Validation of a windowing protocol for accurate in vivo tooth segmentation using i-CAT cone beam computed tomography

Rastegar, Babak; Thumilaire, Brice; Odri, G.; Siciliano, Sergio; Zapała, Jan; Mahy, Pierre; Olszewski, Raphaël

doi:10.17219/acem/68117

Cited by 11 publications

(13 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

Section: Discussionmentioning

confidence: 99%

“…Segmentation using a global threshold based method on image intensity values is common in medical modeling (33,(40)(41)(42). However, it has limitations in dental modeling both because of the variability of the density of the structures within the maxillofacial region and the technical capabilities of the devices and algorithms used in dental imaging (13,33). The density of tooth is heterogeneous from crown to apex and this natural feature leads to less clear difference between the root and the alveolar bone as the contrast between the root and bone decreases.…”

Section: Discussionmentioning

confidence: 99%

“…Performing the experiments under in vitro settings does not accurately mimic clinical environments, and therefore, they do not represent the real efficacy and accuracy of segmentation methods. Only a few studies have tried to determine the accuracy of volumetric analysis of teeth on in vivo images using CBCT (3,4,33).…”

Section: Discussionmentioning

confidence: 99%

“…Recently, the accuracy of tooth segmentation was evaluated using four different windowing protocols for manual method by comparing the volumes obtained from CBCT segmentations. Laser scanning method was employed as the gold standard and the results revealed that only one windowing protocol showed significantly closer volumes to the gold standard (33). Most of these studies evaluating the accuracy of different segmentation procedures have used manual segmentation as the gold standard (5,7,29).…”

Section: Discussionmentioning

confidence: 99%

“…The anatomic location of the object that would be segmented is a very important parameter (3,33) since the outlines of that particular object has to be clearly delineated both manually and automatically. However, easily outlined in vivo hard tissues such as temporomandibular joint and impacted third molars have been utilized in many papers (19,34,35).…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Is Manual Segmentation the Real Gold Standard for Tooth Segmentation? A Preliminary <i>in vivo</i> Study Using Cone-beam Computed Tomography Images

Sabanci¹,

Şener²,

Turhal³

et al. 2021

meandros

View full text Add to dashboard Cite

Objective: This study aimed to assess whether manual segmentation is an accurate method in tooth volume measurement and to compare the outcomes of manual, automatic, and semiautomatic segmentations on cone-beam computed tomography (CBCT) images by comparing each system with the water displacement method, which is the gold standard. Materials and Methods: CBCT images of l0 maxillary impacted teeth were used in this preliminary in vivo study. Following the acquisition of CBCT scans, manual, automatic, and semiautomatic segmentations were completed by the same operator. After surgical removal, the volumes of all impacted teeth were measured with the water displacement method, which was used as the gold standard. The volume of each segmented image was measured in mm 3 using the 3D-Doctor software. The established volumes of each segmented image were compared with those of the gold standard using the 95% confidence interval bootstrap percentiles. Intraobserver reliability was determined using the intraclass correlation coefficient. Results: All segmentation methods revealed significantly different volume values both from the gold standard and from each other (p=0.000). The semiautomatic segmentation demonstrated comparable performance with the manual method, and both systems provided comparable volumes with the gold standard than did the automatic method. Excellent intra-observer intraclass correlations were found for all protocols. Conclusion:The actual volumes of the specimen were not obtained by manual, semiautomatic, and automatic segmentations. Semiautomatic segmentation demonstrated comparable performance to the manual method, whereas automatic segmentation yielded the poorest values. The automatic and semiautomatic segmentations may be improved by the development and utilization of novel or hybrid segmentation algorithms for a faster process and more accurate results.

show abstract