Performance comparison of large language models in treatment planning for the restoration of endodontically treated teeth over time
These videos have been matched automatically. Contact us if they are not relevant.
These videos have been matched automatically. Contact us if they are not relevant.
View abstract on PubMed
Summary
This summary is machine-generated.Gemini excelled in restorative treatment planning for endodontically treated teeth, outperforming other large language models (LLMs). However, no LLM achieved perfect consistency, indicating a need for human oversight in clinical decision-making.
Area Of Science
- Artificial Intelligence in Dentistry
- Clinical Decision Support Systems
- Endodontic Treatment Planning
Background
- Large language models (LLMs) are increasingly explored for clinical applications.
- Evaluating LLM performance in specialized fields like endodontic restorative treatment planning is crucial.
- Understanding LLM response variability over time and with exposure to expert data is essential.
Purpose Of The Study
- To compare the performance of five leading LLMs in endodontic treatment planning.
- To assess LLM accuracy and completeness in restorative treatment planning scenarios.
- To evaluate the impact of repeated exposure and expert examples on LLM responses.
Main Methods
- Five LLMs (ChatGPT 4.5, DeepSeek R1, Gemini 2.5 Pro, Claude 3.7 Sonnet, Microsoft Copilot) were tested.
- Twenty-five case scenarios with a 39-item checklist were used for evaluation.
- LLMs were assessed over three weeks, with and without expert response examples, by blinded evaluators.
Main Results
- Gemini 2.5 Pro demonstrated superior performance in accuracy and completeness compared to DeepSeek and Microsoft Copilot.
- Claude 3.7 Sonnet showed significant accuracy improvement over time.
- Gemini and Claude showed increased completeness after exposure to expert examples, but no LLM achieved perfect repeatability or consistently complete/accurate responses.
Conclusions
- Gemini 2.5 Pro is the most effective LLM for endodontic restorative treatment planning among those tested.
- DeepSeek R1 exhibited the lowest performance.
- Current LLMs should be used as adjunctive tools, requiring professional human supervision due to limitations in consistency and completeness.
These videos have been matched automatically. Contact us if they are not relevant.
These videos have been matched automatically. Contact us if they are not relevant.