Mean AUC Research Articles

Predicting Intensive Care Delirium with Machine Learning: Model Development and External Validation.

Delirium poses significant risks to patients, but countermeasures can be taken to mitigate negative outcomes. Accurately forecasting delirium in intensive care unit (ICU) patients could guide proactive intervention. Our primary objective was to predict ICU delirium by applying machine learning to clinical and physiologic data routinely collected in electronic health records. Two prediction models were trained and tested using a multicenter database (years of data collection 2014 to 2015), and externally validated on two single-center databases (2001 to 2012 and 2008 to 2019). The primary outcome variable was delirium defined as a positive Confusion Assessment Method for the ICU screen, or an Intensive Care Delirium Screening Checklist of 4 or greater. The first model, named "24-hour model," used data from the 24 h after ICU admission to predict delirium any time afterward. The second model designated "dynamic model," predicted the onset of delirium up to 12 h in advance. Model performance was compared with a widely cited reference model. For the 24-h model, delirium was identified in 2,536 of 18,305 (13.9%), 768 of 5,299 (14.5%), and 5,955 of 36,194 (11.9%) of patient stays, respectively, in the development sample and two validation samples. For the 12-h lead time dynamic model, delirium was identified in 3,791 of 22,234 (17.0%), 994 of 6,166 (16.1%), and 5,955 of 28,440 (20.9%) patient stays, respectively. Mean area under the receiver operating characteristics curve (AUC) (95% CI) for the first 24-h model was 0.785 (0.769 to 0.801), significantly higher than the modified reference model with AUC of 0.730 (0.704 to 0.757). The dynamic model had a mean AUC of 0.845 (0.831 to 0.859) when predicting delirium 12 h in advance. Calibration was similar in both models (mean Brier Score [95% CI] 0.102 [0.097 to 0.108] and 0.111 [0.106 to 0.116]). Model discrimination and calibration were maintained when tested on the validation datasets. Machine learning models trained with routinely collected electronic health record data accurately predict ICU delirium, supporting dynamic time-sensitive forecasting.

52 A DEEP-LEARNING APPROACH FOR SCAR DETECTION IN EARLY CONTRAST-ENHANCED CARDIAC COMPUTED TOMOGRAPHY IMAGES

Abstract Diagnosis of myocardial fibrosis has a relevant prognostic and therapeutic role. Scar-tissue analysis is commonly performed with late gadolinium contrast-enhanced (CE) cardiac magnetic resonance (CMR). However, CMR might be contraindicated or unavailable. Coronary computed tomography (CCT), which is the technique of choice for many routine assessments, is emerging as an alternative to CMR. The objective of this study is to evaluate whether artificial intelligence (AI) could allow identification of myocardial fibrosis from routine early CE-CCT images. Fifty consecutive patients with left ventricular dysfunction (LVD), who underwent both CE-CMR and (early and late) CE-CCT were retrospectively selected. According to the late enhancement CMR patterns, patients were classified as with ischemic (n=15, 30%) or non-ischemic (n=25, 70%) LVD. Scar regions were manually traced on late CE-CCT using CE-CMR as ground-truth. On early CE-CCT images, the myocardial sectors were extracted according to AHA 16-segment model and labeled as with scar or not based on the late CE-CCT manual tracing. A deep-learning model was developed to classify each sector (Picture). Of the initial 44187 sectors computed out of the 8285 slices available from the early CE-CCT images, 4594 sectors (10%) presented scar. The CNN approach on the early CE-CCT images yielded a classification accuracy for all sectors of 71% (95% confidence interval (CI): 63%-79%) obtained through 5-fold cross validation. The mean sensitivity, positive predictive value (PPV) and negative predictive value (NPV) for the testing fold resulted in 73% (95% CI: 66%-79%), 56% (95% CI: 48%-65%) and 85% (95% CI: 82%- 88%), respectively. The mean AUC across the five folds was 76% (95% CI: 72%-81%). In a per-segment analysis of the 16-segment AHA model the bull's eye segmental comparison of CE-CMR and respective early CE-CCT findings an 91% agreement was achieved. Artificial intelligence can detect both ischemic and non-ischemic myocardial fibrosis from routine noninvasive coronary scans, without additional contrast-agent administration or radiational dose, thus assisting diagnosis and management of patients with LV dysfunction and coronary artery disease.

Bioequivalence study of BETALOC-XR™ (Drug International Ltd, Bangladesh) and METOPROLOL 100 STADA™ (STADA Pharm Gmbh, Germany) as open-label, two-way crossover and randomized fashion among healthy male volunteers

Bioequivalence study of two oral formulations of metoprolol 100 mg tablet was brought about in 16 healthy male Bangladeshi normal individuals as crossover-randomized fashion. The test and reference formulated products were BETALOC-XR™ (Drug International Ltd, Bangladesh) and METOPROLOL 100 STADA™ (STADA Pharm Gmbh, Germany). Each tablet was delivered with 150 mL of water to every normal individual after whole night fasting on two treatment days separated by more than one week washout period. After administering drug, blood samples were collected serially for a period of 24 hours. A validated HPLC method was used to estimate the plasma concentrations of metoprolol. The pharmacokinetic parameters Cmax, Tmax, AUC0→24h, t1/2, and Kel were determined in this study. The mean (± SD) AUC0→24h for test drug BETALOC-XRTM was 490.3 ± 187.5 ng/hr/mL whereas for reference drug METOPROLOL 100 STADATM it was 559.9 ± 267.2 ng/hr/mL. The relative bioavailability (BETALOC-XRTM/METOPROLOL 100 STADATM ratio) was 87.5%. The Cmax, tmax, half-life of elimination (t1/2) and the rate of elimination (Kel) for test drug was 170.9 ± 78.3 ng/mL, 2.3 ± 0.8 hours, 1.2 ± 2.0 hour and 0.4362 whereas for reference drug was 209.1 ± 91.3 ng/mL, 1.7 ± 0.7 hours, 1.3 ± 2.2 hour and 0.5875 respectively. Based on these statistical hypothesis it was conjecture that a BETALOC-XRTM tablet is bioequivalent to a METOPROLOL 100 STADATM tablet.

Combating medical noisy labels by disentangled distribution learning and consistency regularization

Machine learning, particularly the deep convolutional neural network (CNN), has significantly benefited computer-aided diagnosis technology to improve disease identification performance. Nevertheless, the shortage of training data with high-quality labels hinders training an effective and robust model. Medical image annotation is time-consuming and usually requires domain experts’ knowledge. Thus, most medical image datasets contain label noise including uncertainty and inconsistency. This paper proposes a hybrid hard-soft label learning mechanism and consistency regularization from two perspectives to enhance a single model’s disease detection capability instead of struggling with model ensembling. Specifically, a disentangled distribution learning integrates multiple reference models’ predictions and disentangles them into a majority confident label vector and a description degree score vector for co-training the single-target model, mitigating the negative influence caused by label noise. Furthermore, inter- and intra-instance consistency regularization aim to improve the target model’s robustness to give consistent predictions on images with similar lesion appearances. We conducted substantial experiments on the public chest X-ray and fundus image datasets as well as a computer vision benchmark, Clothing 1M, showing that our model surpasses the state-of-the-art methods with consistent improvements. The mean AUC on the CheXpert dataset is increased by 2.44%. The Q.W. Kappa on the KaggleDR+ and FGADR datasets, is increased by 5.65% and 4.66%, respectively. The classification accuracy on the Clothing 1M dataset is increased by 4.47%.

An Integrated Clinical-MR Radiomics Model to Estimate Survival Time in Patients With Endometrial Cancer.

Determination of survival time in women with endometrial cancer using clinical features remains imprecise. Features from MRI may improve the survival estimation allowing improved treatment planning. To identify clinical features and imaging signatures on T2-weighted MRI that can be used in an integrated model to estimate survival time for endometrial cancer subjects. Retrospective. Four hundred thirteen patients with endometrial cancer as training (N=330, 66.41 ± 11.42 years) and validation (N=83, 67.60 ± 11.89 years) data and an independent set of 82 subjects as testing data (63.26 ± 12.38 years). 1.5-T and 3-T scanners with sagittal T2-weighted spin echo sequence. Tumor regions were manually segmented on T2-weighted images. Features were extracted from segmented masks, and clinical variables including age, cancer histologic grade and risk score were included in a Cox proportional hazards (CPH) model. A group least absolute shrinkage and selection operator method was implemented to determine the model from the training and validation datasets. A likelihood-ratio test and decision curve analysis were applied to compare the models. Concordance index (CI) and area under the receiver operating characteristic curves (AUCs) were calculated to assess the model. Three radiomic features (two image intensity and volume features) and two clinical variables (age and cancer grade) were selected as predictors in the integrated model. The CI was 0.797 for the clinical model (includes clinical variables only) and 0.818 for the integrated model using training and validation datasets, the associated mean AUC value was 0.805 and 0.853. Using the testing dataset, the CI was 0.792 and 0.882, significantly different and the mean AUC was 0.624 and 0.727 for the clinical model and integrated model, respectively. The proposed CPH model with radiomic signatures may serve as a tool to improve estimated survival time in women with endometrial cancer. 4 TECHNICAL EFFICACY: Stage 2.

Hypernetwork Representation Learning Based on Hyperedge Modeling

Most network representation learning approaches only consider the pairwise relationships between the nodes in ordinary networks but do not consider the tuple relationships, namely the hyperedges, among the nodes in the hypernetworks. Therefore, to solve the above issue, a hypernetwork representation learning approach based on hyperedge modeling, abbreviated as HRHM, is proposed, which fully considers the hyperedges to obtain ideal node representation vectors that are applied to downstream machine learning tasks such as node classification, link prediction, community detection, and so on. Experimental results on the hypernetwork datasets show that with regard to the node classification task, the mean node classification accuracy of HRHM approach goes beyond other best baseline approach by about 1% on the MovieLens and wordnet, and with regard to the link prediction task, except for HPHG approach, the mean AUC value of HRHM approach surpasses that of other baseline approaches by about 17%, 18%, and 6%, respectively, on the GPS, drug, and wordnet. The mean AUC value of HRHM approach is very close to that of other best baseline approach on the MovieLens.

Ensemble projections of fish distribution in response to climate changes in the Yellow and Bohai Seas, China

Climate change is an important driving force affecting species distribution, so quantifying the influences of climate change on species distributions is necessary for effective fishery management. To identify the geographical distribution pattern and future potential suitable habitat area of fishes in the Yellow and Bohai Seas (YBS), we built ensemble models of spatial distribution for 22 important fish species using 3185 valid distributional records and 9 environmental variables extracted from multiple databases. The constructed ensemble models showed high accuracy with mean AUC, Kappa and TSS values of 0.97, 0.82 and 0.84. Salinity and temperature proximal to the seabed were the main environmental factors affecting the distribution of fishes. Presently, the number of important fish species (NIFS) tends to be low in the Bohai Sea and high in the Yellow Sea. Future projections indicated that there would be obvious interspecific differences in the geographical distribution of fishes, and the number of species with range contractions is predicted to be greater than that of range expansions. Coastal fishes and cold temperate fishes are predicted to narrow their occupied areas. In the future, the NIFS in the YBS is expected to increase overall. Spatially, sporadic areas in the central and southern Yellow Sea will have a reduced NIFS, while the Bohai Sea, coastal waters near the southern Shandong Peninsula and the northern East China Sea may experience increased NIFS. Our results provide a theoretical basis for predicting the climate-driven range shifts of fishes in one of the world’s most heavily impacted marine ecosystems, that can be extended to develop climate-adaptive management strategies.

A Food-Effect Study to Evaluate the Oral Bioavailability of Desidustat.

This randomized, single-treatment, 2-sequence study evaluated the food effect on oral bioavailability of desidustat. Healthy adult male and female subjects were enrolled and randomly assigned to receive desidustat 50mg orally in a fasting state in one period and a fed state in the other period. A standardized high-fat, high-calorie breakfast was served to assess the food effect. The pharmacokinetic results showed that the time to maximum blood concentration of desidustat was delayed significantly (P<.0001) in the fed state compared to the fasting state. The ingestion of food decreased the maximum blood concentration (Cmax ) compared to the fasting state (mean Cmax , 3248ng/mL in the fed state vs 5496ng/mL in the fasting state). The geometric mean ratio of fed/fasting for log-normal (ln) Cmax was 57. The exposure decreased in the fed state compared to the fasting state (mean area under the concentration-time curve [AUC] over the dosing interval, 25 559ng•h/mL in the fed state versus 33 705ng•h/mL in the fasting state; mean AUC from time 0 to infinity [AUC0-∞ ], 25,910ng•h/mL in the fed state vs 34 233ng•h/mL in the fasting state). The geometric mean ratio of fed/fasting for lnAUC from time 0 to the last quantifiable concentration and ln AUC0- ∞ was 77 and 76, respectively. The 90%CI of fed/fasting ratio of the geometric mean of lnCmax , AUC over the dosing interval, and AUC0-∞ of desidustat 50mg did not fall within 80%-125% margin. Therefore, the absence of food effect could not be established. It can be inferred that food has a significant effect on the oral absorption of desidustat. Therefore, food must not be consumed 1hour before and 2hours after the oral administration of desidustat.

AUC Undermines AD Biomarker Discovery

AbstractBackgroundEfforts are underway to accelerate precision medicine for late‐onset Alzheimer disease (AD). The realization of this endeavor can only be attained by identifying AD subtypes and developing methods to diagnose and treat each subtype. Liquid biopsies offer great potential for mounting this challenge. Identification of promising biomarkers requires evaluation of its ability to diagnose AD, and the area under the receiver operating characteristic curve (AUC) is commonly used for this purpose. This metric is equally weighted on the true positive rate (TPR) and false positive rate (FPR). When a given biomarker represents a single subtype, we expect the TPR to be relatively low as only a subset of AD cases are associated. Consequently, we hypothesized that screening based on AUC may discard valuable subtype biomarkers.MethodAn in silico experiment was conducted in which an ‘ideal’ biomarker for a subtype was modeled and the corresponding AUC was computed. A pseudo biomarker was designed to perfectly identify 20% of the cases and all other individuals were assigned levels near zero. Biomarker levels were generated to create a normal distribution centered on 0, with a standard deviation of 0.01, to represent 1000 controls and 800 cases, and 200 additional cases were represented by a normal distribution centered on 1, with a standard deviation of 0.2, to represent the subtype cases (Figs. 1 and 2). The experiment was conducted 1000 times.ResultThe mean AUC across the 1000 trials was 0.6001, with a maximum of 0.6292 and minimum of 0.5624. The standard deviation for the AUC values was 0.0109. The AUC plot for one of the trials is shown in Fig. 3.ConclusionThe pseudo biomarker perfectly identified 20% of the cases while having FPR = 0, yet it would likely be discarded due to the low AUC value, indicating that AUC is not a suitable metric for evaluating biomarkers for AD subtypes. Alternative evaluation metrics should be considered while traversing the path towards successful precision medicine for AD.

A Dual-Branch Network for Diagnosis of Thorax Diseases from Chest X-rays.

Automated chest X-ray analysis has a great potential for diagnosing thorax diseases since errors in diagnosis have always been a concern among radiologists. Being a multi-label classification problem, achieving accurate classification still remains challenging. Several studies have focused on accurately segmenting the lung regions from the chest X-rays to deal with the challenges involved. The features extracted from the lung regions typically provide precise clues for diseases like nodules. However, such methods ignore the features outside the lung regions, which have been shown to be crucial for diagnosing conditions like cardiomegaly. Therefore, in this work, we explore a dual-branch network-based framework that relies on features extracted from the lung regions as well as the entire chest X-rays. The proposed framework uses a novel network named R-I UNet for segmenting the lung regions. The dual-branch network in the proposed framework employs two pre-trained AlexNet models to extract discriminative features, forming two feature vectors. Each of these feature vectors is fed into a recurrent neural network consisting of a stack of gated recurrent units with skip connections. Finally, the resulting feature vectors are concatenated for classification. The R-I UNet has been evaluated on the JSRT and Montgomery (MC) datasets, while the dual-branch classification network has been evaluated on the NIH ChestXray14 dataset. The proposed models achieve state-of-the-art performance for both segmentation and classification tasks on the above benchmark datasets. Specifically, our lung segmentation model achieves a 5-fold cross-validation accuracy of 98.18 % and 99.14 % on MC and JSRT datasets. For classification, the proposed approach achieves state-of-the-art AUC for 9 out of 14 diseases with a mean AUC of 0.842 on NIH ChestXray14 dataset. The source code is available at https://github.com/JigneshChowdary/CXR_Classificationhttps: //github.com / JigneshChowdary /CXR_Classification.

Gully erosion susceptibility assessment based on machine learning-A case study of watersheds in Tuquan County in the black soil region of Northeast China

Owing to its unique topographic and climatic conditions, agricultural production, and anthropogenic grazing, severe gully erosion has developed in the watersheds of Tuquan County located in the black soil region of Northeast China. Using three machine learning models to assess gully erosion susceptibility in this region, the main purpose of this study was to determine the important factors affecting the occurrence of gully erosion from 25 geo-environmental factors and select the model with the best prediction performance. Field data collection was conducted for 823 gullies. 12,496 pixels corresponding to 823 gullies and 12,496 pixels without gully erosion were randomly selected to form the model calculation database. According to the principle of multiple screening, this study selected 25 geo-environmental factors affecting the occurrence of gully erosion. Two factors, coarse sand content and fine sand content, were excluded by multi-collinearity analysis and using random forest (RF), convolution neural network (CNN), and transformer models, combining the 10-fold cross-validation and related validation metrics to assess the susceptibility of gully erosion in Tuquan County’s small watersheds. The results show that the convergence index (CI), topographical wetness index (TWI), terrain ruggedness index (TRI), and distance from the river based on the RF model are important factors affecting the occurrence of gully erosion. The transformer model (mean AUC of training = 95.338 %, mean AUC of validation = 95.342 %) was superior to the RF (mean AUC of training = 99.997 %, mean AUC of validation = 92.518 %) and CNN (mean AUC of training = 86.995 %, mean AUC of validation = 84.647 %) models in predicting performance. To ensure that the management and decision-making departments take reasonable and effective measures to prevent gully erosion in the study area, the gully erosion susceptibility maps produced using the three machine learning models were beneficial.

Detection of Malicious Websites Using Symbolic Classifier

Malicious websites are web locations that attempt to install malware, which is the general term for anything that will cause problems in computer operation, gather confidential information, or gain total control over the computer. In this paper, a novel approach is proposed which consists of the implementation of the genetic programming symbolic classifier (GPSC) algorithm on a publicly available dataset to obtain a simple symbolic expression (mathematical equation) which could detect malicious websites with high classification accuracy. Due to a large imbalance of classes in the initial dataset, several data sampling methods (random undersampling/oversampling, ADASYN, SMOTE, BorderlineSMOTE, and KmeansSMOTE) were used to balance the dataset classes. For this investigation, the hyperparameter search method was developed to find the combination of GPSC hyperparameters with which high classification accuracy could be achieved. The first investigation was conducted using GPSC with a random hyperparameter search method and each dataset variation was divided on a train and test dataset in a ratio of 70:30. To evaluate each symbolic expression, the performance of each symbolic expression was measured on the train and test dataset and the mean and standard deviation values of accuracy (ACC), AUC, precision, recall and f1-score were obtained. The second investigation was also conducted using GPSC with the random hyperparameter search method; however, 70%, i.e., the train dataset, was used to perform 5-fold cross-validation. If the mean accuracy, AUC, precision, recall, and f1-score values were above 0.97 then final training and testing (train/test 70:30) were performed with GPSC with the same randomly chosen hyperparameters used in a 5-fold cross-validation process and the final mean and standard deviation values of the aforementioned evaluation methods were obtained. In both investigations, the best symbolic expression was obtained in the case where the dataset balanced with the KMeansSMOTE method was used for training and testing. The best symbolic expression obtained using GPSC with the random hyperparameter search method and classic train–test procedure (70:30) on a dataset balanced with the KMeansSMOTE method achieved values of ACC¯, AUC¯, Precsion¯, Recall¯ and F1-score¯ (with standard deviation) 0.9992±2.249×10−5, 0.9995±9.945×10−6, 0.9995±1.09×10−5, 0.999±5.17×10−5, 0.9992±5.17×10−6, respectively. The best symbolic expression obtained using GPSC with a random hyperparameter search method and 5-fold cross-validation on a dataset balanced with the KMeansSMOTE method achieved values of ACC¯, AUC¯, Precsion¯, Recall¯ and F1-score¯ (with standard deviation) 0.9994±1.13×10−5, 0.9994±1.2×10−5, 1.0±0, 0.9988±2.4×10−5, and 0.9994±1.2×10−5, respectively.

The effect of the acute phase of infection on absorption of and exposure to orally administered antibiotics in non-critically ill, hospitalized patients.

During the acute phase of infection, IV antibiotics are preferred to ensure adequate systemic exposure. To assess whether adequate exposure may also be achieved with oral antibiotics, we investigated exposure to oral antibiotics and PTA during the acute phase of infection and after defervescence. We enrolled hospitalized, non-critically ill febrile patients treated with IV antibiotics other than amoxicillin or ciprofloxacin. The study consisted of two visits: when patients had received <24 h IV treatment; and when patients had become afebrile. On both visits, patients received one additional dose of 750 mg amoxicillin, or 500 mg ciprofloxacin, depending on the presumed infection, after which serial blood samples were obtained. The primary endpoint was the ratio of the AUC during the febrile and the afebrile phase. The AUCs were considered to be equivalent when the ratio of the mean AUCs and its 90% CI was contained within the acceptance interval of 80%-125%. The secondary endpoint was PTA. Forty-four patients (15 amoxicillin, 29 ciprofloxacin) completed both study visits. The median time between the two study visits was 65.8 h (range 33.8-427.4). The ratio of the mean AUCs (study visit 1/study visit 2) was 97% (90% CI of 80%-117%) for amoxicillin and 112% (90% CI of 108%-116%) for ciprofloxacin. The PTA for amoxicillin and ciprofloxacin did not differ between the two phases and was adequate to treat common pathogens. The acute phase of infection in non-critically ill febrile patients does not influence the exposure to, or PTA of, orally administered amoxicillin and ciprofloxacin. This might justify earlier IV-to-oral switching.

Machine Learning Models for Prediction of Posttreatment Recurrence in Early-Stage Hepatocellular Carcinoma Using Pretreatment Clinical and MRI Features: A Proof-of-Concept Study.

BACKGROUND. Posttreatment recurrence is an unpredictable complication after liver transplant for hepatocellular carcinoma (HCC) that is associated with poor survival. Biomarkers are needed to estimate recurrence risk before organ allocation. OBJECTIVE. This proof-of-concept study evaluated the use of machine learning (ML) to predict recurrence from pretreatment laboratory, clinical, and MRI data in patients with early-stage HCC initially eligible for liver transplant. METHODS. This retrospective study included 120 patients (88 men, 32 women; median age, 60.0 years) with early-stage HCC diagnosed who were initially eligible for liver transplant and underwent treatment by transplant, resection, or thermal ablation between June 2005 and March 2018. Patients underwent pretreatment MRI and posttreatment imaging surveillance. Imaging features were extracted from postcontrast phases of pretreatment MRI examinations using a pretrained convolutional neural network. Pretreatment clinical characteristics (including laboratory data) and extracted imaging features were integrated to develop three ML models (clinical model, imaging model, combined model) for predicting recurrence within six time frames ranging from 1 through 6 years after treatment. Kaplan-Meier analysis with time to recurrence as the endpoint was used to assess the clinical relevance of model predictions. RESULTS. Tumor recurred in 44 of 120 (36.7%) patients during follow-up. The three models predicted recurrence with AUCs across the six time frames of 0.60-0.78 (clinical model), 0.71-0.85 (imaging model), and 0.62-0.86 (combined model). The mean AUC was higher for the imaging model than the clinical model (0.76 vs 0.68, respectively; p = .03), but the mean AUC was not significantly different between the clinical and combined models or between the imaging and combined models (p > .05). Kaplan-Meier curves were significantly different between patients predicted to be at low risk and those predicted to be at high risk by all three models for the 2-, 3-, 4-, 5-, and 6-year time frames (p < .05). CONCLUSION. The findings suggest that ML-based models can predict recurrence before therapy allocation in patients with early-stage HCC initially eligible for liver transplant. Adding MRI data as model input improved predictive performance over clinical parameters alone. The combined model did not surpass the imaging model's performance. CLINICAL IMPACT. ML-based models applied to currently underutilized imaging features may help design more reliable criteria for organ allocation and liver transplant eligibility.

Pseudo-siamese network combined with dosimetric and clinical factors, radiomics features, CT images and 3D dose distribution for the prediction of radiation pneumonitis: A feasibility study

PurposeRadiation pneumonitis (RP)(grade ≥ 2) can have a considerable impact on patient quality-of life. In previous studies, the traditional method commonly used radiomics and clinical factors for RP prediction. This study aims to develop and evaluate a novel pseudo-siamese network (PSN) to assist radiologists predict RP before radiotherapy based on combination of dosimetric and clinical factors, radiomics features, CT (computed tomography) images, and dose distribution (hybrid model). MethodOne hundred and ten patients with lung cancer (19 RP ≥ 2) who received radiotherapy between 2016 and 2020 were retrospectively enrolled in this study. Dosimetric factors were calculated from DVH (dose-volume histogram), such as lung mean dose, lung V5, and prescription dose. Clinical characteristics were recorded, such as age, sex, smoking status, TN stage, and overall stage. A total of 1419 radiomics features were extracted. Cluster analysis was used for detecting radiomics features that associated with RP. Patients were randomly split into a training set (90 %, 85 non-RP, and 14 RP) and a validation set (10 %, 6 non-RP, and 5 RP). A PSN architecture was designed for combining 1D (dosimetric and clinical factors, radiomics) and 3D (CT images, 3D dose distribution) features. 5-fold cross-validation procedure for estimating the skill of the model on new data. ResultsFor cluster analysis, totally of 106 radiomics features with high correlation were selected. The accuracy was 0.727, 0.636, 0.545, and 0.727 for input dosimetric and clinical factors, dose distribution, CT images, and radiomics features, respectively. The accuracy of hybrid model was 0.818. The sensitivity of hybrid model was 0.800 (95 % confidence interval (CI) [0.299, 0.989]), and specificity was 0.833(95 % CI [0.364, 0.991]). The areas under the receiver operating characteristic curves (AUCs) result in 5-fold cross-validation was 0.77–0.90(mean AUC ± std was 0.85 ± 0.05). ConclusionThis study firstly propose method that the combination of high dimensional and low dimensional features for RP prediction. The results confirm the feasibility of multi-dimensional features predict RP.

A Facile machine learning multi-classification model for Streptococcus agalactiae clonal complexes

BackgroundThe clinical significance of group B streptococcus (GBS) was different among different clonal complexes (CCs), accurate strain typing of GBS would facilitate clinical prognostic evaluation, epidemiological investigation and infection control. The aim of this study was to construct a practical and facile CCs prediction model for S. agalactiae.MethodsA total of 325 non-duplicated GBS strains were collected from clinical samples in Xinhua Hospital, Shanghai, China. Multilocus sequence typing (MLST) method was used for molecular classification, the results were analyzed to derive CCs by Bionumeric 8.0 software. Antibiotic susceptibility test was performed using Vitek-2 Compact system combined with K-B method. Multiplex PCR method was used for serotype identification. A total of 45 virulence genes associated with adhesion, invasion, immune evasion were detected by PCR method and electrophoresis. Three types of features, including antibiotic susceptibility (A), serotypes (S) and virulence genes (V) tests, and XGBoost algorithm was established to develop multi-class CCs identification models. The performance of proposed models was evaluated by the receiver operating characteristic curve (ROC).ResultsThe 325 GBS were divided into 47 STs, and then calculated into 7 major CCs, including CC1, CC10, CC12, CC17, CC19, CC23, CC24. A total of 18 features in three kinds of tests (A, S, V) were significantly different from each CC. The model based on all the features (S&A&V) performed best with AUC 0.9536. The model based on serotype and antibiotic resistance (S&A) only enrolled 5 weighed features, performed well in predicting CCs with mean AUC 0.9212, and had no statistical difference in predicting CC10, CC12, CC17, CC19, CC23 and CC24 when compared with S&A&V model (all p > 0.05).ConclusionsThe S&A model requires least parameters while maintaining a high accuracy and predictive power of CCs prediction. The established model could be used as a promising tool to classify the GBS molecular types, and suggests a substantive improvement in clinical application and epidemiology surveillance in GBS phenotyping.

Comparative polymyxin B pharmacokinetics in critically ill patients with renal insufficiency and in continuous veno-venous hemodialysis.

The aim of this study was to assess polymyxin B pharmacokinetics (PK) in patients with varying degrees of renal dysfunction and in patients who require continuous veno-venous hemodialysis (CVVHD). The study enrolled 37 patients with sepsis, including 13 patients with glomerular filtration rate (GFR) below 80 mL/min and 11 patients on CVVHD. Each patient received a loading dose of polymyxin B (200-300 mg) and at least 3 subsequent doses of 100-150 mg every 12 h. For every patient, 6-8 blood samples were collected between doses. Polymyxin B (PMB) serum concentration was determined using enzyme-linked immunosorbent assay. In sepsis, patients with preserved renal function mean area under the curve over 24 h (AUC0-24 h) value reached 67.8 ± 9.8 mg*h/L, while in patients with GFR below 80 mL/min, mean AUC0-24 h was 87 ± 5.8 mg*h/L. PMB PK in patients with renal insufficiency was characterized by significantly lower clearance (CL) compared to the normal renal function group (2.1 ± 0.1 L/h vs 3.9 ± 0.4 L/h respectively). In patients on CVVHD, mean AUC0-24 h was 110.4 ± 10.3 mg*h/L, while CL reached 2 ± 0.23 L/h. The median recovery rate from dialysate constituted 22%. Simulation of different dosage regimens that indicate a fixed maintenance dose of 100 mg q12h with a loading dose of 200 mg is optimal for patients on CVVHD, and no dosage increase is required. This study demonstrates decreased clearance of PMB in patients with renal insufficiency, which puts them at risk of toxicity. Therefore, patients with extremes of renal function might benefit from therapeutic drug monitoring. For patients with anuria, who require CVVHD, we suggest a fixed dose of 100 mg q12h.

Colon-Targeted Layer-by-Layer Self-assembled Film: Pharmacokinetic Analysis of BCS Class I and Class III Model Drugs.

The aim of the study was to investigate the pharmacokinetic parameters of 5-fluorouracil (5FU) and moxifloxacin HCl (MF) after oral administration using layer-by-layer assembled film in enteric-coated capsule. The layer-by-layer (LbL) film was prepared by sequential layering of chitosan and sodium alginate polyelectrolytes containing either 5FU or MF. The films were in vitro evaluated for physical characteristics, drug loading and release behaviour. In vivo pharmacokinetic evaluation was performed in the rat model for three different drug concentrations after oral administration and compared with intravenous administration. The results showed that the thickness of 10-bilayer film was 147 ± 11.66µm and 212.3 ± 7.19µm after 5FU and MF loading, respectively. The LbL film with backing layer provided directional release of 5FU and MF, where 63.81 ± 4.52% and 101.38 ± 5.08%, respectively, was released in 24h. 5FU showed non-linear pharmacokinetics compared with linear pharmacokinetics shown by MF after oral administration. There is a dose-dependent increase in Cmax after oral administration of 5FU and MF LbL film. The Tmax was found to be 720min and 840min for 5FU and MF after oral administration. The mean residence time and AUC0-24 at 45mg/kg were 871.4 ± 6.45min and 198.6 ± 5.03 × 103min per ng/mL and 1267 ± 142.4min and 1590 ± 55.60 103min per ng/mL for 5FU and MF, respectively. Taken together, colon-targeted LbL film can be developed for oral administration of drugs for local and systemic applications.

Semi-Supervised Approach for EGFR Mutation Prediction on CT Images

The use of deep learning methods in medical imaging has been able to deliver promising results; however, the success of such models highly relies on large, properly annotated datasets. The annotation of medical images is a laborious, expensive, and time-consuming process. This difficulty is increased for the mutations status label since these require additional exams (usually biopsies) to be obtained. On the other hand, raw images, without annotations, are extensively collected as part of the clinical routine. This work investigated methods that could mitigate the labelled data scarcity problem by using both labelled and unlabelled data to improve the efficiency of predictive models. A semi-supervised learning (SSL) approach was developed to predict epidermal growth factor receptor (EGFR) mutation status in lung cancer in a less invasive manner using 3D CT scans.The proposed approach consists of combining a variational autoencoder (VAE) and exploiting the power of adversarial training, intending that the features extracted from unlabelled data to discriminate images can help in the classification task. To incorporate labelled and unlabelled images, adversarial training was used, extending a traditional variational autoencoder. With the developed method, a mean AUC of 0.701 was achieved with the best-performing model, with only 14% of the training data being labelled. This SSL approach improved the discrimination ability by nearly 7 percentage points over a fully supervised model developed with the same amount of labelled data, confirming the advantage of using such methods when few annotated examples are available.

28-day repeated dose toxicity and toxicokinetics study on new melatonergic antidepressant GW117 in beagle dogs.

GW117 is new melatonergic antidepressant being developed to show better antidepressant action than agomelatine. The purpose of this study was to evaluate the toxicity and to determine potential target organs after oral (gavage) administration of the test article GW117 for 28 days and to assess the reversibility after a 4-week recovery phase in beagle dogs. Toxicokinetics was also evaluated. Four groups were designed in this study, including the vehicle control group and the GW117 50, 150 and 500 mg/kg/day groups, with 5 dogs/sex/group. Body weight, hematology, clinical chemistry, gross necropsy, organ weight, histopathology, and other indicators were examined. Results showed that animals dosed at ≥150 mg/kg/day showed gastrointestinal reactions (watery feces and dark green/red brown feces), with a dose-response relationship in the incidence and severity grade. Female dogs at 500 mg/kg/day had an increase in organ weight and ratios of the liver at the end of the dosing phase. Histopathology examination showed that some animals at 500 mg/kg/day, especially female animals, had minimal centrilobular hepatocyte hypertrophy in the liver, which reversed after 28-day recovery. With the exception of the above, no GW117-related abnormality was noted. Meanwhile, there were no sexual differences in drug exposure and accumulation after the first and last dosing. The no observed adverse effect dose level (NOAEL) was 150 mg/kg/day, under which mean Cmax and AUC0 → t were 583.5 and 2767.0 ng/ml*h for females and 663.2 and 4046.3 ng/ml*h for males on Day 28.