DDentalEdu
Ilustracija
IlustracijaIlustracija: DentalEdu (AI)
ZnanostJournal of Periodontology

Deep learning on CBCT reliably measures intrabony periodontal defects

A deep learning model segmented 48 intrabony defects in the mandible, and its measurements strongly correlated with a semi-automatic method (ICC 0.941).

Sažetak pripremio AI-asistent uredništva, uredila i odobrila redakcija prije objave. Uvijek provjerite izvorni rad prije kliničke primjene.

Automatic 3D reconstruction of defects

CBCT provides a detailed view of intrabony periodontal defects, but its full potential is realised only when the image is reconstructed in three dimensions. Manual and semi-automatic segmentation for this purpose are time-consuming and prone to human error, limiting their routine application in practice.

Palkovics et al. tested a multi-stage deep learning model based on the SegResNet architecture on CBCT images of patients with stage III and IV periodontitis. On the 3D models generated by the algorithm, they measured the depth, width, angle, and volume of a total of 48 intrabony defects in the mandible and compared them with the same measurements obtained by semi-automatic segmentation.

The linear and volumetric measurements of the two methods strongly correlated, with an intraclass correlation coefficient of 0.941. The authors conclude that CBCT segmentation using deep learning can be a valuable tool for enhancing periodontal diagnostics, as the three-dimensional representation of the defect is provided without the tedious manual marking of section by section.

For the clinician, this means that the dimensions and shape of the defect could be reliably assessed before flap elevation. Such an assessment facilitates the decision on whether the defect is suitable for regenerative surgery and how much material should be anticipated.

Limitations

The study included only defects in the mandible, so applicability to the maxilla still needs to be confirmed. The reference method was semi-automatic segmentation, not intraoperative measurement.

For your practice

  • In stage III and IV periodontitis with intrabony defects, 3D reconstruction of CBCT helps assess the depth, width, and angle of the defect before regenerative surgery.
  • Evaluate software with automatic segmentation according to published validation; this applies only to the mandible.
  • Continue to determine the indication for CBCT based on clinical need and radiation dose.