Accuracy and precision of manual segmentation of the maxillary sinus in MR images-a method study

TY - JOUR

T1 - Accuracy and precision of manual segmentation of the maxillary sinus in MR images-a method study

AU - Andersen, Tobias N

AU - Darvann, Tron Andre

AU - Murakami, Shumei

AU - Larsen, Per

AU - Senda, Yurie

AU - Bilde, Anders

AU - Buchwald, Christian V

AU - Kreiborg, Sven

PY - 2018/5

Y1 - 2018/5

N2 - Objective: To assess the accuracy and precision of segmentation of the maxillary sinus in MR images to evaluate the potential usefulness of this modality in longitudinal studies of sinus development. Methods: A total of 15 healthy subjects who had been both craniofacial CT and MR scanned were included and the 30 maxillary sinus volumes were evaluated using segmentation. Two of the authors did segmentation of MRI and one of these authors did double segmentation. Agreement in results between CT and MRI as well as inter- and intraexaminer errors were evaluated by statistical and three-dimensional analysis. Results: The intraclass correlation coefficient for volume measurements for both method error, inter- and intraexaminer agreement were > 0.9 [maximal 95% confidence interval of 0.989-0.997, p < 0.001] and the limit of agreement for all parameters were < 5.1%. Segmentation errors were quantified in terms of overlap [Dice Coefficient (DICE) > 0.9 = excellent agreement] and border distance [95% percentile Hausdorff Distance (HD) < 2 mm = acceptable agreement]. The results were replicable and not influenced by systematic errors. Conclusion: We found a high accuracy and precision of manual segmentation of the maxillary sinus in MR images. The largest mean errors were found close to the orbit and the teeth. Advances in knowledge: MRI can be used for 3D models of the paranasal sinuses with equally good results as CT and allows longitudinal follow-up of sinus development.

AB - Objective: To assess the accuracy and precision of segmentation of the maxillary sinus in MR images to evaluate the potential usefulness of this modality in longitudinal studies of sinus development. Methods: A total of 15 healthy subjects who had been both craniofacial CT and MR scanned were included and the 30 maxillary sinus volumes were evaluated using segmentation. Two of the authors did segmentation of MRI and one of these authors did double segmentation. Agreement in results between CT and MRI as well as inter- and intraexaminer errors were evaluated by statistical and three-dimensional analysis. Results: The intraclass correlation coefficient for volume measurements for both method error, inter- and intraexaminer agreement were > 0.9 [maximal 95% confidence interval of 0.989-0.997, p < 0.001] and the limit of agreement for all parameters were < 5.1%. Segmentation errors were quantified in terms of overlap [Dice Coefficient (DICE) > 0.9 = excellent agreement] and border distance [95% percentile Hausdorff Distance (HD) < 2 mm = acceptable agreement]. The results were replicable and not influenced by systematic errors. Conclusion: We found a high accuracy and precision of manual segmentation of the maxillary sinus in MR images. The largest mean errors were found close to the orbit and the teeth. Advances in knowledge: MRI can be used for 3D models of the paranasal sinuses with equally good results as CT and allows longitudinal follow-up of sinus development.

KW - Adolescent

KW - Adult

KW - Female

KW - Human Development/physiology

KW - Humans

KW - Image Processing, Computer-Assisted/methods

KW - Imaging, Three-Dimensional/methods

KW - Magnetic Resonance Imaging/methods

KW - Male

KW - Maxillary Sinus/diagnostic imaging

KW - Observer Variation

KW - Reproducibility of Results

KW - Retrospective Studies

KW - Sensitivity and Specificity

KW - Tomography, X-Ray Computed/methods

KW - Young Adult

U2 - 10.1259/bjr.20170663

DO - 10.1259/bjr.20170663

M3 - Journal article

C2 - 29419324

SN - 0007-1285

VL - 91

JO - British Journal of Radiology

JF - British Journal of Radiology

IS - 1085

M1 - 20170663

ER -