Influenza Screening Using Patient-Generated Health Data in Post COVID-19 Pandemic.

It is very important to ensure reliable performance of deep learning model for future dataset for healthcare. This is more pronounced in the case of patient generated health data such as patient reported symptoms, which are not collected in a controlled environment. Since there has been a big difference in influenza incidence since the COVID-19 pandemic, we evaluated whether the deep learning model can maintain sufficiently robust performance against these changes. We have collected 226,655 episodes from 110,893 users since June 2020 and tested the influenza screening model, our model showed 87.02% sensitivity and 0.8670 of AUROC. The results of COVID-19 pandemic are comparable to that of before COVID-19 pandemic.

Symptom-Based COVID19 Screening Model Combined with Surveillance Information.

As the number of cases for COVID-19 continues to grow unprecedentedly, COVID-19 screening is becoming more important. In this study, we trained machine learning models from the Israel COVID-19 dataset and compared models that used surveillance indices of COVID-19 and those that did not. The AUC scores were 0.8478±0.0037 and 0.8062±0.005 with and without surveillance information, respectively, and there was significant improvement when the surveillance information was used.

Comparative Analysis of Single and Combined Antipyretics Using Patient-Generated Health Data: Retrospective Observational Study.

BACKGROUND: Fever is one of the most common symptoms in children and is the physiological response of the human immune system to external pathogens. However, effectiveness studies of single and combined antipyretic therapy are relatively few due to lack of data. In this study, we used large-scale patient-generated health data from mobile apps to compare antipyretic affects between single and combination antipyretics. OBJECTIVE: We aimed to establish combination patterns of antipyretics and compare antipyretic affects between single and combination antipyretics using large-scale patient-generated health data from mobile apps. METHODS: This study was conducted using medical records of feverish children from July 2015 to June 2017 using the Fever Coach mobile app. In total, 3,584,748 temperature records and 1,076,002 antipyretic records of 104,337 children were analyzed. Antipyretic efficacy was measured by the mean difference in the area under the temperature change curve from baseline for 6 hours, 8 hours, 10 hours, and 12 hours after antipyretic administration in children with a body temperature of ≥38.0 ℃ between single and combination groups. RESULTS: The single antipyretic and combination groups comprised 152,017 and 54,842 cases, respectively. Acetaminophen was the most commonly used single agent (60,929/152,017, 40.08%), and acetaminophen plus dexibuprofen was the most common combination (28,065/54,842, 51.17%). We observed inappropriate use, including triple combination (1205/206,859, 0.58%) and use under 38 ℃ (11,361/206,859, 5.50%). Combination antipyretic use increased with temperature; 23.82% (33,379/140,160) of cases were given a combination treatment when 38 ℃ ≤ temperature < 39 ℃, while 41.40% (1517/3664) were given a combination treatment when 40 ℃ ≤ temperature. The absolute value of the area under the curve at each hour was significantly higher in the single group than in the combination group; this trend was consistently observed, regardless of the type of antipyretics. In particular, the delta fever during the first 6 hours between the two groups showed the highest difference. The combination showed the lowest delta fever among all cases. CONCLUSIONS: Antipyretics combination patterns were analyzed using large-scale data. Approximately 75% of febrile cases used single antipyretics, mostly acetaminophen, but combination usage became more frequent as temperature increased. However, combination antipyretics did not show definite advantages over single antipyretics in defervescence, regardless of the combination. Single antipyretics are effective in reducing fever and relieving discomfort in febrile children.

Influenza Screening via Deep Learning Using a Combination of Epidemiological and Patient-Generated Health Data: Development and Validation Study.

BACKGROUND: Screening for influenza in primary care is challenging due to the low sensitivity of rapid antigen tests and the lack of proper screening tests. OBJECTIVE: The aim of this study was to develop a machine learning-based screening tool using patient-generated health data (PGHD) obtained from a mobile health (mHealth) app. METHODS: We trained a deep learning model based on a gated recurrent unit to screen influenza using PGHD, including each patient's fever pattern and drug administration records. We used meteorological data and app-based surveillance of the weekly number of patients with influenza. We defined a single episode as the set of consecutive days, including the day the user was diagnosed with influenza or another disease. Any record a user entered 24 hours after his or her last record was considered to be the start of a new episode. Each episode contained data on the user's age, gender, weight, and at least one body temperature record. The total number of episodes was 6657. Of these, there were 3326 episodes within which influenza was diagnosed. We divided these episodes into 80% training sets (2664/3330) and 20% test sets (666/3330). A 5-fold cross-validation was used on the training set. RESULTS: We achieved reliable performance with an accuracy of 82%, a sensitivity of 84%, and a specificity of 80% in the test set. After the effect of each input variable was evaluated, app-based surveillance was observed to be the most influential variable. The correlation between the duration of input data and performance was not statistically significant (P=.09). CONCLUSIONS: These findings suggest that PGHD from an mHealth app could be a complementary tool for influenza screening. In addition, PGHD, along with traditional clinical data, could be used to improve health conditions.

Correction: Postvaccination Fever Response Rates in Children Derived Using the Fever Coach Mobile App: A Retrospective Observational Study.

[This corrects the article DOI: 10.2196/12223.].

Deep Convolutional Neural Network-based Software Improves Radiologist Detection of Malignant Lung Nodules on Chest Radiographs.

Background Multicenter studies are required to validate the added benefit of using deep convolutional neural network (DCNN) software for detecting malignant pulmonary nodules on chest radiographs. Purpose To compare the performance of radiologists in detecting malignant pulmonary nodules on chest radiographs when assisted by deep learning-based DCNN software with that of radiologists or DCNN software alone in a multicenter setting. Materials and Methods Investigators at four medical centers retrospectively identified 600 lung cancer-containing chest radiographs and 200 normal chest radiographs. Each radiograph with a lung cancer had at least one malignant nodule confirmed by CT and pathologic examination. Twelve radiologists from the four centers independently analyzed the chest radiographs and marked regions of interest. Commercially available deep learning-based computer-aided detection software separately trained, tested, and validated with 19 330 radiographs was used to find suspicious nodules. The radiologists then reviewed the images with the assistance of DCNN software. The sensitivity and number of false-positive findings per image of DCNN software, radiologists alone, and radiologists with the use of DCNN software were analyzed by using logistic regression and Poisson regression. Results The average sensitivity of radiologists improved (from 65.1% [1375 of 2112; 95% confidence interval {CI}: 62.0%, 68.1%] to 70.3% [1484 of 2112; 95% CI: 67.2%, 73.1%], P < .001) and the number of false-positive findings per radiograph declined (from 0.2 [488 of 2400; 95% CI: 0.18, 0.22] to 0.18 [422 of 2400; 95% CI: 0.16, 0.2], P < .001) when the radiologists re-reviewed radiographs with the DCNN software. For the 12 radiologists in this study, 104 of 2400 radiographs were positively changed (from false-negative to true-positive or from false-positive to true-negative) using the DCNN, while 56 of 2400 radiographs were changed negatively. Conclusion Radiologists had better performance with deep convolutional network software for the detection of malignant pulmonary nodules on chest radiographs than without. © RSNA, 2019 Online supplemental material is available for this article. See also the editorial by Jacobson in this issue.

The Fever Coach Mobile App for Participatory Influenza Surveillance in Children: Usability Study.

BACKGROUND: Effective surveillance of influenza requires a broad network of health care providers actively reporting cases of influenza-like illnesses and positive laboratory results. Not only is this traditional surveillance system costly to establish and maintain but there is also a time lag between a change in influenza activity and its detection. A new surveillance system that is both reliable and timely will help public health officials to effectively control an epidemic and mitigate the burden of the disease. OBJECTIVE: This study aimed to evaluate the use of parent-reported data of febrile illnesses in children submitted through the Fever Coach app in real-time surveillance of influenza activities. METHODS: Fever Coach is a mobile app designed to help parents and caregivers manage fever in young children, currently mainly serviced in South Korea. The app analyzes data entered by a caregiver and provides tailored information for care of the child based on the child's age, sex, body weight, body temperature, and accompanying symptoms. Using the data submitted to the app during the 2016-2017 influenza season, we built a regression model that monitors influenza incidence for the 2017-2018 season and validated the model by comparing the predictions with the public influenza surveillance data from the Korea Centers for Disease Control and Prevention (KCDC). RESULTS: During the 2-year study period, 70,203 diagnosis data, including 7702 influenza reports, were submitted. There was a significant correlation between the influenza activity predicted by Fever Coach and that reported by KCDC (Spearman ρ=0.878; P<.001). Using this model, the influenza epidemic in the 2017-2018 season was detected 10 days before the epidemic alert announced by KCDC. CONCLUSIONS: The Fever Coach app successfully collected data from 7.73% (207,699/2,686,580) of the target population by providing care instruction for febrile children. These data were used to develop a model that accurately estimated influenza activity measured by the central government agency using reports from sentinel facilities in the national surveillance network.

Machine Friendly Machine Learning: Interpretation of Computed Tomography Without Image Reconstruction.

Recent advancements in deep learning for automated image processing and classification have accelerated many new applications for medical image analysis. However, most deep learning algorithms have been developed using reconstructed, human-interpretable medical images. While image reconstruction from raw sensor data is required for the creation of medical images, the reconstruction process only uses a partial representation of all the data acquired. Here, we report the development of a system to directly process raw computed tomography (CT) data in sinogram-space, bypassing the intermediary step of image reconstruction. Two classification tasks were evaluated for their feasibility of sinogram-space machine learning: body region identification and intracranial hemorrhage (ICH) detection. Our proposed SinoNet, a convolutional neural network optimized for interpreting sinograms, performed favorably compared to conventional reconstructed image-space-based systems for both tasks, regardless of scanning geometries in terms of projections or detectors. Further, SinoNet performed significantly better when using sparsely sampled sinograms than conventional networks operating in image-space. As a result, sinogram-space algorithms could be used in field settings for triage (presence of ICH), especially where low radiation dose is desired. These findings also demonstrate another strength of deep learning where it can analyze and interpret sinograms that are virtually impossible for human experts.

An explainable deep-learning algorithm for the detection of acute intracranial haemorrhage from small datasets.

Owing to improvements in image recognition via deep learning, machine-learning algorithms could eventually be applied to automated medical diagnoses that can guide clinical decision-making. However, these algorithms remain a 'black box' in terms of how they generate the predictions from the input data. Also, high-performance deep learning requires large, high-quality training datasets. Here, we report the development of an understandable deep-learning system that detects acute intracranial haemorrhage (ICH) and classifies five ICH subtypes from unenhanced head computed-tomography scans. By using a dataset of only 904 cases for algorithm training, the system achieved a performance similar to that of expert radiologists in two independent test datasets containing 200 cases (sensitivity of 98% and specificity of 95%) and 196 cases (sensitivity of 92% and specificity of 95%). The system includes an attention map and a prediction basis retrieved from training data to enhance explainability, and an iterative process that mimics the workflow of radiologists. Our approach to algorithm development can facilitate the development of deep-learning systems for a variety of clinical applications and accelerate their adoption into clinical practice.

Postvaccination Fever Response Rates in Children Derived Using the Fever Coach Mobile App: A Retrospective Observational Study.

BACKGROUND: Postvaccination fever is a mild adverse event that naturally improves without complications, but is highly prevalent and can be accompanied by febrile convulsions in some cases. These adverse effects may cause parents to delay or avoid vaccinating their children. OBJECTIVE: This study aimed to identify postvaccination fever patterns and the ability of antipyretics to affect changes in these patterns from data collected from a mobile app named Fever Coach. METHODS: Data provided by parents of feverish children derived from a mobile app, Fever Coach, were used to identify postvaccination fever patterns according to vaccinations and the use of antipyretic drugs. We selected single vaccination records that contained five or more body temperature readings performed within 48 hours of vaccination, and we analyzed postvaccination fever onset, offset, duration, and maximum body temperature. Through observing the postvaccination fever response to vaccination, we identified the effects of antipyretic drugs on postvaccination fever onset, offset, and duration times; the extent of fever; and the rate of decline. We also performed logistic regression analysis to determine demographic variables (age, weight, and sex) involved in relatively high fevers (body temperature ≥39°C). RESULTS: The total number of Fever Coach users was 25,037, with 3834 users having entered single vaccination records, including 4448 vaccinations and 55,783 body temperature records. Most records were obtained from children receiving the following vaccinations: pneumococcus (n=2069); Japanese encephalitis (n=911); influenza (n=669); diphtheria, tetanus, and pertussis (n=403); and hepatitis A (n=252). According to the 4448 vaccination records, 3427 (77.05%) children had taken antipyretic drugs, and 3238 (89.15%) children took antibiotics at body temperatures above 38°C. The number of children taking antipyretics at a body temperature of 38°C was more than four times that of those taking antipyretics at 37.9°C (307 vs 67 cases). The number of instances in which this temperature threshold was reached was more than four times greater than the number when the temperature was 37.9°C. A comparative analysis of antipyretic and nonantipyretic cases showed there was no difference in onset time; however, offset and duration times were significantly shorter in nonantipyretic cases than in antipyretic cases (P<.001). In nonantipyretic cases, offset times and duration times were 9.9 and 10.1 hours shorter, respectively, than in antipyretic cases. Body temperatures also decreased faster in nonantipyretic cases. Influenza vaccine-associated fevers lasted relatively longer, whereas pneumococcus vaccine-associated fevers were relatively short-lived. CONCLUSIONS: These findings suggest that postvaccination fever has its own fever pattern, which is dependent on vaccine type and the presence of antipyretic drugs, and that postvaccination temperature monitoring may ease fever phobia and reduce the unnecessary use of antipyretics in medical care.

Beyond Human Perception: Sexual Dimorphism in Hand and Wrist Radiographs Is Discernible by a Deep Learning Model.

Despite the well-established impact of sex and sex hormones on bone structure and density, there has been limited description of sexual dimorphism in the hand and wrist in the literature. We developed a deep convolutional neural network (CNN) model to predict sex based on hand radiographs of children and adults aged between 5 and 70 years. Of the 1531 radiographs tested, the algorithm predicted sex correctly in 95.9% (κ = 0.92) of the cases. Two human radiologists achieved 58% (κ = 0.15) and 46% (κ = - 0.07) accuracy. The class activation maps (CAM) showed that the model mostly focused on the 2nd and 3rd metacarpal base or thumb sesamoid in women, and distal radioulnar joint, distal radial physis and epiphysis, or 3rd metacarpophalangeal joint in men. The radiologists reviewed 70 cases (35 females and 35 males) labeled with sex along with heat maps generated by CAM, but they could not find any patterns that distinguish the two sexes. A small sample of patients (n = 44) with sexual developmental disorders or transgender identity was selected for a preliminary exploration of application of the model. The model prediction agreed with phenotypic sex in only 77.8% (κ = 0.54) of these cases. To the best of our knowledge, this is the first study that demonstrated a machine learning model to perform a task in which human experts could not fulfill.