Previously submitted to: JMIR Public Health and Surveillance (no longer under consideration since May 12, 2021)
Date Submitted: Sep 20, 2019
Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.
The Convolutional Neural Network Combined with the HT Person Fit Statistic to Develop an APP for Detecting Dengue Fever in Children
ABSTRACT
Background:
Dengue fever (DF) is a significant public health issue in Asia. However, DF is extremely hard to detect using the traditional dichotomous (i.e., absent vs present) evaluation with symptoms. The convolution neural network(CNN), a famous deep learning method, can improve the prediction accuracy due to a larger number of parameters used in the model. Whether the HT person fit statistic cab be combined with the CNN to increase the prediction accuracy and then to develop an APP for detecting DF in Children is remains unknown.
Objective:
The aim of this study is to build a model for automatic detection and classification of DF with symptoms for helping patients, family members, or clinicians to identify DF at an earlier stage.
Methods:
We extracted 19 featured variables from DF-related symptoms in 177 pediatric patients (69 diagnosed with DF) using CNN to predict DF risk. Two sets of characteristics(19 symptoms and other 5 variables including person mean, standard deviation, and two HT related statistics) were compared in predicting DF accuracies. Data were separated into training and testing sets-the former were used to predict the latter. We calculated the sensitivity, the specificity and the receiver operating characteristic (ROC) curve (AUC) across studies in comparison.
Results:
We observed that (1) the 24-item model yield a higher accurate rate (=0.95) with AUC(=0.96, 95% CI 0.93-0.99) higher than the 19-item model with the accuracy(=0.92) and AUC(=0.90, 95% CI 0.86-0.94) based on the 177-case training set; (2) the sensitivities are not higher than the specificities (90% in ten scenarios) for predicting the DF ; (3) the sensitivities and the specificities in the 24-item model are not entirely higher than those in the 19-item model.
Conclusions:
The 24-item model earns higher accurate rates than the 19-item model across all 10 study scenario. An APP can be developed for helping patients, family members, or clinicians discriminate DF from other febrile illnesses at an early stage. Clinical Trial: Not available
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.