3D dento-maxillary osteolytic lesion and active contour segmentation pilot study in CBCT: semi-automatic vs manual methods

Abstract

Objectives: This study was designed to evaluate the reliability of a semi-automatic segmentation tool for dento-maxillary osteolytic image analysis compared with manually defined segmentation in CBCT scans.

Methods: Five CBCT scans were selected from patients for whom periapical radiolucency images were available. All images were obtained using a ProMax® 3D Mid Planmeca (Planmeca Oy, Helsinki, Finland) and were acquired with 200-μm voxel size. Two clinicians performed the manual segmentations. Four operators applied three different semi-automatic procedures. The volumes of the lesions were measured. An analysis of dispersion was made for each procedure and each case. An ANOVA was used to evaluate the operator effect. Non-paired t-tests were used to compare semi-automatic procedures with the manual procedure. Statistical significance was set at α = 0.01.

Results: The coefficients of variation for the manual procedure were 2.5-3.5% on average. There was no statistical difference between the two operators. The results of manual procedures can be used as a reference. For the semi-automatic procedures, the dispersion around the mean can be elevated depending on the operator and case. ANOVA revealed significant differences between the operators for the three techniques according to cases.

Conclusions: Region-based segmentation was only comparable with the manual procedure for delineating a circumscribed osteolytic dento-maxillary lesion. The semi-automatic segmentations tested are interesting but are limited to complex surface structures. A methodology that combines the strengths of both methods could be of interest and should be tested. The improvement in the image analysis that is possible through the segmentation procedure and CBCT image quality could be of value.

Keywords: 3D dento-maxillary imaging; CBCT; dento-maxillary osteolytic image; image segmentation; jaw cysts.

Figures

Figure 1

Example of manual segmentation result. A, anterior; I, inferior; L, left; P, posterior; R, right; S, superior.