Development and Evaluation of Deep Learning-Based Automatic Segmentation Model for Skull Zero TE MRI in Children
1
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
Purpose: To develop and evaluate a deep learning technique to automatically segment bone structures in zero echo time (ZTE) for skull magnetic resonance imaging (MRI) in children. Materials and Methods: From January to December 2021, 38 bone ZTE MRIs from infants and children (age range, 1–31 months) were collected for model development. Mask images were generated by manually segmenting the craniofacial bone using a commercial segmentation program. Among them, 35 ZTE series were used to train the three-dimensional (3D)-nnUnet deep learning model and the remaining three series were used for model validation. A temporally different dataset of 19 ZTE bone MRIs obtained in May 2022 from infants and children (age range, 3–168 months) was used to determine the model’s performance. Dice similarity coefficient was calculated for each test case. From 3D volume rendering images, segmentation accuracy, overall image quality, and visibility of cranial sutures were subjectively evaluated on a 5-point scale and compared with ground truth data from manual segmentation. Reasons for segmentation failure were analyzed using axially segmented ZTE images. Results: For the test set, the mean Dice similarity coefficient was 0.985 ± 0.019. The segmentation accuracy was lower than the ground truth without showing a statistically significant difference between the two (3.39 ± 1.11 vs. 3.73 ± 0.77, p = 0.055). The overall image quality and suture visibility showed no significant difference (3.34 ± 0.75 vs. 3.42 ± 0.69, p = 0.317; 3.55 ± 0.97 vs. 3.60 ± 0.95, p = 0.157). Common reasons for low segmentation accuracy were well-pneumatized sinuses, metal artifacts, skin at the vertex level, and bones too thin. Conclusion: The deep learning-based automatic segmentation technique of bone ZTE MRIs showed comparable segmentation performance to manual segmentation. Using the deep learning-based segmentation results, acceptable 3D-volume rendering images of craniofacial bones were generated.
Abstract
Purpose: To develop and evaluate a deep learning technique to automatically segment bone structures in zero echo time (ZTE) for skull magnetic resonance imaging (MRI) in children. Materials and Methods: From January to December 2021, 38 bone ZTE MRIs from infants and children (age range, 1–31 months) were collected for model development. Mask images were generated by manually segmenting the craniofacial bone using a commercial segmentation program. Among them, 35 ZTE series were used to train the three-dimensional (3D)-nnUnet deep learning model and the remaining three series were used for model validation. A temporally different dataset of 19 ZTE bone MRIs obtained in May 2022 from infants and children (age range, 3–168 months) was used to determine the model’s performance. Dice similarity coefficient was calculated for each test case. From 3D volume rendering images, segmentation accuracy, overall image quality, and visibility of cranial sutures were subjectively evaluated on a 5-point scale and compared with ground truth data from manual segmentation. Reasons for segmentation failure were analyzed using axially segmented ZTE images. Results: For the test set, the mean Dice similarity coefficient was 0.985 ± 0.019. The segmentation accuracy was lower than the ground truth without showing a statistically significant difference between the two (3.39 ± 1.11 vs. 3.73 ± 0.77, p = 0.055). The overall image quality and suture visibility showed no significant difference (3.34 ± 0.75 vs. 3.42 ± 0.69, p = 0.317; 3.55 ± 0.97 vs. 3.60 ± 0.95, p = 0.157). Common reasons for low segmentation accuracy were well-pneumatized sinuses, metal artifacts, skin at the vertex level, and bones too thin. Conclusion: The deep learning-based automatic segmentation technique of bone ZTE MRIs showed comparable segmentation performance to manual segmentation. Using the deep learning-based segmentation results, acceptable 3D-volume rendering images of craniofacial bones were generated.
참고문헌 (20)
- Anderson PJ, "Application of three-dimensional computed tomography in craniofacial clinical practice and research", Aust Dent J, 59(Suppl 1), pp.174~185 (2014)
- Committee on Environmental Health, "Risk of ionizing radiation exposure to children : a subject review", Pediatrics, 101(4 Pt 1), pp.717~719 (1998)
- Schieban K, "ZTE imaging with enhanced flip angle using modulated excitation", Magn Reson Med, 74, pp.684~693 (2015)
- Weiger M, "MRI with zero echo time : hard versus sweep pulse excitation", Magn Reson Med, 66, pp.379~389 (2011)
- Wiesinger F, "Zero TE MR bone imaging in the head", Magn Reson Med, 75, pp.107~114 (2016)
- Argentieri EC, "Diagnostic accuracy of zero-echo time MRI for the evaluation of cervical neural foraminal stenosis", Spine(Phila Pa 1976), 43, pp.928~933 (2018)
- Breighner RE, "Technical developments : zero echo time imaging of the shoulder : enhanced osseous detail by using MR imaging", Radiology, 286, pp.960~966 (2018)
- Lee C, "CT-like MRI using the zero-TE technique for osseous changes of the TMJ", Dentomaxillofac Radiol, 49, pp.20190272~ (2020)
- Delso G, "Clinical evaluation of zero-echo-time MR imaging for the segmentation of the skull", J Nucl Med, 56, pp.417~422 (2015)
- Lu A, "Zero TE MRI for craniofacial bone imaging", AJNR Am J Neuroradiol, 40, pp.1562~1566 (2019)
- Cho SB, "Clinical feasibility of zero TE skull MRI in patients with head trauma in comparison with CT : a single-center study", AJNR Am J Neuroradiol, 40, pp.109~115 (2019)
- Patel KB, "3D pediatric cranial bone imaging using high-resolution MRI for visualizing cranial sutures : a pilot study", J Neurosurg Pediatr, 26, pp.311~317 (2020)
- Kralik SF, "Black bone MRI with 3D reconstruction for the detection of skull fractures in children with suspected abusive head trauma", Neuroradiology, 61, pp.81~87 (2019)
- Yoo Seung-Jin, "Automated Lung Segmentation on Chest Computed Tomography Images with Extensive Lung Parenchymal Abnormalities Using a Deep Neural Network", Korean Journal of Radiology, 22(3), pp.476~488 (2021), 대한영상의학회[DOI]
- Lee SB, "Automated segmentation of whole-body CT images for body composition analysis in pediatric patients using a deep neural network", Eur Radiol, 32, pp.8463~8472 (2022)
- Leynes AP, "Zero-echo-time and Dixon deep pseudo-CT(ZeDD CT) : direct generation of pseudo-CT images for pelvic PET", MRI attenuation correction using deep convolutional neural networks with multiparametric MRI, 2018, pp.59, 852~858
- Spadea MF, "Deep learning based synthetic-CT generation in radiotherapy and PET : a review", Med Phys, 48, pp.6537~6566 (2021)
- Wiesinger F, "Zero-TE MRI : principles and applications in the head and neck", Br J Radiol, 95, pp.20220059~ (2022)
- Johnson EM, "Improved cortical bone specificity in UTE MR imaging", Magn Reson Med, 77, pp.684~695 (2017)
- Eley KA, "Automated 3D MRI rendering of the craniofacial skeleton : using ZTE to drive the segmentation of black bone and FIESTA-C images", Neuroradiology, 63, pp.91~98 (2021)