Bone segmentation in contrast enhanced whole-body computed tomography

Leydon, Patrick; O’Connell, Martin; Greene, Derek; Curran, Kathleen M.

doi:10.1088/2057-1976/ac37ab

Cited by 6 publications

(3 citation statements)

References 54 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The values of Dice coefficient obtained for each type of bone were: 86%(Th 7), 85% (L3), 88% (sacrum), 84% (7th rib) and 83%(sternium). The Dice coefficient value between golden standard and segmentation performed by neural networks in discussed works varies from 83% to 97.9%, where only predictions obtained in [33] resulted in higher value of mDice coefficient then described U-net-Resnet model. The obtained results cannot be directly compared because they apply to different range of bone structures of various anatomical characteristics.…”

Section: Discussionmentioning

confidence: 91%

See 1 more Smart Citation

Bulletin of the Polish Academy of Sciences: Technical Sciences

Krawczyk

Starzyński

2021

View full text Add to dashboard Cite

chapter presents the results obtained. The last chapter focuses on conclusions and perspectives for further work.The preliminary results of the work were presented in the CPEE 2020 conference [12]. The paper broadens the subject with segmentation performed by FCN network and modifications of the U-net network. Evaluation of the results was performed on much larger dataset with the use of additional fitting measures. Literature review and goalRecent publications reported that deep neural networks are capable to perform segmentation task on medical images with the accuracy comparable to human expert. One of the common tasks performed successfully by the deep neural network is organ segmentation. In [13] novel superpixel-based and boundary sensitive CNN was used for liver segmentation out of CT data and achieved high Dice coefficient value equal to 97.31 ± 0.36%. Authors of [14] successfully performed segmentation of head and neck organs at risks utilising 3D U-net architecture with performance similar to the expert.The segmentation process is also used in detection of pathological changes in the human body. Authors of [15] performed segmentation of polycystic kidneys with mean volume difference 0.68 ± 2.2.% in comparison to expert segmentation. Various deep convolutional neural networks are also utilised to the task of automatic brain tumour segmentation [16].In addition to the CT and MRI imaging analysis, deep neural networks are also used in the processing of histopathological images for example in segmenting and classifying epithelial and stromal regions [17] or in damaged tissue detection [18].Several papers tackled the subject of bone structure segmentation. In [19] a custom deep learning method was developed for segmentation of 49 bones out of PET/CT datasets.

show abstract

Section: Discussionmentioning

confidence: 91%

“…In [33] the U-net network prediction combined with preprocessing techniques resulted in segmentation of bone structures in whole body CT scans with the value of mDice coefficient equal to 97.9% and 96.5% for two in-house data-sets and 93.4% for external dataset.…”

Section: Discussionmentioning

confidence: 99%

Bulletin of the Polish Academy of Sciences: Technical Sciences

Krawczyk

Starzyński

2021

View full text Add to dashboard Cite

show abstract

“…Bone tissue segmentation from CT has been shown to work well using slicewise 2D CNN-based segmentation algorithms [4][5][6]. The tasks and solutions become more varied when moving from bone-tissue segmentation to distinct bone segmentation (our task) where we distinguish individual bones.…”

Section: Introductionmentioning

confidence: 99%

Improved distinct bone segmentation in upper-body CT through multi-resolution networks

Schnider¹,

Wolleb²,

Huck³

et al. 2023

Preprint

View full text Add to dashboard Cite

Automated distinct bone segmentation from CT scans is widely used in planning and navigation workflows. U-Net variants are known to provide excellent results in supervised semantic segmentation. However, in distinct bone segmentation from upper body CTs a large field of view and a computationally taxing 3D architecture are required. This leads to low-resolution results lacking detail or localisation errors due to missing spatial context when using high-resolution inputs. Methods: We propose to solve this problem by using end-to-end trainable segmentation networks that combine several 3D U-Nets working at different resolutions. Our approach, which extends and generalizes HookNet and MRN, captures spatial information at a lower resolution and skips the encoded information to the target network, which operates on smaller high-resolution inputs. We evaluated our proposed architecture against single resolution networks and performed an ablation study on information concatenation and the number of context networks. Results: Our proposed best network achieves a median DSC of 0.86 taken over all 125 segmented bone classes and reduces the confusion among similar-looking bones in different locations. These results outperform our previously published 3D U-Net baseline results on the task and distinct-bone segmentation results reported by other groups.

show abstract

Efficient lower-limb segmentation for large-scale volumetric CT by using projection view and voxel group attention

Chen

Xie

et al. 2022

Med Biol Eng Comput

View full text Add to dashboard Cite

Bone segmentation in contrast enhanced whole-body computed tomography

Cited by 6 publications

References 54 publications

Bulletin of the Polish Academy of Sciences: Technical Sciences

Bulletin of the Polish Academy of Sciences: Technical Sciences

Improved distinct bone segmentation in upper-body CT through multi-resolution networks

Efficient lower-limb segmentation for large-scale volumetric CT by using projection view and voxel group attention

Contact Info

Product

Resources

About