arXiv subjects

Xiaomeng Li

Publications and source records attributed to Xiaomeng Li.

167 records · Page 10Linked to original sources

arXiv2018

Semi-supervised Skin Lesion Segmentation via Transformation Consistent Self-ensembling Model

Automatic skin lesion segmentation on dermoscopic images is an essential component in computer-aided diagnosis of melanoma. Recently, many fully supervised deep learning based methods have been proposed for automatic skin lesion segmentation. However, these approaches require massive pixel-wise annotation from experienced dermatologists, which is very costly and time-consuming. In this paper, we present a novel semi-supervised method for skin lesion segmentation by leveraging both labeled and unlabeled data. The network is optimized by the weighted combination of a common supervised loss for labeled inputs only and a regularization loss for both labeled and unlabeled data. In this paper, we present a novel semi-supervised method for skin lesion segmentation, where the network is optimized by the weighted combination of a common supervised loss for labeled inputs only and a regularization loss for both labeled and unlabeled data. Our method encourages a consistent prediction for unlabeled images using the outputs of the network-in-training under different regularizations, so that it can utilize the unlabeled data. To utilize the unlabeled data, our method encourages the consistent predictions of the network-in-training for the same input under different regularizations. Aiming for the semi-supervised segmentation problem, we enhance the effect of regularization for pixel-level predictions by introducing a transformation, including rotation and flipping, consistent scheme in our self-ensembling model. With only 300 labeled training samples, our method sets a new record on the benchmark of the International Skin Imaging Collaboration (ISIC) 2017 skin lesion segmentation challenge. Such a result clearly surpasses fully-supervised state-of-the-arts that are trained with 2000 labeled data.

cs.CV↗

arXiv2018

Deeply Supervised Rotation Equivariant Network for Lesion Segmentation in Dermoscopy Images

Automatic lesion segmentation in dermoscopy images is an essential step for computer-aided diagnosis of melanoma. The dermoscopy images exhibits rotational and reflectional symmetry, however, this geometric property has not been encoded in the state-of-the-art convolutional neural networks based skin lesion segmentation methods. In this paper, we present a deeply supervised rotation equivariant network for skin lesion segmentation by extending the recent group rotation equivariant network~\cite{cohen2016group}. Specifically, we propose the G-upsampling and G-projection operations to adapt the rotation equivariant classification network for our skin lesion segmentation problem. To further increase the performance, we integrate the deep supervision scheme into our proposed rotation equivariant segmentation architecture. The whole framework is equivariant to input transformations, including rotation and reflection, which improves the network efficiency and thus contributes to the segmentation performance. We extensively evaluate our method on the ISIC 2017 skin lesion challenge dataset. The experimental results show that our rotation equivariant networks consistently excel the regular counterparts with the same model complexity under different experimental settings. Our best model achieves 77.23\%(JA) on the test dataset, outperforming the state-of-the-art challenging methods and further demonstrating the effectiveness of our proposed deeply supervised rotation equivariant segmentation network. Our best model also outperforms the state-of-the-art challenging methods, which further demonstrate the effectiveness of our proposed deeply supervised rotation equivariant segmentation network.

cs.CV↗

arXiv2018

H-DenseUNet: Hybrid Densely Connected UNet for Liver and Tumor Segmentation from CT Volumes

Liver cancer is one of the leading causes of cancer death. To assist doctors in hepatocellular carcinoma diagnosis and treatment planning, an accurate and automatic liver and tumor segmentation method is highly demanded in the clinical practice. Recently, fully convolutional neural networks (FCNs), including 2D and 3D FCNs, serve as the back-bone in many volumetric image segmentation. However, 2D convolutions can not fully leverage the spatial information along the third dimension while 3D convolutions suffer from high computational cost and GPU memory consumption. To address these issues, we propose a novel hybrid densely connected UNet (H-DenseUNet), which consists of a 2D DenseUNet for efficiently extracting intra-slice features and a 3D counterpart for hierarchically aggregating volumetric contexts under the spirit of the auto-context algorithm for liver and tumor segmentation. We formulate the learning process of H-DenseUNet in an end-to-end manner, where the intra-slice representations and inter-slice features can be jointly optimized through a hybrid feature fusion (HFF) layer. We extensively evaluated our method on the dataset of MICCAI 2017 Liver Tumor Segmentation (LiTS) Challenge and 3DIRCADb Dataset. Our method outperformed other state-of-the-arts on the segmentation results of tumors and achieved very competitive performance for liver segmentation even with a single model.

cs.CV↗

arXiv2018

Extremal functions for Adams' inequalities in dimension four

Let $Ω\subset \mathbb{R}^4$ be a smooth bounded domain, $W_0^{2,2}(Ω)$ be the usual Sobolev space. For any positive integer $\ell$, $λ_{\ell}(Ω)$ is the $\ell$-th eigenvalue of the bi-Laplacian operator. Define $E_{\ell}=E_{λ_1(Ω)}\oplus E_{λ_2(Ω)}\oplus\cdots\oplus E_{λ_{\ell}(Ω)}$, where $E_{λ_i(Ω)}$ is eigenfunction space associated with $λ_i(Ω)$. $E^{\bot}_{\ell}$ denotes the orthogonal complement of $E_\ell$ in $W_0^{2,2}(Ω)$. For $0\leqα<λ_{\ell+1}(Ω)$, we define a norm by $\|u\|_{2,α}^{2}=\|Δu\|^2_2-α\|u\|^2_2$ for $u\in E^\bot_{\ell}$. In this paper, using the blow-up analysis, we prove the following Adams inequalities $$\sup_{u\in E_{\ell}^{\bot},\,\| u\|_{2,α}\leq 1}\int_Ωe^{32π^2u^2}dx<+\infty;$$ moreover, the above supremum can be attained by a function $u_0\in E_{\ell}^{\bot}\cap C^4(\overlineΩ)$ with $\|u_0\|_{2,α}=1$. This result extends that of Yang (J. Differential Equations, 2015), and complements that of Lu and Yang (Adv. Math. 2009) and Nguyen (arXiv: 1701.08249, 2017).

math.AP↗

arXiv2016

Extremal functions for singular Trudinger-Moser inequalities in the entire Euclidean space

In a previous work (Int. Math. Res. Notices 13 (2010) 2394-2426), Adimurthi-Yang proved a singular Trudinger-Moser inequality in the entire Euclidean space $\mathbb{R}^N$ $(N\geq 2)$. Precisely, if $0\leq β<1$ and $0<γ\leq1-β$, then there holds for any $τ>0$, $$\sup_{u\in W^{1,N}(\mathbb{R}^N),\,\int_{\mathbb{R}^N}(|\nabla u|^N+τ|u|^N)dx\leq 1}\int_{\mathbb{R}^N}\frac{1}{|x|^{Nβ}}\left(e^{α_Nγ|u|^{\frac{N}{N-1}}}-\sum_{k=0}^{N-2}\frac{α_N^kγ^k|u|^{\frac{kN}{N-1}}} {k!}\right)dx<\infty,$$ where $α_N=Nω_{N-1}^{1/(N-1)}$ and $ω_{N-1}$ is the area of the unit sphere in $\mathbb{R}^N$. The above inequality is sharp in the sense that if $γ>1-β$, all integrals are still finite but the supremum is infinity. In this paper, we concern extremal functions for these singular inequalities. The regular case $β=0$ has been considered by Li-Ruf (Indiana Univ. Math. J. 57 (2008) 451-480) and Ishiwata (Math. Ann. 351 (2011) 781-804). We shall investigate the singular case $0<β<1$ and prove that for all $τ>0$, $0<β<1$ and $0<γ\leq 1-β$, extremal functions for the above inequalities exist. The proof is based on blow-up analysis.

math.AP↗