A Commentary on Ramezanzade S, Dascalu TL, Bakhshandeh A, Uribe SE, Ibragimov B, Bjørndal L. The impact of training dental students to use an artificial intelligence-based platform for pulp exposure prediction prior to deep caries excavation: a proof-of-concept randomized controlled trial. Int Endod J. 2026;59:1248–56. <a href="https://doi.org/10.1111/iej.70046">https://doi.org/10.1111/iej.70046</a> Design This randomized controlled trial (RCT) evaluated whether a structured educational intervention could improve dental students’ interaction with an artificial intelligence (AI)-based decision-support system developed to predict pulp exposure before excavation of deep carious lesions. Case selection Eighteen dental students were randomly allocated to either an experimental group receiving a one-hour personalized training session on the use of the AI platform or a control group receiving a brief introductory video. Participants subsequently completed a case-based assessment involving radiographic evaluation of deep carious lesions and prediction of pulp exposure risk using the AI system. Data analysis The primary outcome was agreement with AI recommendations (“agreeableness with AI”). Secondary outcomes included diagnostic accuracy, sensitivity, specificity, F1-score, and response time. Outcomes were compared between groups, and the findings were used to estimate the sample size required for a future definitive trial. Results Participants who received AI-focused training demonstrated greater agreement with AI recommendations than controls. However, improvements in agreement were not accompanied by meaningful differences in diagnostic accuracy, sensitivity, specificity, or F1-score. Response times were slightly shorter among trained participants. The findings suggest that targeted instruction may influence how users interact with AI systems, although objective diagnostic performance remained largely unchanged. Conclusions A short, personalized training session may increase dental students’ agreement with AI-generated predictions of pulp exposure during deep caries excavation. However, the intervention did not substantially improve diagnostic performance, and no patient-centered outcomes were assessed. Larger studies are needed to determine whether AI training can enhance clinical decision-making and improve outcomes relevant to the management of deep carious lesions.

Can artificial intelligence training improve clinical decision-making during deep caries excavation?

期刊论文
发表日期：

2026年7月27日

查看全文

Malik Alkabazi,
Melek Tassoker

摘要

A Commentary on

Ramezanzade S, Dascalu TL, Bakhshandeh A, Uribe SE, Ibragimov B, Bjørndal L.

The impact of training dental students to use an artificial intelligence-based platform for pulp exposure prediction prior to deep caries excavation: a proof-of-concept randomized controlled trial. Int Endod J. 2026;59:1248–56. https://doi.org/10.1111/iej.70046

Design

This randomized controlled trial (RCT) evaluated whether a structured educational intervention could improve dental students’ interaction with an artificial intelligence (AI)-based decision-support system developed to predict pulp exposure before excavation of deep carious lesions.

Case selection

Eighteen dental students were randomly allocated to either an experimental group receiving a one-hour personalized training session on the use of the AI platform or a control group receiving a brief introductory video. Participants subsequently completed a case-based assessment involving radiographic evaluation of deep carious lesions and prediction of pulp exposure risk using the AI system.

Data analysis

The primary outcome was agreement with AI recommendations (“agreeableness with AI”). Secondary outcomes included diagnostic accuracy, sensitivity, specificity, F1-score, and response time. Outcomes were compared between groups, and the findings were used to estimate the sample size required for a future definitive trial.

Results

Participants who received AI-focused training demonstrated greater agreement with AI recommendations than controls. However, improvements in agreement were not accompanied by meaningful differences in diagnostic accuracy, sensitivity, specificity, or F1-score. Response times were slightly shorter among trained participants. The findings suggest that targeted instruction may influence how users interact with AI systems, although objective diagnostic performance remained largely unchanged.

Conclusions

A short, personalized training session may increase dental students’ agreement with AI-generated predictions of pulp exposure during deep caries excavation. However, the intervention did not substantially improve diagnostic performance, and no patient-centered outcomes were assessed. Larger studies are needed to determine whether AI training can enhance clinical decision-making and improve outcomes relevant to the management of deep carious lesions.