Concept: Binary classification
INTRODUCTION: We previously derived and validated the AIMS65 score, a mortality prognostic scale for upper GI bleeding (UGIB). OBJECTIVE: To validate the AIMS65 score in a different patient population and compare it with the Glasgow-Blatchford risk score (GBRS). DESIGN: Retrospective cohort study. PATIENTS: Adults with a primary diagnosis of UGIB. MAIN OUTCOME MEASUREMENTS: Primary outcome: inpatient mortality. Secondary outcomes: composite clinical endpoint of inpatient mortality, rebleeding, and endoscopic, radiologic or surgical intervention; blood transfusion; intensive care unit admission; rebleeding; length of stay; timing of endoscopy. The area under the receiver-operating characteristic curve (AUROC) was calculated for each score. RESULTS: Of the 278 study patients, 6.5% died and 35% experienced the composite clinical endpoint. The AIMS65 score was superior in predicting inpatient mortality (AUROC, 0.93 vs 0.68; P < .001), whereas the GBRS was superior in predicting blood transfusions (AUROC, 0.85 vs 0.65; P < .01) The 2 scores were similar in predicting the composite clinical endpoint (AUROC, 0.62 vs 0.68; P = .13) as well as the secondary outcomes. A GBRS of 10 and 12 or more maximized the sum of the sensitivity and specificity for inpatient mortality and rebleeding, respectively. The cutoff was 2 or more for the AIMS65 score for both outcomes. LIMITATIONS: Retrospective, single-center study. CONCLUSION: The AIMS65 score is superior to the GBRS in predicting inpatient mortality from UGIB, whereas the GBRS is superior for predicting blood transfusion. Both scores are similar in predicting the composite clinical endpoint and other outcomes in clinical care and resource use.
Bronchoscopy is frequently used for the evaluation of suspicious pulmonary lesions found on computed tomography, but its sensitivity for detecting lung cancer is limited. Recently, a bronchial genomic classifier was validated to improve the sensitivity of bronchoscopy for lung cancer detection, demonstrating a high sensitivity and negative predictive value among patients at intermediate risk (10-60 %) for lung cancer with an inconclusive bronchoscopy. Our objective for this study was to determine if a negative genomic classifier result that down-classifies a patient from intermediate risk to low risk (<10 %) for lung cancer would reduce the rate that physicians recommend more invasive testing among patients with an inconclusive bronchoscopy.
The accurate diagnosis of asbestos-related diseases is important because of past and current asbestos exposures. This study evaluated the reliability of clinical diagnoses of asbestos-related diseases in former mineworkers using autopsies as the reference standard. Sensitivity, specificity, positive predictive value and negative predictive value were calculated. The 149 cases identified had clinical examinations 0.3-7.4 years before death. More asbestos-related diseases were diagnosed at autopsy rather than clinically: 77 versus 52 for asbestosis, 27 versus 14 for mesothelioma and 22 versus 3 for lung cancer. Sensitivity and specificity values for clinical diagnoses were 50.6% and 81.9% for asbestosis, 40.7% and 97.5% for mesothelioma, and 13.6% and 100.0% for lung cancer. False-negative diagnoses of asbestosis were more likely using radiographs of acceptable (versus good) quality and in cases with pulmonary tuberculosis at autopsy. The low sensitivity values are indicative of the high proportion of false-negative diagnoses. It is unlikely that these were the result of disease manifestation between the last clinical assessment and autopsy. Where clinical features suggest asbestos-related diseases but the chest radiograph is negative, more sophisticated imaging techniques or immunohistochemistry for asbestos-related cancers should be used. Autopsies are useful for the detection of previously undiagnosed and misdiagnosed asbestos-related diseases, and for monitoring clinical practice and delivery of compensation.
In this study, we aimed to examine the clinical value of the pleural effusion (PE) biomarkers, soluble mesothelin-related peptide (SMRP), cytokeratin 19 fragment (CYFRA 21-1) and carcinoembryonic antigen (CEA), and the utility of combining chest computed tomography (CT) findings with these biomarkers, in diagnosing malignant pleural mesothelioma (MPM). We conducted a retrospective cohort study in a single center. Consecutive patients with undiagnosed pleural effusions who underwent PE analysis between September 2014 and August 2016 were reviewed. This study included 240 patients (32 with MPM and 208 non-MPM). SMRP and the CYFRA 21-1/CEA ratio had a sensitivity and specificity for diagnosing MPM of 56.3% and 86.5%, and 87.5% and 74.0%, respectively. Using receiver operating characteristics (ROC) curve analysis of the ability of these markers to distinguish MPM from all other PE causes, the area under the ROC curve (AUC) for SMRP and the CYFRA 21-1/CEA ratio was 0.804 and 0.874, respectively. The sensitivity and specificity of SMRP combined with the CYFRA 21-1/CEA ratio were 93.8% and 64.9%, respectively. The sensitivity of the combination of SMRP, the CYFRA 21-1/CEA ratio, and the presence of Leung’s criteria (a chest CT finding that is suggestive of malignant pleural disease) was 93.8%. In conclusion, the combined PE biomarkers had a high sensitivity for diagnosing MPM, although the addition of chest CT findings did not improve the sensitivity of SMRP combined with the CYFRA 21-1/CEA ratio. Combination of these biomarkers helped to rule out MPM effectively among patients at high risk of suffering MPM and would be valuable especially for old frail patients who have difficulty in undergoing invasive procedures such as thoracoscopy.
Label-free cell analysis is essential to personalized genomics, cancer diagnostics, and drug development as it avoids adverse effects of staining reagents on cellular viability and cell signaling. However, currently available label-free cell assays mostly rely only on a single feature and lack sufficient differentiation. Also, the sample size analyzed by these assays is limited due to their low throughput. Here, we integrate feature extraction and deep learning with high-throughput quantitative imaging enabled by photonic time stretch, achieving record high accuracy in label-free cell classification. Our system captures quantitative optical phase and intensity images and extracts multiple biophysical features of individual cells. These biophysical measurements form a hyperdimensional feature space in which supervised learning is performed for cell classification. We compare various learning algorithms including artificial neural network, support vector machine, logistic regression, and a novel deep learning pipeline, which adopts global optimization of receiver operating characteristics. As a validation of the enhanced sensitivity and specificity of our system, we show classification of white blood T-cells against colon cancer cells, as well as lipid accumulating algal strains for biofuel production. This system opens up a new path to data-driven phenotypic diagnosis and better understanding of the heterogeneous gene expressions in cells.
Binary classifiers are routinely evaluated with performance measures such as sensitivity and specificity, and performance is frequently illustrated with Receiver Operating Characteristics (ROC) plots. Alternative measures such as positive predictive value (PPV) and the associated Precision/Recall (PRC) plots are used less frequently. Many bioinformatics studies develop and evaluate classifiers that are to be applied to strongly imbalanced datasets in which the number of negatives outweighs the number of positives significantly. While ROC plots are visually appealing and provide an overview of a classifier’s performance across a wide range of specificities, one can ask whether ROC plots could be misleading when applied in imbalanced classification scenarios. We show here that the visual interpretability of ROC plots in the context of imbalanced datasets can be deceptive with respect to conclusions about the reliability of classification performance, owing to an intuitive but wrong interpretation of specificity. PRC plots, on the other hand, can provide the viewer with an accurate prediction of future classification performance due to the fact that they evaluate the fraction of true positives among positive predictions. Our findings have potential implications for the interpretation of a large number of studies that use ROC plots on imbalanced datasets.
Purpose To investigate whether multivariate pattern recognition analysis of arterial spin labeling (ASL) perfusion maps can be used for classification and single-subject prediction of patients with Alzheimer disease (AD) and mild cognitive impairment (MCI) and subjects with subjective cognitive decline (SCD) after using the W score method to remove confounding effects of sex and age. Materials and Methods Pseudocontinuous 3.0-T ASL images were acquired in 100 patients with probable AD; 60 patients with MCI, of whom 12 remained stable, 12 were converted to a diagnosis of AD, and 36 had no follow-up; 100 subjects with SCD; and 26 healthy control subjects. The AD, MCI, and SCD groups were divided into a sex- and age-matched training set (n = 130) and an independent prediction set (n = 130). Standardized perfusion scores adjusted for age and sex (W scores) were computed per voxel for each participant. Training of a support vector machine classifier was performed with diagnostic status and perfusion maps. Discrimination maps were extracted and used for single-subject classification in the prediction set. Prediction performance was assessed with receiver operating characteristic (ROC) analysis to generate an area under the ROC curve (AUC) and sensitivity and specificity distribution. Results Single-subject diagnosis in the prediction set by using the discrimination maps yielded excellent performance for AD versus SCD (AUC, 0.96; P < .01), good performance for AD versus MCI (AUC, 0.89; P < .01), and poor performance for MCI versus SCD (AUC, 0.63; P = .06). Application of the AD versus SCD discrimination map for prediction of MCI subgroups resulted in good performance for patients with MCI diagnosis converted to AD versus subjects with SCD (AUC, 0.84; P < .01) and fair performance for patients with MCI diagnosis converted to AD versus those with stable MCI (AUC, 0.71; P > .05). Conclusion With automated methods, age- and sex-adjusted ASL perfusion maps can be used to classify and predict diagnosis of AD, conversion of MCI to AD, stable MCI, and SCD with good to excellent accuracy and AUC values. (©) RSNA, 2016.
Activation of the kynurenine pathway (KP) of tryptophan metabolism results from chronic inflammation and is known to exacerbate progression of neurodegenerative disease. To gain insights into the links between inflammation, the KP and multiple sclerosis (MS) pathogenesis, we investigated the KP metabolomics profile of MS patients. Most significantly, we found aberrant levels of two key KP metabolites, kynurenic acid (KA) and quinolinic acid (QA). The balance between these metabolites is important as it determines overall excitotoxic activity at the N-methyl-D-Aspartate (NMDA) receptor. We also identified that serum KP metabolic signatures in patients can discriminate clinical MS subtypes with high sensitivity and specificity. A C5.0 Decision Tree classification model discriminated the clinical subtypes of MS with a sensitivity of 91%. After validation in another independent cohort, sensitivity was maintained at 85%. Collectively, our studies suggest that abnormalities in the KP may be associated with the switch from early-mild stage MS to debilitating progressive forms of MS and that analysis of KP metabolites in MS patient serum may have application as MS disease biomarkers.
ABSTRACT Objective: To compare the diagnostic accuracy between cone-beam computed tomography (CBCT) and periapical radiography for detecting simulated external apical root resorption (EARR) in vitro. Materials and Methods: The study sample consisted of 160 single-rooted premolar teeth for simulating EARR of varying degrees according to four setups: no (intact teeth), mild (cavity of 1.0 mm in diameter and depth on root surface), moderate (0.4 mm, 0.8 mm, 1.2 mm, and 1.6 mm root shortening), and severe (2.4 mm, 2.8 mm, 3.2 mm, and 3.6 mm root shortening). Two groups of radiographic images were obtained via CBCT and periapical radiography. The absence or presence and the severity for all resorption lesions were evaluated blindly by two calibrated observers. Results: With the CBCT method, the rates of correct classification of no, mild, moderate, and severe EARR were 96.3%, 98.8%, 41.3%, and 87.5%, respectively; with the periapical radiography method, the rates were 82.5%, 41.3%, 68.8%, and 92.5%, respectively. Highly significant differences were found between the two imaging methods for detection of mild (P < .001), moderate (P < .001), and all EARR (P < .001). For detection of all EARR, the sensitivity and specificity values were 75.8% and 96.3% for CBCT, compared with 67.5% and 82.5% for periapical radiography. Conclusion: CBCT is a reliable diagnostic tool to detect simulated EARR, whereas periapical radiography underestimates it. However, if a periapical radiograph is already available to the diagnosis of EARR, CBCT should be used with extreme caution to avoid additional radiation exposure.
Discrimination of edible oils and fats by combination of multivariate pattern recognition and FT-IR spectroscopy: A comparative study between different modeling methods
- Spectrochimica acta. Part A, Molecular and biomolecular spectroscopy
- Published over 7 years ago
By using FT-IR spectroscopy, many researchers from different disciplines enrich the experimental complexity of their research for obtaining more precise information. Moreover chemometrics techniques have boosted the use of IR instruments. In the present study we aimed to emphasize on the power of FT-IR spectroscopy for discrimination between different oil samples (especially fat from vegetable oils). Also our data were used to compare the performance of different classification methods. FT-IR transmittance spectra of oil samples (Corn, Colona, Sunflower, Soya, Olive, and Butter) were measured in the wave-number interval of 450-4000cm(-1). Classification analysis was performed utilizing PLS-DA, interval PLS-DA, extended canonical variate analysis (ECVA) and interval ECVA methods. The effect of data preprocessing by extended multiplicative signal correction was investigated. Whilst all employed method could distinguish butter from vegetable oils, iECVA resulted in the best performances for calibration and external test set with 100% sensitivity and specificity.