Area Under The Receiver Operating Curve Values Research Articles

Model for end-stage liver disease-3.0 vs. model for end-stage liver disease-sodium: mortality prediction in Korea.

The model for end-stage liver disease (MELD) serves as an indicator for short-term mortality among patients diagnosed with liver cirrhosis (LC) and is used to prioritize patients for liver transplantation. In 2021, the updated version of MELD, MELD-3.0, was introduced to improve the accuracy of the mortality prediction of MELD. Therefore, this study aimed to compare the efficacy of MELD 3.0 and MELD-Na in predicting mortality among Korean patients with LC. A retrospective review was conducted using the medical records of patients diagnosed with LC who were admitted to Konkuk University Hospital From 2011 to 2021. The study calculated the predictive values of MELD-Na and MELD-3.0 for 3- and 6-months mortality using the area under the receiver operating curve (AUROC) and compared the results using the DeLong test. Of the 3,034 patients enrolled in the study, 339 (11.2%) died within 3 months and 421 (14.4%) died within 6 months. The AUROCs values for predicting 3 months mortality were 0.846 for MELD-Na and 0.851 for MELD-3.0. The corresponding AUROC values for predicting 6 months mortality were 0.843 for MELD-Na and 0.848 for MELD-3.0. MELD-3.0 exhibited better discrimination ability than MELD-Na for both 3 (p = 0.03) and 6 months mortality (p = 0.01). Our study found a significant difference between the performance of MELD-3.0 and MELD-Na in Korean patients with LC.

AUCReshaping: improved sensitivity at high-specificity

The evaluation of deep-learning (DL) systems typically relies on the Area under the Receiver-Operating-Curve (AU-ROC) as a performance metric. However, AU-ROC, in its holistic form, does not sufficiently consider performance within specific ranges of sensitivity and specificity, which are critical for the intended operational context of the system. Consequently, two systems with identical AU-ROC values can exhibit significantly divergent real-world performance. This issue is particularly pronounced in the context of anomaly detection tasks, a commonly employed application of DL systems across various research domains, including medical imaging, industrial automation, manufacturing, cyber security, fraud detection, and drug research, among others. The challenge arises from the heavy class imbalance in training datasets, with the abnormality class often incurring a considerably higher misclassification cost compared to the normal class. Traditional DL systems address this by adjusting the weighting of the cost function or optimizing for specific points along the ROC curve. While these approaches yield reasonable results in many cases, they do not actively seek to maximize performance for the desired operating point. In this study, we introduce a novel technique known as AUCReshaping, designed to reshape the ROC curve exclusively within the specified sensitivity and specificity range, by optimizing sensitivity at a predetermined specificity level. This reshaping is achieved through an adaptive and iterative boosting mechanism that allows the network to focus on pertinent samples during the learning process. We primarily investigated the impact of AUCReshaping in the context of abnormality detection tasks, specifically in Chest X-Ray (CXR) analysis, followed by breast mammogram and credit card fraud detection tasks. The results reveal a substantial improvement, ranging from 2 to 40%, in sensitivity at high-specificity levels for binary classification tasks.

In-hospital fall prediction using machine learning algorithms and the Morse fall scale in patients with acute stroke: a nested case-control study

BackgroundFalls are one of the most common accidents in medical institutions, which can threaten the safety of inpatients and negatively affect their prognosis. Herein, we developed a machine learning (ML) model for fall prediction in patients with acute stroke and compared its accuracy with that of the existing fall risk prediction tool, the Morse Fall Scale (MFS).MethodsThis is a retrospective nested case-control study. The initial sample size was 8462 admitted to a single cerebrovascular specialty hospital with acute stroke. A total of 156 fall events occurred, and each fall case was randomly matched with six control cases. Six ML algorithms were used, namely, regularized logistic regression, support vector machine, naïve Bayes (NB), k-nearest neighbors, random forest, and extreme-gradient boosting (XGB).ResultsWe included 156 in the fall group and 934 in the non-fall group. The mean ages of the fall and non-fall groups were 68.3 (± 12.2) and 65.3 (± 12.9) years old, respectively. The MFS total score was significantly higher in the fall group (54.3 ± 18.3) than in the non-fall group (37.7 ± 14.7). The area under the receiver operating curve (AUROC) of the MFS in predicting falls was 0.76 (0.73–0.79). XGB had the highest AUROC of 0.85 (0.78–0.92), and XGB and NB had the highest F1 score of 0.44.ConclusionsThe AUROC values of all of ML algorithms were similar to those of the MFS in predicting fall risk in patients with acute stroke, allowing for accurate and efficient fall screening.

Neural network application for assessing thyroid-associated orbitopathy activity using orbital computed tomography

This study aimed to propose a neural network (NN)-based method to evaluate thyroid-associated orbitopathy (TAO) patient activity using orbital computed tomography (CT). Orbital CT scans were obtained from 144 active and 288 inactive TAO patients. These CT scans were preprocessed by selecting eleven slices from axial, coronal, and sagittal planes and segmenting the region of interest. We devised an NN employing information extracted from 13 pipelines to assess these slices and clinical patient age and sex data for TAO activity evaluation. The proposed NN’s performance in evaluating active and inactive TAO patients achieved a 0.871 area under the receiver operating curve (AUROC), 0.786 sensitivity, and 0.779 specificity values. In contrast, the comparison models CSPDenseNet and ConvNeXt were significantly inferior to the proposed model, with 0.819 (p = 0.029) and 0.774 (p = 0.04) AUROC values, respectively. Ablation studies based on the Sequential Forward Selection algorithm identified vital information for optimal performance and evidenced that NNs performed best with three to five active pipelines. This study establishes a promising TAO activity diagnosing tool with further validation.

Automation Radiomics in Predicting Radiation Pneumonitis (RP)

Radiomics has shown great promise in predicting various diseases. Researchers have previously attempted to include radiomics in their automated detection, diagnosis, and segmentation algorithms, taking these steps based on the promising outcomes of radiomics-based studies. As a result of the increased attention given to this topic, numerous institutions have developed their own radiomics software. These packages, on the other hand, have been utilized interchangeably without regard for their fundamental differences. The primary purpose of this study was to explore benefits of predictive model performance for radiation pneumonitis (RP), which is the most frequent side effect of chest radiotherapy, and through this work, we developed a radiomics model based on deep learning that intends to increase RP prediction performance by combining more data points and digging deeper into these data. In order to evaluate the most popular machine learning models, radiographic characteristics were used, and we recorded the most important of them. The high dimensionality of radiomic datasets is a major issue. The method proposed for use in data problems is the synthetic minority oversampling technique, which we used in order to create a balanced dataset by leveraging suitable hardware and open-source software. The present study assessed the efficacy of various machine learning models, including logistic regression (LR), support vector machine (SVM), random forest (RF), and deep neural network (DNN), in predicting radiation pneumonitis by utilizing specific radiomics features. The findings of the study indicate that the four models displayed satisfactory efficacy in forecasting radiation pneumonitis. The DNN model demonstrated the highest area under the receiver operating curve (AUC-ROC) value, which was 0.87, suggesting its superior predictive capacity among the models considered. The AUC-ROC values for the random forest, SVM, and logistic regression models were 0.85, 0.83, and 0.81, respectively.

Prediction of medication-related osteonecrosis of the jaws using machine learning methods from estrogen receptor 1 polymorphisms and clinical information.

The purpose of this study was to evaluate the effect of estrogen receptor 1 (ESR1) polymorphisms on the development of medication-related osteonecrosis of the jaws (MRONJ) in women with osteoporosis. A total of 125 patients taking bisphosphonates was evaluated the relationship between MRONJ occurrence and single nucleotide polymorphisms (SNPs) of ESR1. Clinical information was collected, including current age, treatment duration, and comorbidity. Univariate and Multivariable regression analyzes were performed to evaluate the independent predictive factors for MRONJ occurrence. Predictive models were constructed using machine learning methods such as Lasso regression, Random forest (RF), and Support vector machine (SVM). The area under the receiver-operating curve (AUROC) was used to evaluate the performance of a binary classifier. Two SNPs of ESR1 (rs4870056 and rs78177662) were significantly associated with MRONJ development. Patients with variant allele (A) of rs4870056 showed 2.45 times (95% CI, 1.03-5.87) the odds of MRONJ occurrence compared to those with wild-type homozygote (GG) after adjusting covariates. Additionally, carriers with variant allele (T) of rs78177662 had higher odds than those with wild-type homozygote (CC) (adjusted odds ratio (aOR), 2.64, 95% CI, 1.00-6.94). Among demographic variables, age ≥ 72 years (aOR, 3.98, 95% CI, 1.60-9.87) and bisphosphonate exposure ≥48 months (aOR, 3.16, 95% CI, 1.26-7.93) were also significant risk factors for MRONJ occurrence. AUROC values of machine learning methods ranged between 0.756-0.806 in the study. Our study showed that the MRONJ occurrence was associated with ESR1 polymorphisms in osteoporotic women.

Predicting Bacteremia among Septic Patients Based on ED Information by Machine Learning Methods: A Comparative Study.

Introduction: Bacteremia is a common but life-threatening infectious disease. However, a well-defined rule to assess patient risk of bacteremia and the urgency of blood culture is lacking. The aim of this study is to establish a predictive model for bacteremia in septic patients using available big data in the emergency department (ED) through logistic regression and other machine learning (ML) methods. Material and Methods: We conducted a retrospective cohort study at the ED of National Cheng Kung University Hospital in Taiwan from January 2015 to December 2019. ED adults (≥18 years old) with systemic inflammatory response syndrome and receiving blood cultures during the ED stay were included. Models I and II were established based on logistic regression, both of which were derived from support vector machine (SVM) and random forest (RF). Net reclassification index was used to determine which model was superior. Results: During the study period, 437,969 patients visited the study ED, and 40,395 patients were enrolled. Patients diagnosed with bacteremia accounted for 7.7% of the cohort. The area under the receiver operating curve (AUROC) in models I and II was 0.729 (95% CI, 0.718–0.740) and 0.731 (95% CI, 0.721–0.742), with Akaike information criterion (AIC) of 16,840 and 16,803, respectively. The performance of model II was superior to that of model I. The AUROC values of models III and IV in the validation dataset were 0.730 (95% CI, 0.713–0.747) and 0.705 (0.688–0.722), respectively. There is no statistical evidence to support that the performance of the model created with logistic regression is superior to those created by SVM and RF. Discussion: The advantage of the SVM or RF model is that the prediction model is more elastic and not limited to a linear relationship. The advantage of the LR model is that it is easy to explain the influence of the independent variable on the response variable. These models could help medical staff identify high-risk patients and prevent unnecessary antibiotic use. The performance of SVM and RF was not inferior to that of logistic regression. Conclusions: We established models that provide discrimination in predicting bacteremia among patients with sepsis. The reported results could inspire researchers to adopt ML in their development of prediction algorithms.

Quantitative assessment for characterization of breast lesion tissues using adaptively decomposed ultrasound RF images

Quantitative ultrasound is of significant relevance to the clinical diagnosis of benign and malignant breast lesions. Owing to desmoplastic reactions in the tumor and malignant cells invading the surrounding tissues, the irregular boundary of the lesion is a significant indication of malignancy. Complexity measured using entropy based on ultrasound radio frequency (RF) data is investigated to characterize the disease; however, the performance is reduced due to the slowly varying components in the data. In the present study, adoption of small-window entropy (SWE) based on adaptively decomposed ultrasound RF images is proposed for the characterization of breast lesion tissues. First, down-sampled, dilated and cropped RF (DDC_RF) images are obtained from the original ultrasonic RF data. Subsequently, a fast multivariate empirical mode decomposition algorithm is employed to adaptively decompose the DDC_RF images into intrinsic mode functions (IMFs). The ring regions of interest that are the areas surrounding the tumors are determined from all individual IMF images and their different combinations to calculate the SWEs. The assessment is performed using the Open Access Series of Breast Ultrasonic Data (OASBUD), which includes original RF data and histological assessment results of 48 benign and 48 malignant lesions. The receiver operating curve (ROC) and the area under the ROC (AUC) values are calculated to assess the ability to diagnose breast lesions. The results show that the SWEs based on the combinations of IMFs 1–3, IMFs 1,2, IMFs 1–4, and IMFs 1,3 provided the AUC values of 0.88, 0.87, 0.86, and 0.85, respectively, which are greater than 0.84, obtained from the undecomposed RF images. It is concluded that the SWE based on the adaptively decomposed ultrasound RF images is an improved characteristic for distinguishing between benign and malignant breast lesions, and it is suggested that it may serve as a basis for computer-assisted breast cancer ultrasound diagnosis.

Machine and Deep Learning Prediction Of Prostate Cancer Aggressiveness Using Multiparametric MRI.

Prostate cancer (PCa) is the most frequent male malignancy and the assessment of PCa aggressiveness, for which a biopsy is required, is fundamental for patient management. Currently, multiparametric (mp) MRI is strongly recommended before biopsy. Quantitative assessment of mpMRI might provide the radiologist with an objective and noninvasive tool for supporting the decision-making in clinical practice and decreasing intra- and inter-reader variability. In this view, high dimensional radiomics features and Machine Learning (ML) techniques, along with Deep Learning (DL) methods working on raw images directly, could assist the radiologist in the clinical workflow. The aim of this study was to develop and validate ML/DL frameworks on mpMRI data to characterize PCas according to their aggressiveness. We optimized several ML/DL frameworks on T2w, ADC and T2w+ADC data, using a patient-based nested validation scheme. The dataset was composed of 112 patients (132 peripheral lesions with Prostate Imaging Reporting and Data System (PI-RADS) score ≥ 3) acquired following both PI-RADS 2.0 and 2.1 guidelines. Firstly, ML/DL frameworks trained and validated on PI-RADS 2.0 data were tested on both PI-RADS 2.0 and 2.1 data. Then, we trained, validated and tested ML/DL frameworks on a multi PI-RADS dataset. We reported the performances in terms of Area Under the Receiver Operating curve (AUROC), specificity and sensitivity. The ML/DL frameworks trained on T2w data achieved the overall best performance. Notably, ML and DL frameworks trained and validated on PI-RADS 2.0 data obtained median AUROC values equal to 0.750 and 0.875, respectively, on unseen PI-RADS 2.0 test set. Similarly, ML/DL frameworks trained and validated on multi PI-RADS T2w data showed median AUROC values equal to 0.795 and 0.750, respectively, on unseen multi PI-RADS test set. Conversely, all the ML/DL frameworks trained and validated on PI-RADS 2.0 data, achieved AUROC values no better than the chance level when tested on PI-RADS 2.1 data. Both ML/DL techniques applied on mpMRI seem to be a valid aid in predicting PCa aggressiveness. In particular, ML/DL frameworks fed with T2w images data (objective, fast and non-invasive) show good performances and might support decision-making in patient diagnostic and therapeutic management, reducing intra- and inter-reader variability.

Development and Validation of MRI-Based Radiomics Models for Diagnosing Juvenile Myoclonic Epilepsy.

Radiomic modeling using multiple regions of interest in MRI of the brain to diagnose juvenile myoclonic epilepsy (JME) has not yet been investigated. This study aimed to develop and validate radiomics prediction models to distinguish patients with JME from healthy controls (HCs), and to evaluate the feasibility of a radiomics approach using MRI for diagnosing JME. A total of 97 JME patients (25.6 ± 8.5 years; female, 45.5%) and 32 HCs (28.9 ± 11.4 years; female, 50.0%) were randomly split (7:3 ratio) into a training (n = 90) and a test set (n = 39) group. Radiomic features were extracted from 22 regions of interest in the brain using the T1-weighted MRI based on clinical evidence. Predictive models were trained using seven modeling methods, including a light gradient boosting machine, support vector classifier, random forest, logistic regression, extreme gradient boosting, gradient boosting machine, and decision tree, with radiomics features in the training set. The performance of the models was validated and compared to the test set. The model with the highest area under the receiver operating curve (AUROC) was chosen, and important features in the model were identified. The seven tested radiomics models, including light gradient boosting machine, support vector classifier, random forest, logistic regression, extreme gradient boosting, gradient boosting machine, and decision tree, showed AUROC values of 0.817, 0.807, 0.783, 0.779, 0.767, 0.762, and 0.672, respectively. The light gradient boosting machine with the highest AUROC, albeit without statistically significant differences from the other models in pairwise comparisons, had accuracy, precision, recall, and F1 scores of 0.795, 0.818, 0.931, and 0.871, respectively. Radiomic features, including the putamen and ventral diencephalon, were ranked as the most important for suggesting JME. Radiomic models using MRI were able to differentiate JME from HCs.

Accuracy of the indices of MS-39 anterior segment optical coherence tomography in the diagnosis of keratoconic corneas.

To determine the normative and the cut-off values of various indices available in the MS 39 Anterior Segment Optical Coherence Tomography (MS-39 AS-OCT) for keratoconus (KC) diagnosis, and to detect the accuracy of the variable available parameters. This cross sectional observational study was conducted at Dr Rifay Ophthalmology Center, Rabat, Morocco, on 172 eyes with KC (group 1) and 248 eyes of healthy controls (group 2). Participants were screened using MS-39 AS-OCT (CSO, Firenze, Italy). The investigated indices included: keratometric indices, pachymetric indices, elevation indices, Keratoconus Summary Indices (KSI), aberration indices, and epithelial mapping evaluation. Most of the studied indices were significantly different between both groups (P value <0.001). Twenty two indices had Area Under the Receiver Operating Curve (AUROC) values >0.950, eight of which were from the KSI. The indices with the highest AUROC values were: Root Mean Square (RMS) for the front corneal surface at 6 mm circle (AUROC = 0.996, and best cut-off >1.69 with sensitivity of 97.67% and specificity of 97.98%) and RMS per unit of area for the front surface at 6 mm circle (AUROC = 0.996, and best cut-off >0.06 with sensitivity of 97.67% and specificity of 97.98%). Epithelial thickness-derived metrics were not among the parameters with the highest accuracy. This study presented the mean and the cut-off values for a plethora of parameters available in the MS-39 AS-OCT. The results of this study show that MS-39 AS-OCT is a valuable equipment for diagnosing keratoconic corneas, with a high accuracy detected for many parameters.

Machine Learning Approaches for Predicting Bisphosphonate-Related Osteonecrosis in Women with Osteoporosis Using VEGFA Gene Polymorphisms.

Objective: This nested case–control study aimed to investigate the effects of VEGFA polymorphisms on the development of bisphosphonate-related osteonecrosis of the jaw (BRONJ) in women with osteoporosis. Methods: Eleven single nucleotide polymorphisms (SNPs) of the VEGFA were assessed in a total of 125 patients. Logistic regression was performed for multivariable analysis. Machine learning algorithms, namely, fivefold cross-validated multivariate logistic regression, elastic net, random forest, and support vector machine, were developed to predict risk factors for BRONJ occurrence. Area under the receiver-operating curve (AUROC) analysis was conducted to assess clinical performance. Results: The VEGFA rs881858 was significantly associated with BRONJ development. The odds of BRONJ development were 6.45 times (95% CI, 1.69–24.65) higher among carriers of the wild-type rs881858 allele compared with variant homozygote carriers after adjusting for covariates. Additionally, variant homozygote (GG) carriers of rs10434 had higher odds than those with wild-type allele (OR, 3.16). Age ≥ 65 years (OR, 16.05) and bisphosphonate exposure ≥ 36 months (OR, 3.67) were also significant risk factors for BRONJ occurrence. AUROC values were higher than 0.78 for all machine learning methods employed in this study. Conclusion: Our study showed that the BRONJ occurrence was associated with VEGFA polymorphisms in osteoporotic women.

Accuracy of Two-Dimensional Shear Wave Elastography and Attenuation Imaging for Evaluation of Patients With Nonalcoholic Steatohepatitis

We evaluated the accuracy of a multiparametric approach using attenuation imaging and 2-dimensional shear wave elastography (2D-SWE) for the detection of steatosis and fibrosis in patients with nonalcoholic fatty liver disease (NAFLD). We studied 102 patients with increased levels of liver enzymes or suspicion of NAFLD, examined by attenuation imaging and 2D-SWE, immediately before biopsy collection and analysis (reference standard), from January 2018 to July 2019. We collected data on the attenuation coefficient (dB/cm/MHz) from attenuation imaging, liver stiffness measurements, and shear wave dispersion slope (SWDS, [m/s]/kHz) from 2D-SWE. Multivariate linear regression analysis was performed to identify factors associated with each parameter. Diagnostic performance was determined from area under the receiver operating curve (AUROC) values. The attenuation coefficient was associated with steatosis grade (P < .01) and identified patients with steatosis grades S1 or higher, S2 or higher, and S3 or higher, with AUROC values of 0.93, 0.88, and 0.83, respectively. Liver stiffness associated with fibrosis stage (P < .01) and lobular inflammatory activity was the only factor associated with SWDS (P < .01). SWDS detected inflammation grades I1 or higher, I2 or higher, and I3 or higher with AUROC values of 0.89, 0.85, and 0.78, respectively. We developed a risk scoring system to detect steatohepatitis based on the attenuation coefficient (score of 1 for 0.64 < attenuation coefficient ≤ 0.70; score of 2 for 0.70 < attenuation coefficient ≤ 0.73; and score of 3 for attenuation coefficient >0.73) and SWDS (score of 2 for 10.5 [m/s]/kHz < SWDS ≤ 11.7 [m/s]/kHz; and score of 3 for SWDS >11.7 [m/s]/kHz), using an unweighted sum of each score. Based on histopathology analysis, 55 patients had steatohepatitis. Risk scores correlated with NAFLD activity score (rho= 0.73; P < .01). Our scoring system identified patients with steatohepatitis with an AUROC of 0.93-this valuewassignificantly higher than that of other parameters (P < .05), except SWDS (AUROC, 0.89; P= .18). In the evaluation of patients with suspected NAFLD, the attenuation coefficient can identify patients with steatosis and liver stiffness can detect fibrosis accurately. SWDS was associated significantly with lobular inflammation. We developed a risk scoring system based on the attenuation coefficient and SWDS that might be used to detect steatohepatitis.

Noninvasive Models for Predicting Liver Fibrosis in Individuals with Hepatitis D Virus/Hepatitis B Virus Coinfection in the Brazilian Amazon Region.

Hepatitis D virus (HDV) genotype III is endemic in the western Amazon basin and is considered to cause the most severe form of chronic viral hepatitis. Recently, noninvasive fibrosis scores to determine the stage of liver fibrosis have been evaluated in individuals positive for HDV genotype I, but their utility in HDV genotype III-positive patients is unknown. In this retrospective study conducted in an outpatient viral hepatitis referral clinic in the Brazilian Amazon region, the aspartate aminotransferase (AST) to Aspartate aminotransferase to Platelet Ratio Index (APRI) and Fibrosis Index for Liver Fibrosis (FIB-4) values were calculated and compared with histological fibrosis stages. Among the 50 patients analyzed, the median age at liver biopsy was 35.6 years, 66% were male, and all had compensated liver disease. Histological staging revealed fibrosis stages 0, 1, 2, 3, and 4 in four (8%), eight (16%), 11 (22), 11 (22%), and 16 (32%) patients, respectively. The area under the receiver operating curve (AUROC) of AST-to-alanine aminotransferase (ALT) ratio, APRI, and FIB-4 for detection of significant fibrosis (F ≥ 2) was 0.550 (P = 0.601), 0.853 (P < 0.001), and 0.853 (P < 0.0001), respectively. Lower AUROC values were obtained for cirrhosis: the AST-to-ALT ratio was 0.640 (P = 0.114), APRI was 0.671 (P = 0.053), and FIB-4 was 0.701 (P = 0.023). The optimal cutoff value for significant fibrosis for APRI was 0.708 (sensitivity 84% and specificity 92%) and for FIB-4 was 1.36 (sensitivity 76% and specificity 92%). Aspartate aminotransferase to Platelet Ratio Index and FIB-4 were less useful to predict cirrhosis. In contrast to recent reports from Europe and North America, both APRI and FIB-4 may identify significant fibrosis in HDV-III-infected patients from northwestern Brazil.

Current applications of machine learning in the screening and diagnosis of glaucoma: a systematic review and Meta-analysis.

To compare the effectiveness of two well described machine learning modalities, ocular coherence tomography (OCT) and fundal photography, in terms of diagnostic accuracy in the screening and diagnosis of glaucoma. A systematic search of Embase and PubMed databases was undertaken up to 1st of February 2019. Articles were identified alongside their reference lists and relevant studies were aggregated. A Meta-analysis of diagnostic accuracy in terms of area under the receiver operating curve (AUROC) was performed. For the studies which did not report an AUROC, reported sensitivity and specificity values were combined to create a summary ROC curve which was included in the Meta-analysis. A total of 23 studies were deemed suitable for inclusion in the Meta-analysis. This included 10 papers from the OCT cohort and 13 from the fundal photos cohort. Random effects Meta-analysis gave a pooled AUROC of 0.957 (95%CI=0.917 to 0.997) for fundal photos and 0.923 (95%CI=0.889 to 0.957) for the OCT cohort. The slightly higher accuracy of fundal photos methods is likely attributable to the much larger database of images used to train the models (59 788 vs 1743). No demonstrable difference is shown between the diagnostic accuracy of the two modalities. The ease of access and lower cost associated with fundal photo acquisition make that the more appealing option in terms of screening on a global scale, however further studies need to be undertaken, owing largely to the poor study quality associated with the fundal photography cohort.

Abstract 2958: Development of a superenhancer element-associated transcription factor signature in primary tumors for the identification of liver metastasis in colorectal cancer patients

Abstract Background: More than 50% of patients with colorectal cancer (CRC) develop metastatic disease to liver, which ultimately results in death in more than two thirds of the cases. Hence, earlier detection, together with the identification of patients at high-risk for tumor recurrence could significantly improve patient outcomes. Current paradigm asserts that metastasis must at least in part be a reflection of mutational and transcriptional landscape present in the majority of cells which constitute the primary tumor mass. Among these genomic determinants, transcription factors (TFs) are of particular interest, as they regulate the expression of both protein coding and noncoding RNAs. Keeping this in mind, we undertook an effort to develop a TF-based gene expression signature in primary CRC tissues that can help identify patients with a metastatic disease. Method: A comprehensive list of TFs was established from key relevant databases including TFdb, FANTOM and TFCAT. An in-silico discovery for differentially expressed TFs was performed in multiple, publicly available datasets. The final TF signature was examined in two independent cohorts of CRC patients (N=104 and 151 respectively). Statistical analyses included area under the receiver operating curve (AUROC) estimates, univariate and multivariate logistic regression, and Kaplan Meier survival outcomes. Lastly, we analyzed associations between our selected TF and their binding to various superenhancer elements by analyzing a publicly available H3K27ac dataset on 21 colorectal cell lines. Result: We developed a panel of 14 differentially expressed TFs by analyzing primary CRCs with and without liver metastasis. The AUROC values for liver metastasis-positive cases were 0.79 and 0.86 in the training and validation cohorts, respectively. The AUROC values improved to 0.93 in both cohorts on combining the signature with lymph node metastasis (LNM) status and the CEA levels. This signature was also associated with worse overall survival in the training (H.R (95%C.I) = 4.8(2.1-11.2), p&amp;lt;0.001) and validation cohorts (H.R (95%C.I) 8.9(3.1-25.6), p&amp;lt;0.001). In-silico analyses for superenhancer binding motifs for these TFs indicated that there exists a co-occurrence between many of the motif pairs, and 7 motifs (EHF, KLF7, MECP2, PURA, RARB, TCF4 and ZNF354C) were significantly more enriched within the superenhancers. Conclusions: We present a TF gene expression signature that can identify liver metastasis and can predict poor survival by analyzing primary tumor tissues from CRC patients. Transcriptional cooperativity between 7 of these 14 TFs along with their enriched binding to key superenhancer elements underscores their functional role in metastasis. Citation Format: Roshni Roy, Jesper Grud Madsen, Yasuhide Yamada, Yoshinaga Okugawa, Yuji Toiyama, Susanne Mandrup, Ajay Goel. Development of a superenhancer element-associated transcription factor signature in primary tumors for the identification of liver metastasis in colorectal cancer patients [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 2958.

Increased Survival for Patients With Cirrhosis and Organ Failure in Liver Intensive Care and Validation of the Chronic Liver Failure–Sequential Organ Failure Scoring System

During the past decade, survival has increased among patients admitted to general intensive care units, but it is not clear if it has increased for patients admitted with cirrhosis and organ failure. The chronic liver failure-sequential organ failure assessment (CLIF-SOFA) recently was developed as an adaptation to the SOFA to predict outcomes of patients, but requires validation. We investigated changes in outcomes of patients with cirrhosis and organ failure since 2000, compared the abilities of SOFA and CLIF-SOFA to predict patient survival, and validated the CLIF-SOFA system. In a retrospective study, we collected data from 971 patients (median age, 52 y; age range, 16-90 y; 62% male) with cirrhosis (54% alcohol associated, 12% viral, and 34% other causes). The patients were admitted under emergency conditions from January 1, 2000, to December 31, 2010, to a liver intensive therapy unit in the United Kingdom. Patient survival while in the hospital was compared with measures of illness severity, Acute Physiology and Chronic Health Evaluation (APACHE) II scores, model for end-stage liver disease (MELD) scores, SOFA scores, and CLIF-SOFA scores. Patients had a median APACHE II score of 21 (range, 5-50) and a median MELD score of 23 (range, 6-40). The median APACHE II score at admission decreased from 23 to 22 over the study period (P < .001), whereas the median MELD score at admission decreased from 23 to 18 (P < .001). Overall survival until hospital discharge was 51%; this value increased from 40% in 2000 to 63% in 2010 (P < .001). The unadjusted odds ratio for change in mortality/year was 0.87 (95% confidence interval, 0.83-0.91; P < .001). The APACHE II score adjusted odds ratio for mortality was 0.89 (95% confidence interval, 0.84-0.93; P < .001). The etiology of cirrhosis was not associated with a significant difference in survival. CLIF-SOFA and SOFA scores at the time of admission predicted patient survival with area under the receiver operating curve (AUROC) values of 0.813 and 0.799, respectively; the scores at 48 hours after admission predicted survival with AUROC values of 0.853 and 0.840, and scores after 1 week predicted survival with AUROC values of 0.842 and 0.844, respectively. These AUROC values were higher than those obtained from APACHE II or MELD scores. The proportion of patients with cirrhosis who survived after admission to intensive care increased from 2000 to 2010. SOFA and CLIF-SOFA scores during the first week of critical care appear to have similar abilities to predict patient survival.

Software defect prediction using ensemble learning on selected features

ContextSeveral issues hinder software defect data including redundancy, correlation, feature irrelevance and missing samples. It is also hard to ensure balanced distribution between data pertaining to defective and non-defective software. In most experimental cases, data related to the latter software class is dominantly present in the dataset. ObjectiveThe objectives of this paper are to demonstrate the positive effects of combining feature selection and ensemble learning on the performance of defect classification. Along with efficient feature selection, a new two-variant (with and without feature selection) ensemble learning algorithm is proposed to provide robustness to both data imbalance and feature redundancy. MethodWe carefully combine selected ensemble learning models with efficient feature selection to address these issues and mitigate their effects on the defect classification performance. ResultsForward selection showed that only few features contribute to high area under the receiver-operating curve (AUC). On the tested datasets, greedy forward selection (GFS) method outperformed other feature selection techniques such as Pearson’s correlation. This suggests that features are highly unstable. However, ensemble learners like random forests and the proposed algorithm, average probability ensemble (APE), are not as affected by poor features as in the case of weighted support vector machines (W-SVMs). Moreover, the APE model combined with greedy forward selection (enhanced APE) achieved AUC values of approximately 1.0 for the NASA datasets: PC2, PC4, and MC1. ConclusionThis paper shows that features of a software dataset must be carefully selected for accurate classification of defective components. Furthermore, tackling the software data issues, mentioned above, with the proposed combined learning model resulted in remarkable classification performance paving the way for successful quality control.

Comparison of three-dimensional optical coherence tomography and combining a rotating Scheimpflug camera with a Placido topography system for forme fruste keratoconus diagnosis

ObjectiveTo evaluate the ability of parameters measured by three-dimensional (3D) corneal and anterior segment optical coherence tomography (CAS-OCT) and a rotating Scheimpflug camera combined with a Placido topography system (Scheimpflug...

Noninvasive Assessment of Nonalcoholic Fatty Liver Disease in Obese or Overweight Patients

Reliable noninvasive tools are needed for staging nonalcoholic fatty liver disease (NAFLD). Published scoring systems have not been validated in prospective assessments of unselected patients. We aimed to identify factors that predicted development of nonalcoholic steatohepatitis (NASH) in a large group of overweight or obese patients and compared these with established factors. We performed a prospective analysis of factors associated with the development and severity of NAFLD in patients at a single obesity center. We evaluated liver involvement in 542 patients by a large set of routine and non-routine parameters, including ultrasound and genetic testing. Those suspected of having NAFLD underwent liver biopsy (57.7%). Patients were divided into design cohort (n = 200) and validation cohort (n = 113) to identify factors associated with the presence and severity of NAFLD and NASH. Factors independently associated with development of NASH included increased levels of alanine aminotransferase (ALT), fasting levels of C-peptide, and ultrasound steatosis scores (USSs), with area under the receiver operating curve (AUROC) values of 0.854 in the design cohort and 0.823 in the validation cohort. NASH activity scores also correlated with level of ALT, USS, and fasting level of C-peptide (R(2) = 0.491). Independent predictors of advanced fibrosis included waist circumference and level of aspartate aminotransferase (AUROC values of 0.839 and 0.846 for design and validation cohorts, respectively; negative predictive values of 98% and 97%, respectively, for a cutoff of -2.14). Previously published scoring systems had significantly lower AUROC values. Levels of CK18 and PNPLA3 polymorphisms correlated with development of NASH but did not add value. Parameters routinely analyzed in assessing obese patients can be used to determine the presence of NASH and advanced fibrosis. Non-routine tests do not increase diagnostic accuracy. Previously published scores are significantly less accurate.