Background: The excessive and frequent intake of refined sugar leads to caries. However, the relationship between the amount of sugar intake and the risk of caries is not always consistent. Oral microbial profile and function may impact the link between them. This study aims to identify the plaque microbiota characteristics of caries subjects with low (CL) and high (CH) sugar consumption, and of caries-free subjects with low (FL) and high sugar (FH) consumption. Methods: A total of 40 adolescents were enrolled in the study, and supragingival plaque samples were collected and subjected to metagenomic analyses. The caries status, sugar consumption, and oral-health behaviors of the subjects were recorded. Results: The results indicate that the CL group showed a higher abundance of several cariogenic microorganisms Lactobacillus, A. gerencseriae, A. dentails, S. mutans, C. albicans, S. wiggsiae and P. acidifaciens. C. gingivalis, and P. gingivalis, which were enriched in the FH group. In terms of gene function, the phosphotransferase sugar uptake system, phosphotransferase system, and several two-component responses–regulator pairs were enriched in the CL group. Conclusion: Overall, our data suggest the existence of an increased cariogenic microbial community and sugar catabolism potential in the CL group, and a healthy microbial community in the FH group, which had self-stabilizing functional potential.

Parents may influence children’s intake of sweets and sugar-sweetened beverages through their feeding styles. This study prospectively assessed the association between parental feeding styles and caries risk in preschool children. In November and December 2021, we recruited 1181 preschool children aged 3–4 years in Guangzhou, China, and then conducted a follow-up visit after 15 months. The children were examined for dental caries, and their parents answered the Chinese version of the Parental Feeding Style Questionnaire. Data were analyzed using logistic regression analysis. At baseline, ‘control over eating’ was found to be negatively associated with early childhood caries (OR = 0.60, 95%CI = 0.44, 0.83). After 15 months, no significant association between baseline parental feeding styles and the risk of a new carious lesion was found. When considering changes in parental feeding styles between the baseline and the follow-up visit, we found children with increasing scores on ‘instrumental feeding’ during the 15 months had a higher risk of a new carious lesion (OR = 1.91, 95%CI = 1.18, 3.12). However, children with always-intermediate scores (OR = 0.51, 95%CI = 0.30, 0.86), always-high scores (OR = 0.42, 95%CI = 0.23, 0.77), or increasing scores (OR = 0.50, 95%CI = 0.31, 0.83) on ‘control over eating’ had a lower risk of a new carious lesion. Our study highlighted the influence of parental feeding styles on caries risk in preschool children.

To develop a multimodal deep learning model for dental caries screening in children by integrating intraoral photographs and questionnaire-based data, and to compare its performance against that of a unimodal model utilising images alone. In total, 7671 images (3913 occlusal and 3758 smooth surfaces) were collected from children across different dentition stages. Oral health questionnaires were also conducted. The images were clinically labelled into three categories: caries-free, early caries, and moderate-to-severe caries. A multimodal model integrating intraoral images and questionnaire data was trained and evaluated against an image-only unimodal model. The performance metrics included accuracy, precision, recall, F1 score, and the area under the receiver operating characteristic curve (ROC-AUC). SHapley Additive exPlanations (SHAP) analysis was conducted to interpret feature importance. For the multimodal model, occlusal surface classification achieved an accuracy of 92.3% for the caries-free category, 89.2% for the early caries category, and 93.7% for the moderate-to-severe caries category, with ROC-AUCs of 0.967, 0.911, and 0.972, respectively. The performance of smooth surfaces was similarly strong (an accuracy of 92.1% for the caries-free category, 96.7% for the early caries category, and 93.4% for the moderate-to-severe caries category), with ROC-AUCs of 0.938, 0.908, and 0.963, respectively. Incorporating questionnaire data improved recall for the early caries category without compromising overall accuracy. SHAP analysis provided interpretable insights into model decision-making. Compared with the image-only unimodal model, the multimodal deep learning model increased early caries recall while preserving overall diagnostic performance. This study describes a multimodal caries intelligent screening model, demonstrating its potential as a population-level screening tool. However, further large-scale, multicentre validation is necessary to confirm its generalisability and effectiveness.

Children in the mixed dentition stage are susceptible to dental caries, necessitating frequent examination and preventive measures. In this study, convolutional neural networks (CNNs) were used to develop a model for detecting oral health status in mixed dentition. Intraoral images of dentition spanning 2467 occlusal surfaces were collected from 7–9-year-old children for training and testing. The health status of all teeth, including caries, fillings, intraoral devices, pigments, and black stains, was tested. For the first permanent molar, indicator testing was conducted to determine the eruption conditions, fissure morphology, and sealing conditions. ResNet50 was trained for segmentation tasks, and TResNet was trained for detection tasks. The statistical analysis included calculations of the F1 score (F1) and area under the receiver operating characteristic (ROC) curve (AUC). The model was able to correctly detect caries (F1, 0.777; AUC, 0.964), fillings (F1, 0.746; AUC, 0.893), pigmentation (F1, 0.836; AUC, 0.870), intraoral devices (F1, 0.800; AUC, 1.00) and black stains (F1, 0.867; AUC, 0.988). The model demonstrated good capabilities for detecting deep fissures (F1, 0.927; AUC, 0.951), completely erupted teeth (F1, 0.882; AUC, 0.944), and sealed teeth (F1, 0.947; AUC, 0.986), but it demonstrated a weak ability to detect incompletely erupted teeth (F1, 0.727; AUC, 0.871) and shallow fissures (F1, 0.632; AUC, 0.934). The performance of the model was verified in detecting tooth status and screening teeth suitable for fissure sealants, but the scale of rare labels was also inadequate and requires optimization.