AI-driven segmentation of the pulp cavity system in mandibular molars on CBCT images using convolutional neural networks
2024

AI Tool for Pulp Cavity Segmentation in Mandibular Molars

Sample size: 66 publication 10 minutes Evidence: high

Author Information

Author(s): Slim Marie Louise, Jacobs Reinhilde, de Souza Leal Renata MaĆ­ra, Fontenele Rocharles Cavalcante

Primary Institution: OMFS-IMPATH Research Group, Department of Imaging and Pathology, Faculty of Medicine, KU Leuven, Belgium

Hypothesis

The AI-driven tool would not be able to provide an accurate and time-efficient segmentation of the pulp cavity system in mandibular molars.

Conclusion

The AI-driven segmentation proved to be accurate and time-efficient in segmenting the pulp cavity system in mandibular molars.

Supporting Evidence

  • The AI tool achieved a Dice similarity coefficient of 88% for first molars and 90% for second molars.
  • The 95% Hausdorff distance was lower for AI-driven segmentation compared to manual segmentation.
  • AI-driven segmentation took an average of 4.3 seconds, significantly faster than manual segmentation which took 2349 seconds.

Takeaway

A computer program can quickly and accurately identify the insides of back teeth, helping dentists do their jobs better and faster.

Methodology

The study used 66 CBCT scans divided into training, validation, and testing sets to develop and validate an AI tool for segmentation.

Potential Biases

The study excluded scans with significant artefacts, which may limit the AI tool's performance in more complex cases.

Limitations

The dataset was limited to specific CBCT devices, which may affect generalizability, and the training dataset was relatively small.

Statistical Information

P-Value

p>0.05 for accuracy metrics between first and second molars; p<0.05 for time efficiency comparisons.

Statistical Significance

p<0.05

Digital Object Identifier (DOI)

10.1007/s00784-024-06009-2

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication