Min Yang

Min Yang

Auto-generated profile, based on publicly available information

Min Yang

School of Electronics and Information Engineering, Wuyi University, Jiangmen, China

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

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

Representative Publications

View All Publications
26 September 2024 in IEEE Transactions on Neural Systems and Rehabilitation Engineering
Verified
Journal Article
Open access
XXXiaoshan Xie
Xiaoshan Xie
MYMin Yang
Min Yang
Medical image segmentation is a crucial component of computer-aided clinical diagnosis, with state-of-the-art models often being variants of U-Net. Despite their success, these models’ skip connections introduce an unnecessary semantic gap between the encoder and decoder, which hinders their ability to achieve the high precision required for clinical applications. Awareness of this semantic gap and its detrimental influences have increased over time. However, a quantitative understanding of how this semantic gap compromises accuracy and reliability remains lacking, emphasizing the need for effective mitigation strategies. In response, we present the first quantitative evaluation of the semantic gap between corresponding layers of U-Net and identify two key characteristics: 1) The direct skip connection (DSC) exhibits a semantic gap that negatively impacts models’ performance; 2) The magnitude of the semantic gap varies across different layers. Based on these findings, we re-examine this issue through the lens of skip connections. We introduce a Multichannel Fusion Transformer (MCFT) and propose a novel USCT-UNet architecture, which incorporates U-shaped skip connections (USC) to replace DSC, allocates varying numbers of MCFT blocks based on the semantic gap magnitude at different layers, and employs a spatial channel cross-attention (SCCA) module to facilitate the fusion of features between the decoder and USC. We evaluate USCT-UNet on four challenging datasets, and the results demonstrate that it effectively eliminates the semantic gap. Compared to using DSC, our USC and SCCA strategies achieve maximum improvements of 4.79% in the Dice coefficient, 5.70% in mean intersection over union (MIoU), and 3.26 in Hausdorff distance.

25 December 2024 in IEEE Access
Verified
Journal Article
Open access
FZFudong Zhong
Fudong Zhong
CQChuanbo Qin
Chuanbo Qin
YFYue Feng
Yue Feng
JZJunying Zeng
Junying Zeng
XJXudong Jia
Xudong Jia
MYMin Yang
Min Yang
In Traditional Chinese Medicine (TCM) diagnostics, the appearance of the tongue is a crucial indicator of health. TCM practitioners traditionally assess the tongue’s shape, color, texture, and other features to aid diagnosis. With advancements in technology, digitizing and analyzing tongue images using computer has become possible, making tongue image segmentation an important step in realizing automated tongue diagnosis. While existing network models have shown good results, they often struggle with tongue images against complex backgrounds, particularly on resource-limited edge devices. To tackle this challenge, this paper introduces a novel, more efficient automated tongue image segmentation model (MEAT-SAM). MEAT-SAM leverages a lightweight large model, a first in the field of tongue image segmentation. Compared to previous models, MEAT-SAM reduces the overall model parameters, improves segmentation speed, and enables operation on more resource-constrained edge devices. Despite its efficiency, MEAT-SAM maintains performance close to the state-of-the-art (SOTA) even with complex tongue image backgrounds. Tested on three different datasets (TongueDataset01, TongueDataset02, TongueDataset03), MEAT-SAM achieved IoU of 96.63%, 95.58%, and 95.07%, demonstrating excellent generalization and robustness against various complex background conditions in tongue images. Furthermore, MEAT-SAM can run effectively on the computationally limited Jetson Nano single-board computer, achieving similar segmentation effects as in experimental testing.

17 February 2025 in IEEE Access
Verified
Journal Article
Open access
JLJia Li
Jia Li
LHLi Huang
Li Huang
HOHao Ouyang
Hao Ouyang
DCDingjian Cai
Dingjian Cai
SZShuoqi Zheng
Shuoqi Zheng
MYMin Yang
Min Yang
The precise segmentation of retinal vessel can facilitate more accurate evaluation of the health of subject by medical professionals. The raid advancement of deep learning technology has led to the remarkable performance of U-Net and its variant networks in retinal vessel segmentation tasks. However, the existing algorithms still face challenges in obtaining high-precision segmentation results, manifested in the complex and variable structure of the retinal vessel itself; the existing U-Net series networks have problems such as a lack of linguistic information, limited receptive field and insufficient learning capacity for small vessel.We propose a novel retinal vessel segmentation network, named multi-scale dense attention cascade network (MDAC-Net), which incorporates multi-scale feature fusion and co-optimization of multiple attention mechanisms to deal with the above challenges. 1) We propose a new inter-stage attention module (NIAM) to realize the information exchange between the subnetworks, which promote the fine particles segment subnetwork (FPSS) to pay more attention to small vessel region. 2) Considering the context information and vessel details, we adopt dilated convolution module (DCM) to replace the traditional convolution operation. 3) We integrate a multi-scale convolutional attention module (MBAM) in the subnetwork, which suppress noise and irrelevant information. 4) We use Grad-CAM to visualize the segmentation results, which can guide doctors to analyze the pathogenic factors of different subjects more accurately. The experimental results demonstrate that MDAC-Net, while remaining relatively lightweight, achieves accuracy of more than 97% on three public datasets, especially remarkable segmentation effects for small vessel.