DENTEX CHALLENGE 2023

Dental Enumeration and Diagnosis on Panoramic X-rays


METRICS for ASSESSMENT

In order to evaluate the performance of the participating teams, the DENTEX Challenge utilizes a comprehensive set of metrics, including AP50, AP75, AP, and AR. These performance evaluation metrics are calculated based on a fully annotated testing dataset that encompasses quadrant-enumeration-diagnosis data for the quadrant, enumeration, and diagnosis aspects of dental image analysis. This results in a total of 12 metrics that are employed for evaluating the performance of the participating teams.

Following the VerSe'19 challenge, the VerSe'20 challenge, and the Brain Tumor Segmentation (BraTS) challenge, we use a point-based ranking process described in VerSe. This ranking process for the DENTEX Challenge involves comparing every possible pair of teams for each metric using the Wilcoxon Signed Rank Test. In each comparison, the team that performs statistically better (with a p-value of 0.001) earns one point. After all comparisons have been completed, each team has a "total point count," which reflects the number of times it outperformed its counterparts. To further enhance the robustness and reliability of the ranking system, bootstrapping techniques are incorporated by resampling 10% of the data and repeating the ranking process for the remaining x-ray images. This generates a "total point count" for each team for every excluded x-ray image. The "total point counts" obtained are then combined to establish the final ranking for the DENTEX Challenge.

The final top three ranked participating teams, according to their evaluation against the testing data, will be invited to MICCAI 2023 for a presentation of their methods and results and to receive their monetary awards.