scholarly journals Using self-supervised feature learning to improve the use of pulse oximeter signals to predict paediatric hospitalization

2021 ◽  
Vol 6 ◽  
pp. 248
Author(s):  
Paul Mwaniki ◽  
Timothy Kamanu ◽  
Samuel Akech ◽  
Dustin Dunsmuir ◽  
J. Mark Ansermino ◽  
...  
Keyword(s):  
Machine Learning ◽  
Logistic Regression ◽  
Deep Learning ◽  
Pulse Oximeter ◽  
Regression Models ◽  
Data Sets ◽  
Learning Models ◽  
Data Set ◽  
Logistic Regression Models ◽  
End To End

Background: The success of many machine learning applications depends on knowledge about the relationship between the input data and the task of interest (output), hindering the application of machine learning to novel tasks. End-to-end deep learning, which does not require intermediate feature engineering, has been recommended to overcome this challenge but end-to-end deep learning models require large labelled training data sets often unavailable in many medical applications. In this study, we trained machine learning models to predict paediatric hospitalization given raw photoplethysmography (PPG) signals obtained from a pulse oximeter. We trained self-supervised learning (SSL) for automatic feature extraction from PPG signals and assessed the utility of SSL in initializing end-to-end deep learning models trained on a small labelled data set with the aim of predicting paediatric hospitalization.Methods: We compared logistic regression models fitted using features extracted using SSL with end-to-end deep learning models initialized either randomly or using weights from the SSL model. We also compared the performance of SSL models trained on labelled data alone (n=1,031) with SSL trained using both labelled and unlabelled signals (n=7,578). Results: The SSL model trained on both labelled and unlabelled PPG signals produced features that were more predictive of hospitalization compared to the SSL model trained on labelled PPG only (AUC of logistic regression model: 0.78 vs 0.74). The end-to-end deep learning model had an AUC of 0.80 when initialized using the SSL model trained on all PPG signals, 0.77 when initialized using SSL trained on labelled data only, and 0.73 when initialized randomly. Conclusions: This study shows that SSL can improve the classification of PPG signals by either extracting features required by logistic regression models or initializing end-to-end deep learning models. Furthermore, SSL can leverage larger unlabelled data sets to improve performance of models fitted using small labelled data sets.

scholarly journals Training confounder-free deep learning models for medical applications

2020 ◽  
Vol 11 (1) ◽  
Author(s):  
Qingyu Zhao ◽  
Ehsan Adeli ◽  
Kilian M. Pohl
Keyword(s):  
Machine Learning ◽  
Deep Learning ◽  
Bone Age ◽  
Magnetic Resonance Images ◽  
Data Sets ◽  
Large Set ◽  
Learning Models ◽  
End To End ◽  
The Impact ◽  
The Relationship

AbstractThe presence of confounding effects (or biases) is one of the most critical challenges in using deep learning to advance discovery in medical imaging studies. Confounders affect the relationship between input data (e.g., brain MRIs) and output variables (e.g., diagnosis). Improper modeling of those relationships often results in spurious and biased associations. Traditional machine learning and statistical models minimize the impact of confounders by, for example, matching data sets, stratifying data, or residualizing imaging measurements. Alternative strategies are needed for state-of-the-art deep learning models that use end-to-end training to automatically extract informative features from large set of images. In this article, we introduce an end-to-end approach for deriving features invariant to confounding factors while accounting for intrinsic correlations between the confounder(s) and prediction outcome. The method does so by exploiting concepts from traditional statistical methods and recent fair machine learning schemes. We evaluate the method on predicting the diagnosis of HIV solely from Magnetic Resonance Images (MRIs), identifying morphological sex differences in adolescence from those of the National Consortium on Alcohol and Neurodevelopment in Adolescence (NCANDA), and determining the bone age from X-ray images of children. The results show that our method can accurately predict while reducing biases associated with confounders. The code is available at https://github.com/qingyuzhao/br-net.

Machine Learning Diagnostic Modeling for Classifying Fibromyalgia Using B-mode Ultrasound Images

2020 ◽  
Vol 42 (3) ◽  
pp. 135-147 ◽  
Author(s):  
Michael Behr ◽  
Saba Saiel ◽  
Valerie Evans ◽  
Dinesh Kumbhare
Keyword(s):  
Machine Learning ◽  
Logistic Regression ◽  
Regression Models ◽  
Trapezius Muscle ◽  
Image Texture ◽  
Learning Models ◽  
Test Set ◽  
Logistic Regression Models ◽  
Performance Accuracy ◽  
Machine Learning Models

Fibromyalgia (FM) diagnosis remains a challenge for clinicians due to a lack of objective diagnostic tools. One proposed solution is the use of quantitative ultrasound (US) techniques, such as image texture analysis, which has demonstrated discriminatory capabilities with other chronic pain conditions. From this, we propose the use of image texture variables to construct and compare two machine learning models (support vector machine [SVM] and logistic regression) for differentiating between the trapezius muscle in healthy and FM patients. US videos of the right and left trapezius muscle were acquired from healthy ( n = 51) participants and those with FM ( n = 57). The videos were converted into 64,800 skeletal muscle regions of interest (ROIs) using MATLAB. The ROIs were filtered by an algorithm using the complex wavelet structural similarity index (CW-SSIM), which removed ROIs that were similar. Thirty-one texture variables were extracted from the ROIs, which were then used in nested cross-validation to construct SVM and elastic net regularized logistic regression models. The generalized performance accuracy of both models was estimated and confirmed with a final validation on a holdout test set. The predicted generalized performance accuracy of the SVM and logistic regression models was computed to be 83.9 ± 2.6% and 65.8 ± 1.7%, respectively. The models achieved accuracies of 84.1%, and 66.0% on the final holdout test set, validating performance estimates. Although both machine learning models differentiate between healthy trapezius muscle and that of patients with FM, only the SVM model demonstrated clinically relevant performance levels.

scholarly journals High performance logistic regression for privacy-preserving genome analysis

2021 ◽  
Vol 14 (1) ◽  
Author(s):  
Martine De Cock ◽  
Rafael Dowsley ◽  
Anderson C. A. Nascimento ◽  
Davis Railsback ◽  
Jianwei Shen ◽  
...  
Keyword(s):  
Machine Learning ◽  
Logistic Regression ◽  
Genome Analysis ◽  
Local Area Network ◽  
Local Area ◽  
Activation Function ◽  
Area Network ◽  
Learning Models ◽  
Data Set ◽  
Machine Learning Models

Abstract Background In biomedical applications, valuable data is often split between owners who cannot openly share the data because of privacy regulations and concerns. Training machine learning models on the joint data without violating privacy is a major technology challenge that can be addressed by combining techniques from machine learning and cryptography. When collaboratively training machine learning models with the cryptographic technique named secure multi-party computation, the price paid for keeping the data of the owners private is an increase in computational cost and runtime. A careful choice of machine learning techniques, algorithmic and implementation optimizations are a necessity to enable practical secure machine learning over distributed data sets. Such optimizations can be tailored to the kind of data and Machine Learning problem at hand. Methods Our setup involves secure two-party computation protocols, along with a trusted initializer that distributes correlated randomness to the two computing parties. We use a gradient descent based algorithm for training a logistic regression like model with a clipped ReLu activation function, and we break down the algorithm into corresponding cryptographic protocols. Our main contributions are a new protocol for computing the activation function that requires neither secure comparison protocols nor Yao’s garbled circuits, and a series of cryptographic engineering optimizations to improve the performance. Results For our largest gene expression data set, we train a model that requires over 7 billion secure multiplications; the training completes in about 26.90 s in a local area network. The implementation in this work is a further optimized version of the implementation with which we won first place in Track 4 of the iDASH 2019 secure genome analysis competition. Conclusions In this paper, we present a secure logistic regression training protocol and its implementation, with a new subprotocol to securely compute the activation function. To the best of our knowledge, we present the fastest existing secure multi-party computation implementation for training logistic regression models on high dimensional genome data distributed across a local area network.

45 Application of Machine Learning Models to Thermal Burn Patient Outcome Predictions in the Aftermath of a Nuclear Event

2021 ◽  
Vol 42 (Supplement_1) ◽  
pp. S33-S34
Author(s):  
Morgan A Taylor ◽  
Randy D Kearns ◽  
Jeffrey E Carter ◽  
Mark H Ebell ◽  
Curt A Harris
Keyword(s):  
Machine Learning ◽  
Logistic Regression ◽  
Length Of Stay ◽  
Regression Models ◽  
Large Scale ◽  
Prediction Models ◽  
Burn Patients ◽  
Thermal Burn ◽  
Logistic Regression Models ◽  
Burn Patient

Abstract Introduction A nuclear disaster would generate an unprecedented volume of thermal burn patients from the explosion and subsequent mass fires (Figure 1). Prediction models characterizing outcomes for these patients may better equip healthcare providers and other responders to manage large scale nuclear events. Logistic regression models have traditionally been employed to develop prediction scores for mortality of all burn patients. However, other healthcare disciplines have increasingly transitioned to machine learning (ML) models, which are automatically generated and continually improved, potentially increasing predictive accuracy. Preliminary research suggests ML models can predict burn patient mortality more accurately than commonly used prediction scores. The purpose of this study is to examine the efficacy of various ML methods in assessing thermal burn patient mortality and length of stay in burn centers. Methods This retrospective study identified patients with fire/flame burn etiologies in the National Burn Repository between the years 2009 – 2018. Patients were randomly partitioned into a 67%/33% split for training and validation. A random forest model (RF) and an artificial neural network (ANN) were then constructed for each outcome, mortality and length of stay. These models were then compared to logistic regression models and previously developed prediction tools with similar outcomes using a combination of classification and regression metrics. Results During the study period, 82,404 burn patients with a thermal etiology were identified in the analysis. The ANN models will likely tend to overfit the data, which can be resolved by ending the model training early or adding additional regularization parameters. Further exploration of the advantages and limitations of these models is forthcoming as metric analyses become available. Conclusions In this proof-of-concept study, we anticipate that at least one ML model will predict the targeted outcomes of thermal burn patient mortality and length of stay as judged by the fidelity with which it matches the logistic regression analysis. These advancements can then help disaster preparedness programs consider resource limitations during catastrophic incidents resulting in burn injuries.

Association Between Opioid Use and Diabetes in Patients With Ankle Fracture Repair

2021 ◽  
pp. 107110072110581
Author(s):  
Wenye Song ◽  
Naohiro Shibuya ◽  
Daniel C. Jupiter
Keyword(s):  
Diabetes Mellitus ◽  
Logistic Regression ◽  
Ankle Fracture ◽  
Regression Models ◽  
Fracture Repair ◽  
Opioid Use ◽  
Data Set ◽  
Opioid Prescription ◽  
Logistic Regression Models ◽  
Patients With Diabetes

Background: Ankle fractures in patients with diabetes mellitus have long been recognized as a challenge to practicing clinicians. Ankle fracture patients with diabetes may experience prolonged healing, higher risk of hardware failure, an increased risk of wound dehiscence and infection, and higher pain scores pre- and postoperatively, compared to patients without diabetes. However, the duration of opioid use among this patient cohort has not been previously evaluated. The purpose of this study is to retrospectively compare the time span of opioid utilization between ankle fracture patients with and without diabetes mellitus. Methods: We conducted a retrospective cohort study using our institution’s TriNetX database. A total of 640 ankle fracture patients were included in the analysis, of whom 73 had diabetes. All dates of opioid use for each patient were extracted from the data set, including the first and last date of opioid prescription. Descriptive analysis and logistic regression models were employed to explore the differences in opioid use between patients with and without diabetes after ankle fracture repair. A 2-tailed P value of .05 was set as the threshold for statistical significance. Results: Logistic regression models revealed that patients with diabetes are less likely to stop using opioids within 90 days, or within 180 days, after repair compared to patients without diabetes. Female sex, neuropathy, and prefracture opioid use are also associated with prolonged opioid use after ankle fracture repair. Conclusion: In our study cohort, ankle fracture patients with diabetes were more likely to require prolonged opioid use after fracture repair. Level of Evidence: Level III, prognostic.

scholarly journals Predicting Fusarium Head Blight Epidemics with Boosted Regression Trees

2014 ◽  
Vol 104 (7) ◽  
pp. 702-714 ◽  
Author(s):  
D. A. Shah ◽  
E. D. De Wolf ◽  
P. A. Paul ◽  
L. V. Madden
Keyword(s):  
Logistic Regression ◽  
Relative Humidity ◽  
Fusarium Head Blight ◽  
Regression Models ◽  
Boosted Regression Trees ◽  
Misclassification Rate ◽  
Head Blight ◽  
Data Set ◽  
Logistic Regression Models ◽  
Daily Relative Humidity

Predicting major Fusarium head blight (FHB) epidemics allows for the judicious use of fungicides in suppressing disease development. Our objectives were to investigate the utility of boosted regression trees (BRTs) for predictive modeling of FHB epidemics in the United States, and to compare the predictive performances of the BRT models with those of logistic regression models we had developed previously. The data included 527 FHB observations from 15 states over 26 years. BRTs were fit to a training data set of 369 FHB observations, in which FHB epidemics were classified as either major (severity ≥ 10%) or non-major (severity < 10%), linked to a predictor matrix consisting of 350 weather-based variables and categorical variables for wheat type (spring or winter), presence or absence of corn residue, and cultivar resistance. Predictive performance was estimated on a test (holdout) data set consisting of the remaining 158 observations. BRTs had a misclassification rate of 0.23 on the test data, which was 31% lower than the average misclassification rate over 15 logistic regression models we had presented earlier. The strongest predictors were generally one of mean daily relative humidity, mean daily temperature, and the number of hours in which the temperature was between 9 and 30°C and relative humidity ≥ 90% simultaneously. Moreover, the predicted risk of major epidemics increased substantially when mean daily relative humidity rose above 70%, which is a lower threshold than previously modeled for most plant pathosystems. BRTs led to novel insights into the weather–epidemic relationship.

scholarly journals The influence of training sample size on the accuracy of deep learning models for the prediction of soil properties with near-infrared spectroscopy data

SOIL ◽  
2020 ◽  
Vol 6 (2) ◽  
pp. 565-578
Author(s):  
Wartini Ng ◽  
Budiman Minasny ◽  
Wanderson de Sousa Mendes ◽  
José Alexandre Melo Demattê
Keyword(s):  
Machine Learning ◽  
Deep Learning ◽  
Soil Properties ◽  
Sample Size ◽  
Training Sample ◽  
Calibration Data ◽  
Learning Models ◽  
Data Set ◽  
Calibration Data Set ◽  
Machine Learning Models

Abstract. The number of samples used in the calibration data set affects the quality of the generated predictive models using visible, near and shortwave infrared (VIS–NIR–SWIR) spectroscopy for soil attributes. Recently, the convolutional neural network (CNN) has been regarded as a highly accurate model for predicting soil properties on a large database. However, it has not yet been ascertained how large the sample size should be for CNN model to be effective. This paper investigates the effect of the training sample size on the accuracy of deep learning and machine learning models. It aims at providing an estimate of how many calibration samples are needed to improve the model performance of soil properties predictions with CNN as compared to conventional machine learning models. In addition, this paper also looks at a way to interpret the CNN models, which are commonly labelled as a black box. It is hypothesised that the performance of machine learning models will increase with an increasing number of training samples, but it will plateau when it reaches a certain number, while the performance of CNN will keep improving. The performances of two machine learning models (partial least squares regression – PLSR; Cubist) are compared against the CNN model. A VIS–NIR–SWIR spectra library from Brazil, containing 4251 unique sites with averages of two to three samples per depth (a total of 12 044 samples), was divided into calibration (3188 sites) and validation (1063 sites) sets. A subset of the calibration data set was then created to represent a smaller calibration data set ranging from 125, 300, 500, 1000, 1500, 2000, 2500 and 2700 unique sites, which is equivalent to a sample size of approximately 350, 840, 1400, 2800, 4200, 5600, 7000 and 7650. All three models (PLSR, Cubist and CNN) were generated for each sample size of the unique sites for the prediction of five different soil properties, i.e. cation exchange capacity, organic carbon, sand, silt and clay content. These calibration subset sampling processes and modelling were repeated 10 times to provide a better representation of the model performances. Learning curves showed that the accuracy increased with an increasing number of training samples. At a lower number of samples (< 1000), PLSR and Cubist performed better than CNN. The performance of CNN outweighed the PLSR and Cubist model at a sample size of 1500 and 1800, respectively. It can be recommended that deep learning is most efficient for spectra modelling for sample sizes above 2000. The accuracy of the PLSR and Cubist model seems to reach a plateau above sample sizes of 4200 and 5000, respectively, while the accuracy of CNN has not plateaued. A sensitivity analysis of the CNN model demonstrated its ability to determine important wavelengths region that affected the predictions of various soil attributes.

scholarly journals A Tweet Sentiment Classification Approach Using a Hybrid Stacked Ensemble Technique

Information ◽  
2021 ◽  
Vol 12 (9) ◽  
pp. 374
Author(s):  
Babacar Gaye ◽  
Dezheng Zhang ◽  
Aziguli Wulamu
Keyword(s):  
Machine Learning ◽  
Logistic Regression ◽  
Deep Learning ◽  
Sentiment Analysis ◽  
Language Processing ◽  
Short Term Memory ◽  
State Of The Art ◽  
Accuracy Score ◽  
Learning Models ◽  
Proposed Model

With the extensive availability of social media platforms, Twitter has become a significant tool for the acquisition of peoples’ views, opinions, attitudes, and emotions towards certain entities. Within this frame of reference, sentiment analysis of tweets has become one of the most fascinating research areas in the field of natural language processing. A variety of techniques have been devised for sentiment analysis, but there is still room for improvement where the accuracy and efficacy of the system are concerned. This study proposes a novel approach that exploits the advantages of the lexical dictionary, machine learning, and deep learning classifiers. We classified the tweets based on the sentiments extracted by TextBlob using a stacked ensemble of three long short-term memory (LSTM) as base classifiers and logistic regression (LR) as a meta classifier. The proposed model proved to be effective and time-saving since it does not require feature extraction, as LSTM extracts features without any human intervention. We also compared our proposed approach with conventional machine learning models such as logistic regression, AdaBoost, and random forest. We also included state-of-the-art deep learning models in comparison with the proposed model. Experiments were conducted on the sentiment140 dataset and were evaluated in terms of accuracy, precision, recall, and F1 Score. Empirical results showed that our proposed approach manifested state-of-the-art results by achieving an accuracy score of 99%.

Sign in / Sign up

Export Citation Format

Share Document