-
Journal of Imaging Mar 2023The rapid spread of deceptive information on the internet can have severe and irreparable consequences. As a result, it is important to develop technology that can...
The rapid spread of deceptive information on the internet can have severe and irreparable consequences. As a result, it is important to develop technology that can detect fake news. Although significant progress has been made in this area, current methods are limited because they focus only on one language and do not incorporate multilingual information. In this work, we propose Multiverse-a new feature based on multilingual evidence that can be used for fake news detection and improve existing approaches. Our hypothesis that cross-lingual evidence can be used as a feature for fake news detection is supported by manual experiments based on a set of true (legit) and fake news. Furthermore, we compared our fake news classification system based on the proposed feature with several baselines on two multi-domain datasets of general-topic news and one fake COVID-19 news dataset, showing that (in combination with linguistic features) it yields significant improvements over the baseline models, bringing additional useful signals to the classifier.
PubMed: 37103228
DOI: 10.3390/jimaging9040077 -
BMC Medical Research Methodology Mar 2023In pre-post designs, analysis of covariance (ANCOVA) is a standard technique to detect the treatment effect with a continuous variable measured at baseline and...
BACKGROUND
In pre-post designs, analysis of covariance (ANCOVA) is a standard technique to detect the treatment effect with a continuous variable measured at baseline and follow-up. For measurements subject to a high degree of variability, it may be advisable to repeat the pre-treatment and/or follow-up assessments. In general, repeating the follow-up measurements is more advantageous than repeating the pre-treatment measurements, while the latter can still be valuable and improve efficiency in clinical trials.
METHODS
In this article, we report investigations of using multiple pre-treatment and post-treatment measurements in randomized clinical trials. We consider the sample size formula for ANCOVA under general correlation structures with the pre-treatment mean included as the covariate and the mean follow-up value included as the response. We propose an optimal experimental design of multiple pre-post allocations under a specified constraint, that is, given the total number of pre-post treatment visits. The optimal number of the pre-treatment measurements is derived. For non-linear models, closed-form formulas for sample size/power calculations are generally unavailable, but we conduct Monte Carlo simulation studies instead.
RESULTS
Theoretical formulas and simulation studies show the benefits of repeating the pre-treatment measurements in pre-post randomized studies. The optimal pre-post allocation derived from the ANCOVA extends well to binary measurements in simulation studies, using logistic regression and generalized estimating equations (GEE).
CONCLUSIONS
Repeating baselines and follow-up assessments is a valuable and efficient technique in pre-post design. The proposed optimal pre-post allocation designs can minimize the sample size, i.e., achieve maximum power.
Topics: Humans; Randomized Controlled Trials as Topic; Research Design; Sample Size; Computer Simulation; Logistic Models
PubMed: 36978004
DOI: 10.1186/s12874-023-01893-w -
Magnetic Resonance in Medicine May 2019Subject motion and static field (B ) drift are known to reduce the quality of single voxel MR spectroscopy data due to incoherent averaging. Retrospective correction has...
PURPOSE
Subject motion and static field (B ) drift are known to reduce the quality of single voxel MR spectroscopy data due to incoherent averaging. Retrospective correction has previously been shown to improve data quality by adjusting the phase and frequency offset of each average to match a reference spectrum. In this work, a new method (RATS) is developed to be tolerant to large frequency shifts (>7 Hz) and baseline instability resulting from inconsistent water suppression.
METHODS
In contrast to previous approaches, the variable-projection method and baseline fitting is incorporated into the correction procedure to improve robustness to fluctuating baseline signals and optimization instability. RATS is compared to an alternative method, based on time-domain spectral registration (TDSR), using simulated data to model frequency, phase, and baseline instability. In addition, a J-difference edited glutathione in-vivo dataset is processed using both approaches and compared.
RESULTS
RATS offers improved accuracy and stability for large frequency shifts and unstable baselines. Reduced subtraction artifacts are demonstrated for glutathione edited MRS when using RATS, compared with uncorrected or TDSR corrected spectra.
CONCLUSIONS
The RATS algorithm has been shown to provide accurate retrospective correction of SVS MRS data in the presence of large frequency shifts and baseline instability. The method is rapid, generic and therefore readily incorporated into MRS processing pipelines to improve lineshape, SNR, and aid quality assessment.
Topics: Algorithms; Artifacts; Brain; Computer Simulation; Glutathione; Healthy Volunteers; Humans; Image Processing, Computer-Assisted; Lipids; Magnetic Resonance Spectroscopy; Motion; Signal-To-Noise Ratio; Water
PubMed: 30417937
DOI: 10.1002/mrm.27605 -
Scientific Reports Dec 2022It is critical for hospitals to accurately predict patient length of stay (LOS) and mortality in real-time. We evaluate temporal convolutional networks (TCNs) and data...
It is critical for hospitals to accurately predict patient length of stay (LOS) and mortality in real-time. We evaluate temporal convolutional networks (TCNs) and data rebalancing methods to predict LOS and mortality. This is a retrospective cohort study utilizing the MIMIC-III database. The MIMIC-Extract pipeline processes 24 hour time-series clinical objective data for 23,944 unique patient records. TCN performance is compared to both baseline and state-of-the-art machine learning models including logistic regression, random forest, gated recurrent unit with decay (GRU-D). Models are evaluated for binary classification tasks (LOS > 3 days, LOS > 7 days, mortality in-hospital, and mortality in-ICU) with and without data rebalancing and analyzed for clinical runtime feasibility. Data is split temporally, and evaluations utilize tenfold cross-validation (stratified splits) followed by simulated prospective hold-out validation. In mortality tasks, TCN outperforms baselines in 6 of 8 metrics (area under receiver operating characteristic, area under precision-recall curve (AUPRC), and F-1 measure for in-hospital mortality; AUPRC, accuracy, and F-1 for in-ICU mortality). In LOS tasks, TCN performs competitively to the GRU-D (best in 6 of 8) and the random forest model (best in 2 of 8). Rebalancing improves predictive power across multiple methods and outcome ratios. The TCN offers strong performance in mortality classification and offers improved computational efficiency on GPU-enabled systems over popular RNN architectures. Dataset rebalancing can improve model predictive power in imbalanced learning. We conclude that temporal convolutional networks should be included in model searches for critical care outcome prediction systems.
Topics: Humans; Prospective Studies; Retrospective Studies
PubMed: 36481828
DOI: 10.1038/s41598-022-25472-z -
Radiology. Artificial Intelligence Mar 2023To assess whether transfer learning with a bidirectional encoder representations from transformers (BERT) model, pretrained on a clinical corpus, can perform...
PURPOSE
To assess whether transfer learning with a bidirectional encoder representations from transformers (BERT) model, pretrained on a clinical corpus, can perform sentence-level anatomic classification of free-text radiology reports, even for anatomic classes with few positive examples.
MATERIALS AND METHODS
This retrospective study included radiology reports of patients who underwent whole-body PET/CT imaging from December 2005 to December 2020. Each sentence in these reports (6272 sentences) was labeled by two annotators according to body part ("brain," "head & neck," "chest," "abdomen," "limbs," "spine," or "others"). The BERT-based transfer learning approach was compared with two baseline machine learning approaches: bidirectional long short-term memory (BiLSTM) and the count-based method. Area under the precision-recall curve (AUPRC) and area under the receiver operating characteristic curve (AUC) were computed for each approach, and AUCs were compared using the DeLong test.
RESULTS
The BERT-based approach achieved a macro-averaged AUPRC of 0.88 for classification, outperforming the baselines. AUC results for BERT were significantly higher than those of BiLSTM for all classes and those of the count-based method for the "brain," "chest," "abdomen," and "others" classes ( values < .025). AUPRC results for BERT were superior to those of baselines even for classes with few labeled training data (brain: BERT, 0.95, BiLSTM, 0.11, count based, 0.41; limbs: BERT, 0.74, BiLSTM, 0.28, count based, 0.46; spine: BERT, 0.82, BiLSTM, 0.53, count based, 0.69).
CONCLUSION
The BERT-based transfer learning approach outperformed the BiLSTM and count-based approaches in sentence-level anatomic classification of free-text radiology reports, even for anatomic classes with few labeled training data. Anatomy, Comparative Studies, Technology Assessment, Transfer Learning © RSNA, 2023.
PubMed: 37035437
DOI: 10.1148/ryai.220097 -
Anesthesiology Jan 2002Artifact robustness (i.e., size of deviation of an electroencephalographic parameter value from baseline caused by artifacts) and baseline stability (i.e., consistency...
BACKGROUND
Artifact robustness (i.e., size of deviation of an electroencephalographic parameter value from baseline caused by artifacts) and baseline stability (i.e., consistency of median baseline values) of electroencephalographic parameters profoundly influence electroencephalography-based pharmacodynamic parameter estimation and the usefulness of the processed electroencephalogram as measure of the arousal state of the central nervous system (depth of anesthesia). In this study, the authors compared the artifact robustness and the interindividual and intraindividual baseline stability of several univariate descriptors of the electroencephalogram (Shannon entropy, approximate entropy, spectral edge frequency 95, delta ratio, and canonical univariate parameter).
METHODS
Electroencephalographic data of 16 healthy volunteers before and after administration of an intravenous bolus of propofol (2 mg/kg body weight) were analyzed. Each volunteer was studied twice. The baseline electroencephalogram was recorded for a median of 18 min before drug administration. For each electroencephalographic descriptor, the authors calculated the following: (1) baseline variability (= (median baseline - median effect) [i.e., signal]/SD baseline [i.e., noise]) without artifact rejection; (2) baseline variability with artifact rejection; and (3) baseline stability within and between individuals (= (median baseline - median effect) averaged over all volunteers/SD of all median baselines).
RESULTS
Without artifact rejection, Shannon entropy and canonical univariate parameter displayed the highest signal-to-noise ratio. After artifact rejection, approximate entropy, Shannon entropy, and the canonical univariate parameter displayed the highest signal-to-noise ratio. Baseline stability within and between individuals was highest for approximate entropy.
CONCLUSIONS
With regard to robustness against artifacts, the electroencephalographic entropy parameters and the canonical univariate parameter were superior to spectral edge frequency 95 and delta ratio. Electroencephalographic approximate entropy displayed the best interindividual and intraindividual baseline stability.
Topics: Adult; Aged; Artifacts; Electroencephalography; Humans; Middle Aged; Thermodynamics
PubMed: 11753002
DOI: 10.1097/00000542-200201000-00015 -
Journal of Digital Imaging Oct 2022To visualise the tumours inside the body on a screen, a long and thin tube is inserted with a light source and a camera at the tip to obtain video frames inside organs...
To visualise the tumours inside the body on a screen, a long and thin tube is inserted with a light source and a camera at the tip to obtain video frames inside organs in endoscopy. However, multiple artefacts exist in these video frames that cause difficulty during the diagnosis of cancers. In this research, deep learning was applied to detect eight kinds of artefacts: specularity, bubbles, saturation, contrast, blood, instrument, blur, and imaging artefacts. Based on transfer learning with pre-trained parameters and fine-tuning, two state-of-the-art methods were applied for detection: faster region-based convolutional neural networks (Faster R-CNN) and EfficientDet. Experiments were implemented on the grand challenge dataset, Endoscopy Artefact Detection and Segmentation (EAD2020). To validate our approach in this study, we used phase I of 2,200 frames and phase II of 331 frames in the original training dataset with ground-truth annotations as training and testing dataset, respectively. Among the tested methods, EfficientDet-D2 achieves a score of 0.2008 (mAP[Formula: see text]0.6+mIoU[Formula: see text]0.4) on the dataset that is better than three other baselines: Faster-RCNN, YOLOv3, and RetinaNet, and competitive to the best non-baseline result scored 0.25123 on the leaderboard although our testing was on phase II of 331 frames instead of the original 200 testing frames. Without extra improvement techniques beyond basic neural networks such as test-time augmentation, we showed that a simple baseline could achieve state-of-the-art performance in detecting artefacts in endoscopy. In conclusion, we proposed the combination of EfficientDet-D2 with suitable data augmentation and pre-trained parameters during fine-tuning training to detect the artefacts in endoscopy.
Topics: Humans; Artifacts; Neural Networks, Computer; Endoscopy; Machine Learning
PubMed: 35478060
DOI: 10.1007/s10278-022-00627-6 -
Journal of Geophysical Research. Solid... Nov 2022Accurate earthquake location and magnitude estimation play critical roles in seismology. Recent deep learning frameworks have produced encouraging results on various...
Accurate earthquake location and magnitude estimation play critical roles in seismology. Recent deep learning frameworks have produced encouraging results on various seismological tasks (e.g., earthquake detection, phase picking, seismic classification, and earthquake early warning). Many existing machine learning earthquake location methods utilize waveform information from a single station. However, multiple stations contain more complete information for earthquake source characterization. Inspired by recent successes in applying graph neural networks (GNNs) in graph-structured data, we develop a Spatiotemporal Graph Neural Network (STGNN) for estimating earthquake locations and magnitudes. Our graph neural network leverages geographical and waveform information from multiple stations to construct graphs automatically and dynamically by adaptive message passing based on graphs' edges. Using a recent graph neural network and a fully convolutional neural network as baselines, we apply STGNN to earthquakes recorded by the Southern California Seismic Network from 2000 to 2019 and earthquakes collected in Oklahoma from 2014 to 2015. STGNN yields more accurate earthquake locations than those obtained by the baseline models and performs comparably in terms of depth and magnitude prediction, though the ability to predict depth and magnitude remains weak for all tested models. Our work demonstrates the potential of using GNNs and multiple stations for better automatic estimation of earthquake epicenters.
PubMed: 37033773
DOI: 10.1029/2022JB024401 -
Sensors (Basel, Switzerland) May 2019Spaceborne multistatic synthetic aperture radar (SAR) tomography (SMS-TomoSAR) systems take full advantage of the flexible configuration of multistatic SAR in the space,...
Spaceborne multistatic synthetic aperture radar (SAR) tomography (SMS-TomoSAR) systems take full advantage of the flexible configuration of multistatic SAR in the space, time, phase, and frequency dimensions, and simultaneously achieve high-precision height resolution and low-deformation measurement of three-dimensional ground scenes. SMS-TomoSAR currently poses a series of key issues to solve, such as baseline optimization, spatial transmission error estimation and compensation, and the choice of imaging algorithm, which directly affects the performance of height-dimensional imaging and surface deformation measurement. This paper explores the impact of baseline distribution on height-dimensional imaging performance for the baseline optimization issue, and proposes a feasible baseline optimization method. Firstly, the multi-base multi-pass baselines of an SMS-TomoSAR system are considered equivalent to a group of multi-pass baselines from monostatic SAR. Secondly, we establish the equivalent baselines as a symmetric-geometric model to characterize the non-uniform characteristic of baseline distribution. Through experimental simulation and model analysis, an approximately uniform baseline distribution is shown to have better SMS-TomoSAR imaging performance in the height direction. Further, a baseline design method under uniform-perturbation sampling with Gaussian distribution error is proposed. Finally, the imaging performance of different levels of perturbation is compared, and the maximum baseline perturbation allowed by the system is given.
PubMed: 31067712
DOI: 10.3390/s19092106 -
Journal of Biomedical Informatics Jul 2019Information curation and literature surveillance efforts that synthesize the current knowledge about the impact of genetic variability on disease states and drug...
BACKGROUND
Information curation and literature surveillance efforts that synthesize the current knowledge about the impact of genetic variability on disease states and drug responses are vitally important for the practise of evidence-based precision medicine. For these efforts, finding the relevant and comprehensive set of articles from the ever growing scientific literature is a challenge.
METHODS
We have designed and developed Article Retrieval for Precision Medicine (ARtPM), an end-to-end article retrieval system that employs multi-stage architecture to retrieve and rank relevant articles for a given medical case summary (genetic variants, disease, demographic, and other medical conditions). We compared ARtPM with five baselines, including PubMed Best Match, the improved search functionality recently introduced by PubMed.
RESULTS
The differences in the performance of ARtPM and five baselines were statistically significant for four metrics that quantify different aspects of search effectiveness (P-values for P@10, R-prec, infNDCG, Recall@1000 were <.001, <.001,.003,.009, respectively). Pairwise systems' comparisons show that ARtPM is comparable or better than the best performing baseline on three metrics (R-prec: 0.324 vs 0.299, P-value=.06; infNDCG: 0.556 vs 0.465, P-value=.08; R@1000: 0.665 vs 0.572, P-value=.007), but performance in P@10 (0.603 vs 0.630, P-value:.64) needs to improve.
CONCLUSION
The recall-focused phase of the ARtPM is effective at retrieving more relevant articles. The precision-focused ranking phase performs well at deeper ranks but needs further work on early ranks (e.g., richer feature set). Overall, the ARtPM system effectively facilitates evidence-based precision medicine practice, and provides a robust search framework for further work in this direction.
Topics: Biomedical Research; Data Curation; Databases, Factual; Humans; Information Storage and Retrieval; Periodicals as Topic; Precision Medicine
PubMed: 31200123
DOI: 10.1016/j.jbi.2019.103224