Explainable Deep Learning Approaches for Risk Screening of Periodontitis

Abstract

Several pieces of evidence have been reported regarding the association between periodontitis and systemic diseases. Despite the emphasized significance of prevention and early diagnosis of periodontitis, there is still a lack of a clinical tool for early screening of this condition. Therefore, this study aims to use explainable artificial intelligence (XAI) technology to facilitate early screening of periodontitis. This is achieved by analyzing various clinical features and providing individualized risk assessment using XAI. We used 1,012 variables for a total of 30,465 participants data from National Health and Nutrition Examination Survey (NHANES). After preprocessing, 9,632 and 5,601 participants were left for all age groups and the over 50 y age group, respectively. They were used to train deep learning and machine learning models optimized for opportunistic screening and diagnosis analysis of periodontitis based on Centers for Disease Control and Prevention/ American Academy of Pediatrics case definition. Local interpretable model-agnostic explanations (LIME) were applied to evaluate potential associated factors, including demographic, lifestyle, medical, and biochemical factors. The deep learning models showed area under the curve values of 0.858 ± 0.011 for the opportunistic screening and 0.865 ± 0.008 for the diagnostic dataset, outperforming baselines. By using LIME, we elicited important features and assessed the combined impact and interpretation of each feature on individual risk. Associated factors such as age, sex, diabetes status, tissue transglutaminase, and smoking status have emerged as crucial features that are about twice as important than other features, while arthritis, sleep disorders, high blood pressure, cholesterol levels, and overweight have also been identified as contributing factors to periodontitis. The feature contribution rankings generated with XAI offered insights that align well with clinically recognized associated factors for periodontitis. These results highlight the utility of XAI in deep learning-based associated factor analysis for detecting clinically associated factors and the assistance of XAI in developing early detection and prevention strategies for periodontitis in medical checkups.

Keywords: artificial intelligence; dental health; diagnosis; explainable artificial intelligence; opportunistic screening; risk factor.

Figure 1.

Flowchart of participant screening and data preprocessing. (A) Flowchart of participant screening. (B) Flowchart of data cleaning and preprocessing.

Figure 2.

Receiver-operating characteristic (ROC) curves for 10-fold cross-validation on each dataset of deep learning and machine learning models. (A, B) Each ROC curve includes the results from each of the 10 folds and the micro-average results with the area under the curve (AUC). It also encompasses the ±1 standard deviation area around the mean ROC curve and the range of AUC values. (C, D) Each ROC curve encompasses the results of 9 machine learning classifiers, with each outcome including the micro-average results and the AUC from 10-fold cross-validation. The curves contain the range of the ±1 standard deviation of the AUC for the mean ROC curve. Opportunistic Screening dataset: normal, mild + moderate + severe; Diagnosis dataset: normal + mild, moderate, severe.

Figure 3.

Summary plot and force plot representing the feature analysis with local interpretable model-agnostic explanations (LIME). (A, B) The resulting coefficients using LIME, which is denoted as feature contribution values are normalized with respect to features in a range of −0.5 through 0.5. Each marker corresponds to an individual participant, with different colors indicating the feature values that are normalized in the range of 0 to 1. For the sex feature, 1 and 0 represent male and female, respectively, while the rest of the categorical features with large feature values represent the positive answer. The dots placed on the right side of the graph with positive impact values have a large contribution on being classified as higher stages of periodontitis and those on the left side with negative impact values contribute to being classified as normal. (C) Individualized risk assessments from the feature contribution values of top significant features to the decision-making process of the deep learning model. The positive values denote the increase in severe periodontitis risk and the negative values the increase in the probability of being classified into the normal class. The “education level” is divided into 5 classes according to the highest level of school the participants graduated, and “lifetime smoking status” refers to the question “smoked at least 100 cigarettes in life.” Opportunistic Screening dataset: normal, mild + moderate + severe; Diagnosis dataset: normal + mild, moderate, severe.