The prognostic utility of serum thyrotropin in hospitalized Covid-19 patients: statistical and machine learning approaches

PURPOSE

To assess the prognostic value of serum TSH in Greek patients with COVID-19 and compare it with that of commonly used prognostic biomarkers.

METHODS

Retrospective study of 128 COVID-19 in patients with no history of thyroid disease. Serum TSH, albumin, CRP, ferritin, and D-dimers were measured at admission. Outcomes were classified as "favorable" (discharge from hospital) and "adverse" (intubation or in-hospital death of any cause). The prognostic performance of TSH and other indices was assessed using binary logistic regression, machine learning classifiers, and ROC curve analysis.

RESULTS

Patients with adverse outcomes had significantly lower TSH compared to those with favorable outcomes (0.61 versus 1.09 mIU/L, p < 0.001). Binary logistic regression with sex, age, TSH, albumin, CRP, ferritin, and D-dimers as covariates showed that only albumin (p < 0.001) and TSH (p = 0.006) were significantly predictive of the outcome. Serum TSH below the optimal cut-off value of 0.5 mIU/L was associated with an odds ratio of 4.13 (95% C.I.: 1.41-12.05) for adverse outcome. Artificial neural network analysis showed that the prognostic importance of TSH was second only to that of albumin. However, the prognostic accuracy of low TSH was limited, with an AUC of 69.5%, compared to albumin's 86.9%. A Naïve Bayes classifier based on the combination of serum albumin and TSH levels achieved high prognostic accuracy (AUC 99.2%).

CONCLUSION

Low serum TSH is independently associated with adverse outcome in hospitalized Greek patients with COVID-19 but its prognostic utility is limited. The integration of serum TSH into machine learning classifiers in combination with other biomarkers enables outcome prediction with high accuracy.

A Semi-Supervised Method for Tumor Segmentation in Mammogram Images

Background:

Breast cancer is one of the most common cancers in women. Mammogram images have an important role in the treatment of various states of this cancer. In recent years, machine learning methods have been widely used for tumor segmentation in mammogram images. Pixel-based segmentation methods have been presented using both supervised and unsupervised learning approaches. Supervised learning methods are usually fast and accurate, but they usually use a large number of labeled data. Besides, providing these samples is very hard and usually expensive. Unsupervised learning methods do not require the labels of the training data for decision making and they completely ignore the prior knowledge that may lead to a low performance. Semi-supervised learning methods which use a small number of labeled data solve the problem of providing the high number of samples in supervised methods, while they usually result in a higher accuracy in comparison to the unsupervised methods.

Methods:

In this study, we used a semisupervised method for tumor segmentation in which the pixel information is used for the classification. The static and gray level run length matrix features for each pixel are considered as the features, and Fisher discriminant analysis (FDA) is used for feature reduction. A cotraining algorithm based on support vector machine and Bayes classifiers is proposed for tumor segmentation on MIAS data set.

Results and Conclusion:

The results show that the proposed method outperforms both supervised methods.

Automated Assessment of Hematoma Volume of Rodents Subjected to Experimental Intracerebral Hemorrhagic Stroke by Bayes Segmentation Approach

Simulating a clinical condition of intracerebral hemorrhage (ICH) in animals is key to research on the development and testing of diagnostic or treatment strategies for this high-mortality disease. In order to study the mechanism, pathology, and treatment for hemorrhagic stroke, various animal models have been developed. Measurement of hematoma volume is an important assessment parameter to evaluate post-ICH outcomes. However, due to tissue preservation conditions and variables in digitization, quantification of hematoma volume is usually labor intensive and sometimes even subjective. The objective of this study is to develop an automated method that can accurately and efficiently obtain unbiased cerebral hematoma volume. We developed an application (MATLAB program) that can delineate the brain slice from the background and use the Hue information in the Hue/Saturation/Value (HSV) color space to segment the hematoma region. The segmentation threshold of Hue is calculated based on the Bayes classifier theorem so that the minimum error is mathematically ensured and automated processing is enabled. To validate the developed method, we compared the outcomes from the developed method with the hemoglobin content by the spectrophotometric assay method. The results were linearly correlated with statistical significance. The method was also validated by digital phantoms with an error less than 5% compared with the ground truth from the phantoms. Hematoma volumes yielded by the automated processing and those obtained by the operator's manual operation are highly correlated. This automated segmentation approach can be potentially used to quantify hemorrhagic outcomes in rodent stroke models in an unbiased and efficient way.

Grouting Quality Evaluation in Post-Tensioning Tendon Ducts Using Wavelet Packet Transform and Bayes Classifier

The grouting quality of tendon ducts is very important for post-tensioning technology in order to protect the prestressing reinforcement from environmental corrosion and to make a smooth stress distribution. Unfortunately, various grouting defects occur in practice, and there is no efficient method to evaluate grouting compactness yet. In this study, a method based on wavelet packet transform (WPT) and Bayes classifier was proposed to evaluate grouting conditions using stress waves generated and received by piezoelectric transducers. Six typical grouting conditions with both partial grouting and cavity defects of different dimensions were experimentally investigated. The WPT was applied to explore the energy of received stress waves at multi-scales. After that, the Bayes classifier was employed to identify the grouting conditions, by taking the traditionally used total energy and the proposed energy vector of WPT components as input, respectively. The experimental results demonstrated that the Bayes classifier input with the energy vector could identify different grouting conditions more accurately. The proposed method has the potential to be applied at key spots of post-tensioning tendon ducts in practice.

New-Onset Alzheimer's Disease and Normal Subjects 100% Differentiated by P300

Previous work has suggested that evoked potential analysis might allow the detection of subjects with new-onset Alzheimer's disease, which would be useful clinically and personally. Here, it is described how subjects with new-onset Alzheimer's disease have been differentiated from healthy, normal subjects to 100% accuracy, based on the back-projected independent components (BICs) of the P300 peak at the electroencephalogram electrodes in the response to an oddball, auditory-evoked potential paradigm. After artifact removal, clustering, selection, and normalization processes, the BICs were classified using a neural network, a Bayes classifier, and a voting strategy. The technique is general and might be applied for presymptomatic detection and to other conditions and evoked potentials, although further validation with more subjects, preferably in multicenter studies is recommended.

Robust probabilistic classification applicable to irregularly sampled functional data

A robust probabilistic classifier for functional data is developed to predict class membership based on functional input measurements and to provide a reliable probability estimates for class membership. The method combines a Bayes classifier and semi-parametric mixed effects model with robust tuning parameter to make the method robust to outlying curves, and to improve the accuracy of the risk or uncertainty estimates, which is crucial in medical diagnostic applications. The approach applies to functional data with varying ranges and irregular sampling without making parametric assumptions on the within-curve covariance. Simulation studies evaluate the proposed method and competitors in terms of sensitivity to heavy tailed functional distributions and outlying curves. Classification performance is evaluated by both error rate and logloss, the latter of which imposes heavier penalties on highly confident errors than on less confident errors. Runtime experiments on the R implementation indicate that the proposed method scales well computationally. Illustrative applications include data from quantitative ultrasound analysis and phoneme recognition.

A Dirichlet-Multinomial Bayes Classifier for Disease Diagnosis with Microbial Compositions

By incorporating prior information on disease prevalence, Bayes classifiers have the potential to estimate disease probability better than other common machine-learning methods. Thus, it is important to develop Bayes classifiers specifically tailored for microbiome data. Our method shows higher classification accuracy than the only existing Bayesian classifier and the popular random forest method, and thus provides an alternative option for using microbial compositions for disease diagnosis.

Dysbiosis of microbial communities is associated with various human diseases, raising the possibility of using microbial compositions as biomarkers for disease diagnosis. We have developed a Bayes classifier by modeling microbial compositions with Dirichlet-multinomial distributions, which are widely used to model multicategorical count data with extra variation. The parameters of the Dirichlet-multinomial distributions are estimated from training microbiome data sets based on maximum likelihood. The posterior probability of a microbiome sample belonging to a disease or healthy category is calculated based on Bayes’ theorem, using the likelihood values computed from the estimated Dirichlet-multinomial distribution, as well as a prior probability estimated from the training microbiome data set or previously published information on disease prevalence. When tested on real-world microbiome data sets, our method, called DMBC (for Dirichlet-multinomial Bayes classifier), shows better classification accuracy than the only existing Bayesian microbiome classifier based on a Dirichlet-multinomial mixture model and the popular random forest method. The advantage of DMBC is its built-in automatic feature selection, capable of identifying a subset of microbial taxa with the best classification accuracy between different classes of samples based on cross-validation. This unique ability enables DMBC to maintain and even improve its accuracy at modeling species-level taxa. The R package for DMBC is freely available at https://github.com/qunfengdong/DMBC.

IMPORTANCE By incorporating prior information on disease prevalence, Bayes classifiers have the potential to estimate disease probability better than other common machine-learning methods. Thus, it is important to develop Bayes classifiers specifically tailored for microbiome data. Our method shows higher classification accuracy than the only existing Bayesian classifier and the popular random forest method, and thus provides an alternative option for using microbial compositions for disease diagnosis.

Bayesian pretest probability estimation for primary malignant bone tumors based on the Surveillance, Epidemiology and End Results Program (SEER) database

PURPOSE

In the diagnostic process of primary bone tumors, patient age, tumor localization and to a lesser extent sex affect the differential diagnosis. We therefore aim to develop a pretest probability calculator for primary malignant bone tumors based on population data taking these variables into account.

METHODS

We access the SEER (Surveillance, Epidemiology and End Results Program of the National Cancer Institute, 2015 release) database and analyze data of all primary malignant bone tumors diagnosed between 1973 and 2012. We record age at diagnosis, tumor localization according to the International Classification of Diseases (ICD-O-3) and sex. We take relative probability of the single tumor entity as a surrogate parameter for unadjusted pretest probability. We build a probabilistic (naïve Bayes) classifier to calculate pretest probabilities adjusted for age, tumor localization and sex.

RESULTS

We analyze data from 12,931 patients (647 chondroblastic osteosarcomas, 3659 chondrosarcomas, 1080 chordomas, 185 dedifferentiated chondrosarcomas, 2006 Ewing's sarcomas, 281 fibroblastic osteosarcomas, 129 fibrosarcomas, 291 fibrous malignant histiocytomas, 289 malignant giant cell tumors, 238 myxoid chondrosarcomas, 3730 osteosarcomas, 252 parosteal osteosarcomas, 144 telangiectatic osteosarcomas). We make our probability calculator accessible at http://ebm-radiology.com/bayesbone/index.html . We provide exhaustive tables for age and localization data. Results from tenfold cross-validation show that in 79.8 % of cases the pretest probability is correctly raised.

CONCLUSIONS

Our approach employs population data to calculate relative pretest probabilities for primary malignant bone tumors. The calculator is not diagnostic in nature. However, resulting probabilities might serve as an initial evaluation of probabilities of tumors on the differential diagnosis list.

Prognostic and diagnostic value of EEG signal coupling measures in coma

OBJECTIVE

Our aim was to assess the diagnostic and predictive value of several quantitative EEG (qEEG) analysis methods in comatose patients.

METHODS

In 79 patients, coupling between EEG signals on the left-right (inter-hemispheric) axis and on the anterior-posterior (intra-hemispheric) axis was measured with four synchronization measures: relative delta power asymmetry, cross-correlation, symbolic mutual information and transfer entropy directionality. Results were compared with etiology of coma and clinical outcome. Using cross-validation, the predictive value of measure combinations was assessed with a Bayes classifier with mixture of Gaussians.

RESULTS

Five of eight measures showed a statistically significant difference between patients grouped according to outcome; one measure revealed differences in patients grouped according to the etiology. Interestingly, a high level of synchrony between the left and right hemisphere was associated with mortality on intensive care unit, whereas higher synchrony between anterior and posterior brain regions was associated with survival. The combination with the best predictive value reached an area-under the curve of 0.875 (for patients with post anoxic encephalopathy: 0.946).

CONCLUSIONS

EEG synchronization measures can contribute to clinical assessment, and provide new approaches for understanding the pathophysiology of coma.

SIGNIFICANCE

Prognostication in coma remains a challenging task. qEEG could improve current multi-modal approaches.

Metagenome fragment classification based on multiple motif-occurrence profiles

A vast amount of metagenomic data has been obtained by extracting multiple genomes simultaneously from microbial communities, including genomes from uncultivable microbes. By analyzing these metagenomic data, novel microbes are discovered and new microbial functions are elucidated. The first step in analyzing these data is sequenced-read classification into reference genomes from which each read can be derived. The Naïve Bayes Classifier is a method for this classification. To identify the derivation of the reads, this method calculates a score based on the occurrence of a DNA sequence motif in each reference genome. However, large differences in the sizes of the reference genomes can bias the scoring of the reads. This bias might cause erroneous classification and decrease the classification accuracy. To address this issue, we have updated the Naïve Bayes Classifier method using multiple sets of occurrence profiles for each reference genome by normalizing the genome sizes, dividing each genome sequence into a set of subsequences of similar length and generating profiles for each subsequence. This multiple profile strategy improves the accuracy of the results generated by the Naïve Bayes Classifier method for simulated and Sargasso Sea datasets.