LASSO and random forest models and used the area under the receiver operating curve (AUROC) to evaluate the ability to identify amyloid positivity. The random forest model reached AUROC of 0.74 with higher specificity than sensitivity (AUROC 95% CI:0.73 - 0.76, Sensitivity 0.50, Specificity 0.88) on the held-out test set; higher than the LASSO model (0.68 (95% CI:0.68 - 0.69)). The 10 features with the highest importance from the random forest model are: picture sequence memory, cognition total composite, cognition fluid composite, list sorting working memory, words-in-noise test (hearing), pattern comparison processing speed, odor identification, 2-minutes-walk endurance, 4-meter walk gait speed, and picture vocabulary. Overall, our model revealed the validity of measurements in cognition, motor, and sensation domains, in associating with AD biomarkers. Our results support the utilization of the NIH toolbox as an efficient and standardizable AD biomarker measurement that is better at identifying amyloid negative (i.e., high specificity) than positive cases (i.e., low sensitivity).

Advancing precision prognostication in neuro-oncology: Machine learning models for data-driven personalized survival predictions in IDH-wildtype glioblastoma.

Glioblastoma (GBM) remains associated with a dismal prognoses despite standard therapies. While population-level survival statistics are established, generating individualized prognosis remains challenging. We aim to develop machine learning (ML) models that generate personalized survival predictions for GBM patients to enhance prognostication. Adult patients with histologically confirmed IDH-wildtype GBM from the National Cancer Database (NCDB) were analyzed. ML models were developed with TabPFN, TabNet, XGBoost, LightGBM, and Random Forest algorithms to predict mortality at 6, 12, 18, and 24 months postdiagnosis. SHapley Additive exPlanations (SHAP) were employed to enhance the interpretability of the models. Models were primarily evaluated using the area under the receiver operating characteristic (AUROC) values, and the top-performing models indicated by the highest AUROCs for each outcome were deployed in a web application that was created for individualized predictions. A total of 7537 patients were retrieved from the NCDB. Performance evaluation revealed the top-performing models for each outcome were built using the TabPFN algorithm. The TabPFN models yielded mean AUROCs of 0.836, 0.78, 0.732, and 0.724 in predicting 6, 12, 18, and 24 month mortality, respectively. This study establishes ML models tailored to individual patients to enhance GBM prognostication. Future work should focus on external validation and dynamic updating as new data emerge.

Prototypical Learning Guided Context-Aware Segmentation Network for Few-Shot Anomaly Detection.

Few-shot anomaly detection (FSAD) denotes the identification of anomalies within a target category with a limited number of normal samples. Existing FSAD methods largely rely on pretrained feature representations to detect anomalies, but the inherent domain gap between pretrained representations and target FSAD scenarios is often overlooked. This study proposes a prototypical learning-guided context-aware segmentation network (PCSNet) to address the domain gap, thereby improving feature descriptiveness in target scenarios and enhancing FSAD performance. In particular, PCSNet comprises a prototypical feature adaption (PFA) subnetwork and a context-aware segmentation (CAS) subnetwork. PFA extracts prototypical features as guidance to ensure better feature compactness for normal data while distinct separation from anomalies. A pixel-level disparity classification (PDC) loss is also designed to make subtle anomalies more distinguishable. Then a CAS subnetwork is introduced for pixel-level anomaly localization, where pseudo anomalies are exploited to facilitate the training process. Experimental results on MVTec AD and metal part defect detection (MPDD) demonstrate the superior FSAD performance of PCSNet, with 94.9% and 80.2% image-level area under the receiver operating characteristics (AUROCs) in an eight-shot scenario, respectively. Real-world applications on automotive plastic part inspection further demonstrate that PCSNet can achieve promising results with limited training samples. The code is available at https://github.com/yuxin-jiang/PCSNet.

Surface-enhanced Raman spectroscopy for characterization of filtrates of blood serum samples from patients with tuberculosis obtained by 50 kDa filtration devices.

The ability of surface-enhanced Raman spectroscopy (SERS) to generate spectroscopic fingerprints has made it an emerging tool for biomedical applications. The objective of this study is to confirm the potential use of Raman spectroscopy for early disease diagnosis based on blood serum. In this study, a total of sixty blood serum samples, consisting of forty from diseased patients and twenty (controls) from healthy individuals, was used. Because disease biomarkers, found in the lower molecular weight fraction, are suppressed by higher molecular weight proteins, 50 kDa Amicon ultrafiltration centrifugation devices were used to produce two fractions from whole blood serum consisting of a filtrate, which is a low molecular weight fraction, and a residue, which is a high molecular weight fraction. These fractions were then analyzed, and their SERS spectral data were compared with those of healthy fractions. The SERS technique was utilized on blood serum, filtrate and residue of patients with tuberculosis to identify characteristic SERS spectral features associated with the development of disease, which can be used to differentiate them from healthy samples using silver nanoparticles as a SERS substrate. For further analysis, the effective chemometric technique of principal component analysis (PCA) was used to qualitatively differentiate all the analyzed samples based on their SERS spectral features. Partial least squares discriminant analysis (PLS-DA) accurately classified the filtrate portions of healthy and tuberculosis samples with 97% accuracy, 97% specificity, 98% sensitivity, and an area under the receiver operating characteristic (AUROC) curve of 0.74.

Groundwater potential zone mapping using GIS and Remote Sensing based models for sustainable groundwater management

The present research is conducted in the southern region of Khyber Pakhtunkhwa, Pakistan, to identify groundwater potential zones (GWPZ). We used three models including Weight of Evidence (WOE), Frequency Ratio (FR), and Information Value (IV) with twelve parameters (elevation, slope, aspect, curvature, drainage network, LULC, precipitation, geology, Lineament, NDVI, road, and soil texture, that have been prepared and integrated into ArcGIS 10.8. The reliability of the applied models’ results was validated using Area Under the Receiver Operating Characteristics (AUROC). The GWPZ were reclassified into five classes, i.e. very low, low, medium, high, and very high zone. The area occupied by mentioned classes using WOE are very low (10.14%), low (19.58%), medium (26.75%), high (27.10%), very high (16.40%), while using FR are very low (20.93%), low (32.38%), medium (18.92%), high (13.13%), very high (14.61%) and using IV are very low (14.41%), low (17.17%), medium (29.01%), high (25.85%), and very High (13.53%). The Success Rate Curve of WOE, FR, and IV were 0.86, 0.91, and 0.87, while the Predicted Rate Curve values were 0.89, 0.93, and 0.90, respectively. The results revealed that all applied statistical models performed very well to delineate GWPZ. However, use of the FR technique is strongly encouraged to evaluate the GWPZ, and its findings are especially useful for managing groundwater resources in urban planning. Our approaches for assessing the GWPZ mapping can be applied in any region with similar scenarios and are recommended as a helpful tool for policymakers to manage groundwater.

Using Advanced Convolutional Neural Network Approaches to Reveal Patient Age, Gender, and Weight Based on Tongue Images.

The human tongue has been long believed to be a window to provide important insights into a patient's health in medicine. The present study introduced a novel approach to predict patient age, gender, and weight inferences based on tongue images using pretrained deep convolutional neural networks (CNNs). Our results demonstrated that the deep CNN models (e.g., ResNeXt) trained on dorsal tongue images produced excellent results for age prediction with a Pearson correlation coefficient of 0.71 and a mean absolute error (MAE) of 8.5 years. We also obtained an excellent classification of gender, with a mean accuracy of 80% and an AUC (area under the receiver operating characteristic curve) of 88%. ResNeXt model also obtained a moderate level of accuracy for weight prediction, with a Pearson correlation coefficient of 0.39 and a MAE of 9.06 kg. These findings support our hypothesis that the human tongue contains crucial information about a patient. This study demonstrated the feasibility of using the pretrained deep CNNs along with a large tongue image dataset to develop computational models to predict patient medical conditions for noninvasive, convenient, and inexpensive patient health monitoring and diagnosis.

Predictive Value of Perfusion Index for Assessing Success of Supraclavicular Brachial Plexus Block: A Prospective Observational Study

Abstract Background: Perfusion index (PI) is a simple, objective, and noninvasive method for evaluating the success of brachial plexus blocks. There is only one study which assessed the time point at which the PI had the best predictive value. Aims: Of the 5 measured outcome variables (i.e.; PI at baseline, at 5 min, at 10 min, and at 15 min and PI ratio), we wanted to determine the one which had the best predictive value for block success. Materials and Methods: It is a prospective observational study done in a tertiary care teaching hospital. Sixty-nine patients of either sex, American Society of Anesthesiologists Physical Status I and II, between the ages of 18 and 65 years posted for elective upper limb surgery were included. Patients were given supraclavicular blocks with a peripheral nerve stimulator. PI recorded at baseline, 5 min, 10 min, and 15 min. A PI ratio was calculated. Sensory and motor blocks were assessed at 5-min intervals. Statistical Analysis: Descriptive analysis was applied by mean and standard deviation for quantitative variables and frequency and proportion for categorical variables. Receiver operating characteristic (ROC) curves were constructed. SPSS version 22 was used to detect an area under the ROCs (AUROC) curve and calculated to assess how good a test PI at 10 and 15 min and PI ratio are in predicting the outcome of a block. Results: The mean PI increased continuously from the baseline till 15 min in successful blocks, but in unsuccessful blocks, the rise was not seen. ROC curves showed an AUROC curve of 0.93 in case of PI at 15 min and 0.84 for PI ratio. Conclusion: We conclude that PI at 15 min is the best in our study and PI ratio is the next best as a predictor for evaluating success of supraclavicular blocks.

Advancing delirium classification: A clinical notes-based natural language processing-supported machine learning model

ObjectiveThe study of the epidemiology of delirium in hospitalized patients is challenging. We aimed to identify the presence or absence of delirium from clinical text notes using natural language processing (NLP) techniques and machine learning (ML) models. Materials and methodsWe developed a delirium predictive model using 942 clinical notes from hospitalized patients with an ICD-10 delirium hospital discharge code. Moreover, we implemented ML models using a) delirium-suggestive words from an expert-defined dictionary or b) free text in clinical notes. Both strategies considered positive and negative delirium-associated words. ResultsAt the note level, for the dictionary method, the logistic regression model achieved an area under the receiver-operating curve (AUROC) of 0.917 for positive words and 0.914 for combined positive and negative words. The areas under the precision-recall curve (AUPR) were 0.893 and 0.897, respectively. For the free-text method, the model achieved an AUROC of 0.826 and 0.830 and AUPR of 0.852 and 0.856, respectively. DiscussionNLP-based ML models accurately identified the presence of delirium in clinical notes. The dictionary-based method was superior to the free-text method. The use of negative features improved performance in both methods. ConclusionOur proposed NLP-based ML model identified delirium in clinical notes. This model could automatically screen millions of notes and facilitate the study of the epidemiology of in-hospital delirium.

A Machine Learning Approach to Understanding the Physical Properties of Magnetic Flux Ropes in the Solar Wind at 1 au

Interplanetary magnetic flux ropes (MFRs) are commonly observed structures in the solar wind, categorized as magnetic clouds (MCs) and small-scale MFRs (SMFRs) depending on whether they are associated with coronal mass ejections. We apply machine learning to systematically compare SMFRs, MCs, and ambient solar wind plasma properties. We construct a data set of 3-minute averaged sequential data points of the solar wind’s instantaneous bulk fluid plasma properties using about 20 years of measurements from Wind. We label samples by the presence and type of MFRs containing them using a catalog based on Grad–Shafranov (GS) automated detection for SMFRs and NASA's catalog for MCs (with samples in neither labeled non-MFRs). We apply the random forest machine learning algorithm to find which categories can be more easily distinguished and by what features. MCs were distinguished from non-MFRs with an area under the receiver-operator curve (AUC) of 94% and SMFRs with an AUC of 89%, and had distinctive plasma properties. In contrast, while SMFRs were distinguished from non-MFRs with an AUC of 86%, this appears to rely solely on the 〈B〉 > 5 nT threshold applied by the GS catalog. The results indicate that SMFRs have virtually the same plasma properties as the ambient solar wind, unlike the distinct plasma regimes of MCs. We interpret our findings as additional evidence that most SMFRs at 1 au are generated within the solar wind. We also suggest that they should be considered a salient feature of the solar wind’s magnetic structure rather than transient events.

Prediction of tuberculosis clusters in the riverine municipalities of the Brazilian Amazon with machine learning.

Tuberculosis (TB) is the second most deadly infectious disease globally, posing a significant burden in Brazil and its Amazonian region. This study focused on the "riverine municipalities" and hypothesizes the presence of TB clusters in the area. Wealso aimed to train a machine learning model to differentiate municipalities classified as hot spots vs. non-hot spots using disease surveillance variables as predictors. Data regarding the incidence of TB from 2019 to 2022 in the riverine town was collected from the Brazilian Health Ministry Informatics Department. Moran's I was used to assess global spatial autocorrelation, while the Getis-Ord GI* method was employed to detect high and low-incidence clusters. A Random Forest machine-learning model was trained using surveillance variables related to TB cases to predict hot spots among non-hot spot municipalities. Our analysis revealed distinct geographical clusters with high and low TB incidence following a west-to-east distribution pattern. The Random Forest Classification model utilizes six surveillance variables to predict hot vs. non-hot spots. The machine learning model achieved an Area Under the Receiver Operator Curve (AUC-ROC) of 0.81. Municipalities with higher percentages of recurrent cases, deaths due to TB, antibiotic regimen changes, percentage of new cases, and cases with smoking history were the best predictors of hot spots. This prediction method can be leveraged to identify the municipalities at the highest risk of being hot spots for the disease, aiding policymakers with an evidenced-based tool to direct resource allocation for disease control in the riverine municipalities.

A Dual-Camera Eye-Tracking Platform for Rapid Real-Time Diagnosis of Acute Delirium: A Pilot Study.

Delirium, an acute confusional state, affects 20-80% of patients in Intensive Care Units (ICUs), one in three medically hospitalized patients, and up to 50% of all patients who have had surgery. Its development is associated with short- and long-term morbidity, and increased risk of death. Yet, we lack any rapid, objective, and automated method to diagnose delirium. Here, we detail the prospective deployment of a novel dual-camera contextual eye-tracking platform. We then use the data from this platform to contemporaneously classify delirium. We recruited 42 patients, resulting in 210 (114 with delirium, 96 without) recordings of hospitalized patients in ICU across two centers, as part of a prospective multi-center feasibility pilot study. All recordings made with our platform were usable for analysis. We divided the collected data into training and validation cohorts based on the data originating center. We trained two Temporal Convolutional Network (TCN) models that can classify delirium using a pre-existing manual scoring system (Confusion Assessment Method in ICU (CAM-ICU)) as the training target. The first model uses eye movements only which achieves an Area Under the Receiver Operator Curve (AUROC) of 0.67 and a mean Average Precision (mAP) of 0.68. The second model uses the point of regard, the part of the scene the patient is looking at, and increases the AUROC to 0.76 and the mAP to 0.81. These models are the first to classify delirium using continuous non-invasive eye-tracking but will require further clinical prospective validation prior to use as a decision-support tool. Eye-tracking is a biological signal that can be used to identify delirium in patients in ICU. The platform, alongside the trained neural networks, can automatically, objectively, and continuously classify delirium aiding in the early detection of the deteriorating patient. Future work is aimed at prospective evaluation and clinical translation.