VP-Nets : Efficient automatic localization of key brain structures in 3D fetal neurosonography
<p>Three-dimensional (3D) fetal neurosonography is used clinically to detect cerebral abnormalities and to assess growth in the developing brain. However, manual identification of key brain structures in 3D ul- trasound images requires expertise to perform and even then is tedious. Inspired by...
| Hoofdauteurs: | , , |
|---|---|
| Formaat: | Journal article |
| Taal: | English |
| Gepubliceerd in: |
Elsevier
2018
|
| _version_ | 1826306289389535232 |
|---|---|
| author | Huang, R Xie, W Noble, JA |
| author_facet | Huang, R Xie, W Noble, JA |
| author_sort | Huang, R |
| collection | OXFORD |
| description | <p>Three-dimensional (3D) fetal neurosonography is used clinically to detect cerebral abnormalities and to assess growth in the developing brain. However, manual identification of key brain structures in 3D ul- trasound images requires expertise to perform and even then is tedious. Inspired by how sonographers view and interact with volumes during real-time clinical scanning, we propose an efficient automatic method to simultaneously localize multiple brain structures in 3D fetal neurosonography. The proposed View-based Projection Networks (VP-Nets) , uses three view-based Convolutional Neural Networks (CNNs), to simplify 3D localizations by directly predicting 2D projections of the key structures onto three anatom- ical views.</p> <br/> <p>While designed for efficient use of data and GPU memory, the proposed VP-Nets allows for full- resolution 3D prediction. We investigated parameters that influence the performance of VP-Nets, e.g. depth and number of feature channels. Moreover, we demonstrate that the model can pinpoint the struc- ture in 3D space by visualizing the trained VP-Nets, despite only 2D supervision being provided for a sin- gle stream during training. For comparison, we implemented two other baseline solutions based on Ran- dom Forest and 3D U-Nets. In the reported experiments, VP-Nets consistently outperformed other meth- ods on localization. To test the importance of loss function, two identical models are trained with binary corss-entropy and dice coefficient loss respectively. Our best VP-Net model achieved prediction center deviation: 1.8 ±1.4 mm, size difference: 1.9 ±1.5 mm, and 3D Intersection Over Union (IOU): 63.2 ±14.7% when compared to the ground truth. To make the whole pipeline intervention free, we also implement a skull-stripping tool using 3D CNN, which achieves high segmentation accuracy. As a result, the proposed processing pipeline takes a raw ultrasound brain image as input, and output a skull-stripped image with five detected key brain structures.</p> |
| first_indexed | 2024-03-07T06:45:42Z |
| format | Journal article |
| id | oxford-uuid:fac60e26-034b-471d-9a8b-6aefec9dacb4 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-07T06:45:42Z |
| publishDate | 2018 |
| publisher | Elsevier |
| record_format | dspace |
| spelling | oxford-uuid:fac60e26-034b-471d-9a8b-6aefec9dacb42022-03-27T13:08:54ZVP-Nets : Efficient automatic localization of key brain structures in 3D fetal neurosonographyJournal articlehttp://purl.org/coar/resource_type/c_dcae04bcuuid:fac60e26-034b-471d-9a8b-6aefec9dacb4EnglishSymplectic Elements at OxfordElsevier2018Huang, RXie, WNoble, JA<p>Three-dimensional (3D) fetal neurosonography is used clinically to detect cerebral abnormalities and to assess growth in the developing brain. However, manual identification of key brain structures in 3D ul- trasound images requires expertise to perform and even then is tedious. Inspired by how sonographers view and interact with volumes during real-time clinical scanning, we propose an efficient automatic method to simultaneously localize multiple brain structures in 3D fetal neurosonography. The proposed View-based Projection Networks (VP-Nets) , uses three view-based Convolutional Neural Networks (CNNs), to simplify 3D localizations by directly predicting 2D projections of the key structures onto three anatom- ical views.</p> <br/> <p>While designed for efficient use of data and GPU memory, the proposed VP-Nets allows for full- resolution 3D prediction. We investigated parameters that influence the performance of VP-Nets, e.g. depth and number of feature channels. Moreover, we demonstrate that the model can pinpoint the struc- ture in 3D space by visualizing the trained VP-Nets, despite only 2D supervision being provided for a sin- gle stream during training. For comparison, we implemented two other baseline solutions based on Ran- dom Forest and 3D U-Nets. In the reported experiments, VP-Nets consistently outperformed other meth- ods on localization. To test the importance of loss function, two identical models are trained with binary corss-entropy and dice coefficient loss respectively. Our best VP-Net model achieved prediction center deviation: 1.8 ±1.4 mm, size difference: 1.9 ±1.5 mm, and 3D Intersection Over Union (IOU): 63.2 ±14.7% when compared to the ground truth. To make the whole pipeline intervention free, we also implement a skull-stripping tool using 3D CNN, which achieves high segmentation accuracy. As a result, the proposed processing pipeline takes a raw ultrasound brain image as input, and output a skull-stripped image with five detected key brain structures.</p> |
| spellingShingle | Huang, R Xie, W Noble, JA VP-Nets : Efficient automatic localization of key brain structures in 3D fetal neurosonography |
| title | VP-Nets : Efficient automatic localization of key brain structures in 3D fetal neurosonography |
| title_full | VP-Nets : Efficient automatic localization of key brain structures in 3D fetal neurosonography |
| title_fullStr | VP-Nets : Efficient automatic localization of key brain structures in 3D fetal neurosonography |
| title_full_unstemmed | VP-Nets : Efficient automatic localization of key brain structures in 3D fetal neurosonography |
| title_short | VP-Nets : Efficient automatic localization of key brain structures in 3D fetal neurosonography |
| title_sort | vp nets efficient automatic localization of key brain structures in 3d fetal neurosonography |
| work_keys_str_mv | AT huangr vpnetsefficientautomaticlocalizationofkeybrainstructuresin3dfetalneurosonography AT xiew vpnetsefficientautomaticlocalizationofkeybrainstructuresin3dfetalneurosonography AT nobleja vpnetsefficientautomaticlocalizationofkeybrainstructuresin3dfetalneurosonography |