Two Parallel Stages Deep Learning Network for Anterior Visual Pathway Segmentation

Li, Siqi; Chen, Zan; Guo, Wenlong; Zeng, Qingrun; Feng, Yuanjing

doi:10.1007/978-3-030-73018-5_22

Cited by 7 publications

(5 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For the OT segmentation, we built a deep learning model called TFusionSegmenter to receive as input the T1w image, FA image, and peaks image. We chose the FA image because the OT region in FA images generally has a high contrast 42 . We chose the peaks image as input because a previous study 32,43 verified that the fiber orientation distribution function (FOD) peaks could be used in the white matter tract segmentation.…”

Section: Methodsmentioning

confidence: 99%

“…We chose the FA image because the OT region in FA images generally has a high contrast. 42 We chose the peaks image as input because a previous study 32,43 verified that the fiber orientation distribution function (FOD) peaks could be used in the white matter tract segmentation. For the OC/ON segmentation, we built a model called NCSegmenter that only receives as input the T1w image.…”

Section: Template Bundlesmentioning

confidence: 99%

See 1 more Smart Citation

A unified global tractography framework for automatic visual pathway reconstruction

Yao

Zeng

et al. 2023

NMR in Biomedicine

Self Cite

View full text Add to dashboard Cite

The human visual pathway starts from the retina, passes through the retinogeniculate visual pathway, the optic radiation, and finally connects to the primary visual cortex. Diffusion MRI tractography is the only technology that can noninvasively reconstruct the visual pathway. However, complete and accurate visual pathway reconstruction is challenging because of the skull base environment and complex fiber geometries. Specifically, the optic nerve within the complex skull base environment can cause abnormal diffusion signals. The crossing and fanning fibers at the optic chiasm, and a sharp turn of Meyer's loop at the optic radiation, contribute to complex fiber geometries of the visual pathway. A fiber trajectory distribution (FTD) function‐based tractography method of our previous work and several high sensitivity tractography methods can reveal these complex fiber geometries, but are accompanied by false‐positive fibers. Thus, the related studies of the visual pathway mostly applied the expert region of interest selection strategy. However, interobserver variability is an issue in reconstructing an accurate visual pathway. In this paper, we propose a unified global tractography framework to automatically reconstruct the visual pathway. We first extend the FTD function to a high‐order streamline differential equation for global trajectory estimation. At the global level, the tractography process is simplified as the estimation of global trajectory distribution coefficients by minimizing the cost between trajectory distribution and the selected directions under the prior guidance by introducing the tractography template as anatomic priors. Furthermore, we use a deep learning‐based method and tractography template prior information to automatically generate the mask for tractography. The experimental results demonstrate that our proposed method can successfully reconstruct the visual pathway with high accuracy.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Template Bundlesmentioning

confidence: 99%

A unified global tractography framework for automatic visual pathway reconstruction

Yao

Zeng

et al. 2023

NMR in Biomedicine

Self Cite

View full text Add to dashboard Cite

show abstract

“…They targeted the entire anterior visual pathway and did not separate both ONs and chiasm. 26 To accurately quantify the ON and its disease-related changes, it is necessary to differentiate the ON from the surrounding CSF rather than segmenting them as a single structure. Few studies have aimed at isolating the ON from surrounding CSF to obtain precise quantitative measures of the ON itself.…”

Section: Introductionmentioning

confidence: 99%

“…Li et al. 26 proposed a parallel stages network composed of two 2D U-Nets for segmentation of the entire anterior visual pathway, employing T1-weighted MR images and fractional anisotropy images for feature extraction. They targeted the entire anterior visual pathway and did not separate both ONs and chiasm.…”

Section: Introductionmentioning

confidence: 99%

“…Dai et al 25 recently developed a regional convolutional neural network for segmenting OARs on MRI, incorporating several advanced architectures, such as a deep attention feature pyramid network and mask scoring networks, but their ON segmentation performance was low compared with other organs included in the study. Li et al 26 proposed a parallel stages network composed of two 2D U-Nets for segmentation of the entire anterior visual pathway, employing T1-weighted MR images and fractional anisotropy images for feature extraction. They targeted the entire anterior visual pathway and did not separate both ONs and chiasm.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Automatic segmentation and quantification of the optic nerve on MRI using a 3D U-Net

van Elst,

de Bloeme,

Noteboom

et al. 2023

J. Med. Imag.

View full text Add to dashboard Cite

Purpose: Pathological conditions associated with the optic nerve (ON) can cause structural changes in the nerve. Quantifying these changes could provide further understanding of disease mechanisms. We aim to develop a framework that automatically segments the ON separately from its surrounding cerebrospinal fluid (CSF) on magnetic resonance imaging (MRI) and quantifies the diameter and cross-sectional area along the entire length of the nerve.Approach: Multicenter data were obtained from retinoblastoma referral centers, providing a heterogeneous dataset of 40 high-resolution 3D T2-weighted MRI scans with manual ground truth delineations of both ONs. A 3D U-Net was used for ON segmentation, and performance was assessed in a tenfold cross-validation (n ¼ 32) and on a separate test-set (n ¼ 8) by measuring spatial, volumetric, and distance agreement with manual ground truths. Segmentations were used to quantify diameter and cross-sectional area along the length of the ON, using centerline extraction of tubular 3D surface models. Absolute agreement between automated and manual measurements was assessed by the intraclass correlation coefficient (ICC). Results:The segmentation network achieved high performance, with a mean Dice similarity coefficient score of 0.84, median Hausdorff distance of 0.64 mm, and ICC of 0.95 on the test-set. The quantification method obtained acceptable correspondence to manual reference measurements with mean ICC values of 0.76 for the diameter and 0.71 for the cross-sectional area. Compared with other methods, our method precisely identifies the ON from surrounding CSF and accurately estimates its diameter along the nerve's centerline. Conclusions:Our automated framework provides an objective method for ON assessment in vivo.

show abstract

Deep Learning Methods for Identification of White Matter Fiber Tracts: Review of State-of-the-Art and Future Prospective

2023

View full text Add to dashboard Cite

Two Parallel Stages Deep Learning Network for Anterior Visual Pathway Segmentation

Cited by 7 publications

References 23 publications

A unified global tractography framework for automatic visual pathway reconstruction

A unified global tractography framework for automatic visual pathway reconstruction

Automatic segmentation and quantification of the optic nerve on MRI using a 3D U-Net

Deep Learning Methods for Identification of White Matter Fiber Tracts: Review of State-of-the-Art and Future Prospective

Contact Info

Product

Resources

About