Identifying Disease of Interest With Deep Learning Using Diagnosis Code

Background@#Autoencoder (AE) is one of the deep learning techniques that uses an artificial neural network to reconstruct its input data in the output layer. We constructed a novel supervised AE model and tested its performance in the prediction of a co-existence of the disease of interest only using diagnostic codes. @*Methods@#Diagnostic codes of one million randomly sampled patients listed in the Korean National Health Information Database in 2019 were used to train, validate, and test the prediction model. The first used AE solely for a feature engineering tool for an input of a classifier. Supervised Multi-Layer Perceptron (sMLP) was added to train a classifier to predict a binary level with latent representation as an input (AE + sMLP). The second model simultaneously updated the parameters in the AE and the connected MLP classifier during the learning process (End-to-End Supervised AE [EEsAE]). We tested the performances of these two models against baseline models, eXtreme Gradient Boosting (XGB) and naïve Bayes, in the prediction of co-existing gastric cancer diagnosis. @*Results@#The proposed EEsAE model yielded the highest F1-score and highest area under the curve (0.86). The EEsAE and AE + sMLP gave the highest recalls. XGB yielded the highest precision. Ablation study revealed that iron deficiency anemia, gastroesophageal reflux disease, essential hypertension, gastric ulcers, benign prostate hyperplasia, and shoulder lesion were the top 6 most influential diagnoses on performance. @*Conclusion@#A novel EEsAE model showed promising performance in the prediction of a disease of interest.

Relationship among the use of food-related content, dietary behaviors, and dietary self-efficacy of high school students in Seoul and Gyeonggi areas / 한국영양학회지

PURPOSE: This study examined the relationship among the use of food-related content (FRC), dietary behaviors, and dietary self-efficacy to demonstrate the need for nutrition education to help adolescents build healthy eating habits and provide evidence for developing nutrition education programs for adolescents. METHODS: Three hundred and eighty-one high school students in Seoul and Gyeonggi areas participated in the study. The subjects were divided into three groups (low, medium, and high) according to the level of use of the FRC, and their general characteristics, dietary behaviors, and dietary self-efficacy were analyzed. Correlation analysis was performed between FRC usage, dietary behaviors, and dietary self-efficacy, and the mediating effects of dietary self-efficacy on the relationship between the level of the use of FRC and dietary behaviors were estimated. RESULTS: A higher level of FRC usage was associated with an increased daily cost of eating out and snacking, but no difference was observed in the BMI range. The subjects in a group with a high level of FRC usage ate convenience store or instant foods instead of homemade meals (p = 0.033), had a late-night meal or snack (p = 0.024), and turned to emotional eating under stress (p < 0.001) more than those in the low level group. In addition, the high level group checked the nutrition facts label more carefully when purchasing processed foods (p = 0.016) and exercised at least 30 minutes daily, not considering physical education classes (p = 0.057). The higher level of FRC use, the lower the dietary self-efficacy, whereby the subscales ‘environmental stimulus control efficacy’ and ‘affective factor control efficacy’ showed complete mediating effects. CONCLUSION: Given that FRC has been increased recently, adolescents are in need of support to help them control and enhance their dietary self-efficacy as well as develop healthy dietary behaviors through proper nutrition education programs.