Xiaodan Deng

Dr. Xiaodan Deng


Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
6 January 2024 in Multimedia Tools and Applications
Verified
Journal Article
Closed access
PDF file
XDXiaodan Deng
Xiaodan Deng

23 September 2022
Verified
Conference Paper
Open access
SWShuo Wang
Shuo Wang
JZJunli Zhao
Junli Zhao
WLWeiwei Li
Weiwei Li
JTJieying Tang
Jieying Tang
XDXiaodan Deng
Xiaodan Deng
MZMingquan Zhou
Mingquan Zhou
The similarity evaluation of skull data is a research work of great significance and has been widely used in forensic, archaeology, and other fields. In this paper, we propose a 3D skull similarity evaluation method based on Siamese Network. Firstly, the original 3D skull data are mapped onto 2D depth images, which maintains the original information of the 3D data to the greatest extent. Then, we introduce the residual structure into the Siamese network to extract the main features of any pair of skull depth image data and map it to a new space for representation. Finally, a mapping function is proposed to calculate a reasonable similarity value of 0-1 between the two input skull data through the sigmoid layer. We conducted objective experiments on 3D skull models and compared them with the subjective skull similarity. The experimental results show that the 3D skull similarity evaluation method based on Siamese network with residual structure can effectively evaluate the similarity of skulls. The results are basically consistent with people's subjective evaluation, providing an effective direction for craniofacial reconstruction and identification.


24 August 2025 in Electronics
Verified
Journal Article
Open access
PDF file
PYPeng Yang
Peng Yang
ZCZhuochao Chen
Zhuochao Chen
XSXiaoxuan Sun
Xiaoxuan Sun
XDXiaodan Deng
Xiaodan Deng
Automatic skin lesion segmentation is essential for early melanoma diagnosis, yet the scarcity and limited diversity of annotated training data hinder progress. We introduce a two-stage framework that first employs a denoising diffusion probabilistic model (DDPM) enhanced with dilated convolutions and self-attention to synthesize unseen, high-fidelity dermoscopic images. In the second stage, segmentation models—including a dilated U-Net variant that leverages dilated convolutions to enlarge the receptive field—are trained on the augmented dataset. Experimental results demonstrate that this approach not only enhances segmentation accuracy across various architectures with an increase in DICE of more than 0.4, but also enables compact and computationally efficient segmentation models to achieve performance comparable to or even better than that of models with 10 times the parameters. Moreover, our diffusion-based data augmentation strategy consistently improves segmentation performance across multiple architectures, validating its effectiveness for developing accurate and deployable clinical tools.