2D and 3D Vascular Structures Enhancement via Multiscale Fractional Anisotropy Tensor

Alhasson, Haifa F.; Alharbi, Shuaa S.; Obara, Bogusław

doi:10.1007/978-3-030-11024-6_26

Cited by 10 publications

(15 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our hybrid vesselness filter for vascular structures enhancement takes the advantages of vesselness enhancement diffusion, and integrates the improved Frangi's filter based on the ratio of eigenvalues of the Hessian matrix. The proposed method is mainly compared with Frangi's filter [14], Jerman's method [20] and multiscale fractional anisotropy tensor (MFAT) method [23]. The main reasons are summarized in the following aspects: 1) Frangi's filter is widely used, since it is easy to implement, and returns very high response uniformity on objects with uniform intensities.…”

Section: Resultsmentioning

confidence: 99%

“…The proposed method is found to achieve 96.96 ± 0.70, 3.04 ± 0.70, 5.39 ± 1.25 and 96.56 ± 0.94 percentage of TP, FN, FP, and OM rate for all 10 datasets, respectively. Moreover, low level of noise was added to the 2011 Vas-cuSynth Sample Data [30], [31], and the proposed method was tested compared with Jerman's method [20] and the Multiscale Fractional Anisotropy Tensor (MFAT) method [23]. The comparison results are presented in Table 3.…”

Section: Table 2 Segmentation Results On 2013 Vascusynth Sample (%)mentioning

confidence: 99%

“…Thus, FAT has great potential in medical images processing, like diffusion tensor regularization. For example, the method in [23] explores the feasibility of detecting vascular structures based on the Hessian matrix.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

2D and 3D Vascular Structures Enhancement Via Improved Vesselness Filter and Vessel Enhancing Diffusion

2019

View full text Add to dashboard Cite

In medical imaging practice, vascular enhancement filtering has been widely performed before vessel segmentation and centerline detection, which provides important pathological information and holds great significance for vessel quantification. In the literature, numerous well known vesselness filtering approaches have been developed. For example, some techniques explore the Hessian matrix of the original images and construct the vessel filter based on the eigenvalues of the Hessian matrix. In this work we develop a hybrid technique for fast and accurate vascular enhancement filter, which contains two main steps: vesselness diffusion and improved vesselness filter based on the eigenvalues ratio. This novel approach is quantitatively and qualitatively tested on the public 2D retinal datasets and 3D synthetic vascular structure models. Experimental results demonstrate that the proposed filter outperforms other existing approaches for curvilinear structure enhancement from noisy images. Moreover, the novel approach is further evaluated on real patient Coronary Computed Tomography Angiography (CCTA) datasets with ground truth regions labelled by professional cardiologist. Our method is proven to produce more accurate coronary artery segmentation results. Given the accuracy and efficiency, the proposed vesselness filter can be further used in medical practice for vascular structures enhancement before vessel segmentation and quantification.INDEX TERMS Vascular structures detection, improved enhancement filter, Hessian matrix, vesselness diffusion, vessel segmentation.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Table 2 Segmentation Results On 2013 Vascusynth Sample (%)mentioning

confidence: 99%

See 1 more Smart Citation

2D and 3D Vascular Structures Enhancement Via Improved Vesselness Filter and Vessel Enhancing Diffusion

2019

View full text Add to dashboard Cite

show abstract

“…To emulate the multiscale Hessian-based Vesselness approach used in phenoVein (Bühler et al, 2015), the MATLAB fibremetric implementation of the Frangi 'Vesselness' filter (Frangi et al, 1998) was used with vein thickness between 4 and 9 pixels, applied over four scales using a Gaussian image pyramid (Burt & Adelson, 1983) to cover the largest veins. We also tested improved Hessian-based enhancement techniques using Multiscale Fractional Anisotropy Tensors (MFAT), in both their eigenvalue-based (MFAT λ ) and probability-based (MFAT p ) form (Alhasson et al, 2019), and the intensity-independent, multiscale phase-congruency enhancement developed by Kovesi (Kovesi, 1999(Kovesi, , 2000, which was previously used to segment fungal (Obara et al, 2012), slime mould (Fricker et al, 2017) and ER networks (Pain et al, 2019). We used the normalised local weighted mean phase angle ('Feature Type') to give intensity-independent enhancement of vein structures, initially calculated over 3-5 scales and six orientations, and then applied to an image pyramid to cover larger scales.…”

Section: Comparison With Other Vein Segmentation Methodsmentioning

confidence: 99%

Automated and accurate segmentation of leaf venation networks via deep learning

Blonder

Jodra

et al. 2020

New Phytologist

View full text Add to dashboard Cite

Leaf vein network geometry can predict levels of resource transport, defence and mechanical support that operate at different spatial scales. However, it is challenging to quantify network architecture across scales due to the difficulties both in segmenting networks from images and in extracting multiscale statistics from subsequent network graph representations. Here we developed deep learning algorithms using convolutional neural networks (CNNs) to automatically segment leaf vein networks. Thirty-eight CNNs were trained on subsets of manually defined ground-truth regions from >700 leaves representing 50 southeast Asian plant families. Ensembles of six independently trained CNNs were used to segment networks from larger leaf regions (c. 100 mm 2). Segmented networks were analysed using hierarchical loop decomposition to extract a range of statistics describing scale transitions in vein and areole geometry. The CNN approach gave a precision-recall harmonic mean of 94.5% AE 6%, outperforming other current network extraction methods, and accurately described the widths, angles and connectivity of veins. Multiscale statistics then enabled the identification of previously undescribed variation in network architecture across species. We provide a LEAFVEINCNN software package to enable multiscale quantification of leaf vein networks, facilitating the comparison across species and the exploration of the functional significance of different leaf vein architectures.

show abstract

“…Hessian-based enhancement techniques using Multiscale Fractional Anisotropy Tensors (MFAT), in both their eigenvalue-based (MFATλ) and probability-based (MFATp) form (Alhasson et al, 2018), and the intensity-independent, multiscale phase-congruency enhancement developed by Kovesi (Kovesi, 1999;Kovesi, 2000), that we have previously used to segment fungal (Obara et al, 2012), slime mold (Fricker et al, 2017) and ER networks (Pain et al, 2019). We used the normalized local weighted mean phase angle ('Feature Type') to give intensity-independent enhancement of vein structures, initially calculated over 3-5 scales and 6 orientations, and then applied to an image pyramid to cover larger scales.…”

Section: Comparison With Other Vein Segmentation Methodsmentioning

confidence: 99%

Automated and accurate segmentation of leaf venation networks via deep learning

Blonder

Jodra

et al. 2020

Preprint

View full text Add to dashboard Cite

Leaf vein network geometry can predict levels of resource transport, defence, and mechanical support that operate at different spatial scales. However, it is challenging to quantify network architecture across scales, due to the difficulties both in segmenting networks from images, and in extracting multi-scale statistics from subsequent network graph representations. Here we develop deep learning algorithms using convolutional neural networks (CNNs) to automatically segment leaf vein networks. Thirty-eight CNNs were trained on subsets of manually-defined ground-truth regions from >700 leaves representing 50 southeast Asian plant families. Ensembles of 6 independently trained CNNs were used to segment networks from larger leaf regions (~100 mm2). Segmented networks were analysed using hierarchical loop decomposition to extract a range of statistics describing scale transitions in vein and areole geometry. The CNN approach gave a precision-recall harmonic mean of 94.5% ± 6%, outperforming other current network extraction methods, and accurately described the widths, angles, and connectivity of veins. Multi-scale statistics then enabled identification of previously-undescribed variation in network architecture across species. We provide a LeafVeinCNN software package to enable multi-scale quantification of leaf vein networks, facilitating comparison across species and exploration of the functional significance of different leaf vein architectures.

show abstract

2D and 3D Vascular Structures Enhancement via Multiscale Fractional Anisotropy Tensor

Cited by 10 publications

References 16 publications

2D and 3D Vascular Structures Enhancement Via Improved Vesselness Filter and Vessel Enhancing Diffusion

2D and 3D Vascular Structures Enhancement Via Improved Vesselness Filter and Vessel Enhancing Diffusion

Automated and accurate segmentation of leaf venation networks via deep learning

Automated and accurate segmentation of leaf venation networks via deep learning

Contact Info

Product

Resources

About