 |
| Title |
Little Evidence to Support Artificial Intelligence (AI) as a Tool for Proximal caries Progression Assessment for Pediatric Dentists |
| Clinical Question |
In children aged 5-12 years with proximal caries, can artificial intelligence, compared to pediatric dentists, predict proximal caries progression? |
| Clinical Bottom Line |
Current evidence demonstrates Artificial Intelligence's (AI) capability in detecting proximal caries, but its ability to predict caries progression remains unsupported. This limitation is clinically significant in pediatric dentistry, where early detection and accurate progression prediction are crucial for guiding treatment decisions. Failure to identify and monitor caries progression can result in delayed interventions, leading to more invasive treatments and poorer patient outcomes such as increased severity, pain and discomfort, Impacts on growth and development and negative behavioral and psychological effects over time. |
| Best Evidence |
(you may view more info by clicking on the PubMed ID link) |
| PubMed ID |
Author / Year |
Patient Group |
Study type
(level of evidence) |
| #1) 35399834 | Talpur/2022 | 12 studies included | Systematic review of non-randomized trials | | Key results | This systematic review included 12 articles that primarily employed Convolutional Neural Networks (CNN) and Artificial Neural Networks (ANN) as key components to detect caries using dental x-rays. The studies, published between 2008 and 2019, featured varying sample sizes for training data (ranging from 80 to 9,630) and testing data (ranging from 80 to 2,380). Data from 2020 lacked clear details regarding training and testing sample sizes, while a decreasing trend in the ratio of training to testing data was observed in studies from 2021 to 2022. The selected studies were limited to the diagnosis of proximal, root, and occlusal caries. Although different imaging modalities were explored, only 2D or 3D radiographs were used for dental caries diagnosis. | | #2) 38016620 | Arsiwala-Scheppach/2024 | 22 dentists who assessed 20 bitewings | Randomized Controlled Trial | | Key results | Dentists were employed at the dental hospital of Charité, Universitätsmedizin Berlin or in private practices in Berlin, Germany. The comparison of "fixation" between dentists assisted by AI (median = 68, IQR = 31, 116) and those without AI assistance (median = 47, IQR = 19, 100) revealed a significant difference (P = 0.01). Notably, dentists using AI demonstrated a less systematic visual search pattern for caries, with a lower frequency of lateral transitions from tooth to tooth.The analysis of fixation counts revealed that dentists using AI exhibited a higher median fixation count on teeth with carious lesions and/or restorations (median = 163, IQR = 104, 234) compared to those without AI (median = 138, IQR = 87, 204), with a statistically significant difference (p = 0.004).When focusing on teeth with lesions and/or restorations, the observed difference primarily originated from teeth with restorations. Dentists using AI had a higher median fixation count on these teeth (median = 68, IQR = 31, 116) compared to those without AI (median = 47, IQR = 19, 100), with a significant difference (p = 0.01).For teeth with carious lesions, dentists using AI had fewer fixations on teeth with D2 lesions (median = 17, IQR = 9, 32) compared to dentists without AI (median = 43, IQR = 20, 51), with a significant difference (p = 0.03).Among dentists using AI, the median fixation duration was significantly longer for teeth with carious lesions (median = 412 ms, IQR = 245, 692) compared to teeth with restorations (median = 292 ms, IQR = 221, 369), with a highly significant difference (p < 0.001). | | #3) 3456656 | Mertens/2022 | 20 bitewings of permanent teeth analyzed by 22 dentists | Randomized Controlled Trial | | Key results | Dentists assisted by AI demonstrated a significantly higher mean area under the Receiver Operating Characteristic (ROC) curve (0.89; 95% CI: 0.87–0.90) compared to those without AI (0.85; 95% CI: 0.83–0.86; p < 0.05). This improvement was primarily driven by a significantly higher sensitivity (0.81; 95% CI: 0.74–0.87 vs. 0.72; 95% CI: 0.64–0.79; p < 0.05), while specificity remained unaffected (p > 0.05). The rise in sensitivity was noted for enamel lesions but was not seen for early or advanced dentin lesions. However, the heightened sensitivity also led to an increase in both non-invasive and invasive treatment decisions (p < 0.05). | |
| Evidence Search |
proximal caries AND (artificial intelligence OR machine learning OR deep learning) |
Comments on
The Evidence |
Neural networks have demonstrated high accuracy as diagnostic adjuncts for caries detection; however, there is currently no standardized protocol for the data used to train and test these models. Trials comparing caries diagnosis using neural networks against the histological gold standard are limited, and the potential for generalizing these results is further constrained by highly specialized equipment used to analyze and inform the results. As a result, the quality of evidence is considered low, and the findings should be interpreted with caution. While AI enhances dentists' diagnostic accuracy—particularly in detecting enamel lesions—it may also lead to increased decisions favoring invasive treatments. Additionally, AI systems could introduce diagnostic bias in how dentists interpret radiographs. Although neural networks excel at caries detection, they currently lack the capability to classify caries progression based on severity, such as distinguishing between acute, chronic, or arrested stages. |
| Applicability |
AI's adjunctive role in diagnosing early caries in children facilitates timely detection of carious lesions, enabling conservative treatments that reduce the need for invasive procedures. Furthermore, non-invasive and minimally invasive treatments can lower overall costs while promoting better health outcomes by preventing complications associated with untreated dental issues. Accurate early diagnoses also contribute to more positive dental experiences, fostering a favorable attitude toward dental care in children. Ultimately, AI has the potential to enhance the quality of life for young patients and encourage preventive behaviors among parents and caregivers. |
| Specialty/Discipline |
(Public Health) (General Dentistry) (Pediatric Dentistry) (Restorative Dentistry) |
| Keywords |
Artificial Intelligence, deep learning, neural networks, machine learning, proximal caries, caries progression
|
| ID# |
3566 |
| Date of submission: |
10/18/2024 |
| E-mail |
chachappan@livemail.uthscsa.edu , nguyend19@livemail.uthscsa.edu |
| Author |
Dayl Chachappan, Dylan Nguyen |
| Co-author(s) |
|
| Co-author(s) e-mail |
Maria Jose Cervantes, D.D.S, M.S., F.A.A.P.D. |
| Faculty mentor/Co-author |
cervantesmen@uthscsa.edu |
| Faculty mentor/Co-author e-mail |
|
Basic Science Rationale
(Mechanisms that may account for and/or explain the clinical question, i.e. is the answer to the clinical question consistent with basic biological, physical and/or behavioral science principles, laws and research?) |
post a rationale |
| None available | |
|
Comments and Evidence-Based Updates on the CAT
(FOR PRACTICING DENTISTS', FACULTY, RESIDENTS and/or STUDENTS COMMENTS ON PUBLISHED CATs) |
post a comment |
| None available | |
|
|
