Implementing a Corrective Annotation No Code Artificial Intelligence-based Software to Detect Several Radiographic Features Associated With Unsatisfactory Endodontic Treatment: A Randomized Controlled Trial
Trial Snapshot
- Phase
- Not Applicable
- Status
- Not yet recruiting
- Sponsor
- University of Copenhagen
- Enrollment
- 80
- Primary Endpoint
- Accuracy
Study Overview
Brief Summary
Developing neural network-based models for image analysis can be time-consuming, requiring dataset design and model training. No-code AI platforms allow users to annotate object features without coding. Corrective annotation, a "human-in-the-loop" approach, refines AI segmentations iteratively. Dentistry has seen success with no-code AI for segmenting dental restorations. This study aims to assess radiographic features related to root canal treatment quality using a "human-in-the-loop" approach.
Detailed Description
The emergence of artificial intelligence (AI) and specifically deep learning (DL) have shown great potential in finding radiographic features and treatment planning in the field of cariology and endodontics. A growing body of literature suggests that DL models might assist dental practitioners in detecting radiographic features such as carious lesions, and periapical lesions, as well as predicting the risk of pulp exposure when doing caries excavation therapy. Although, the current literature lacks sufficient research on the interaction of participants and AI in an AI-based platform for detecting features associated with technical quality of endodontic treatment. This prospective randomized controlled trial aims to assess the performance of students when using an AI-based platform for detecting features associated with technical quality of endodontic treatment and predicting the long term prognosis of the treatment. The hypothesis is that participants' performance in the group with access to AI responses is similar to the control group without access to AI responses.
Study Design
- Study Type
- Interventional
- Allocation
- Randomized
- Intervention Model
- Parallel
- Primary Purpose
- Diagnostic
- Masking
- Double (Participant, Outcomes Assessor)
Eligibility Criteria
- Ages
- 20 Years to 40 Years (Adult)
- Sex
- All
- Accepts Healthy Volunteers
- Yes
Inclusion Criteria
- •1.Being a last year dental student at the university of Copenhagen
Exclusion Criteria
- •Having any previous AI-related experiences
- •Not accepting to sign the informed consent
Arms & Interventions
participants using guidance from artificial Intelligence
the experimental arm refers to the group of participants who have access to the AI-based platform for detecting features associated with the technical quality of endodontic treatment. These participants will utilize the AI assistance during the study.
Intervention: AI guidance for finding radiographic features (Device)
Control arm without any guidance from artificial Intelligence
the control arm consists of participants who do not have access to the AI-based platform. They will perform the same tasks or assessments as those in the experimental arm but without the assistance of AI.
Outcomes
Primary Outcomes
Accuracy
Time Frame: through data collection, an average of 6 months
Accuracy represents how closely a result aligns with the true value or standard. Accuracy of participants at experiment and control group in correctly finding the radiographic features and predicting the outcomes is one of our primary outcomes. the reference for the comparison is the consensus of three experts in dentistry.
Sensitivity
Time Frame: through data collection, an average of 6 months
This measure quantifies the proportion of true positive results (correctly identified cases) out of all positive cases. High sensitivity indicates that one is good at detecting the condition. Sensitivity of participants at experiment and control group in correctly finding the radiographic features and predicting the outcomes is one of our primary outcomes. The comparison is made against the consensus judgment of three experts in dentistry.
Specificity
Time Frame: through data collection, an average of 6 months
Specificity measures the proportion of true negative results out of all negative cases. Specificity of participants at experiment and control group in correctly finding the radiographic features and predicting the outcomes is one of our primary outcomes. The comparison is made against the consensus judgment of three experts in dentistry.
Secondary Outcomes
No secondary outcomes reported
Investigators
Lars Bjørndal
Associate Professor
University of Copenhagen