Xue Hao

Xue Hao

Auto-generated profile, based on publicly available information

Xue Hao

Faculty of Computing, Universiti Teknologi Malaysia (UTM), Johor, Malaysia

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
30 January 2024 in IEEE Access
Verified
Journal Article
Open access
QSQichen Su
Qichen Su
HHHaza Nuzly Abdull Hamed
Haza Nuzly Abdull Hamed
MIMOHD ADHAM ISA
MOHD ADHAM ISA
XHXue Hao
Xue Hao
XDXin Dai
Xin Dai
Skin cancer is one of the most common types of cancer globally. Despite the remarkable advancements of deep learning methods in computer vision, automatic diagnosis of skin diseases still faces challenges such as limited data and class imbalance. Generative Adversarial Networks (GANs), which can synthesize realistic data, appear as an alternative to mitigate these issues. However, for imbalanced data, unconditional GANs either generate uneven data distribution or neglect universal knowledge of the whole dataset, while state-of-the-art (SOTA) conditional GANs suffer from performance degradation due to the mode collapse of minority classes. This paper proposes a two-stage GAN-based method to synthesize fine-grained and diverse 256×256 pixels skin lesion images for the imbalanced dataset, named Self-Transfer GAN (STGAN). STGAN first learns universal knowledge from all classes then transfers this shared knowledge to each class and fuses it with class-specific knowledge to synthesize high-quality images. Furthermore, based on STGAN, a framework to enhance the classification performance is established. Both data generation and classification tasks are evaluated on HAM10000 dataset. In terms of Fréchet Inception Distance (FID), Inception Score, Precision, and Recall, STGAN improved by 16%, 16%, 4%, 33% compared with SOTA conditional StyleGAN2. For classification, the STGAN-based framework achieves remarkable results, with an Accuracy of 98.23%, Sensitivity of 88.85%, Precision of 90.23%, F1-score of 89.48%, and Specificity of 98.34%.

13 February 2025 in IEEE Access
Verified
Journal Article
Open access
XHXue Hao
Xue Hao
HHHaza Nuzly Abdull Hamed
Haza Nuzly Abdull Hamed
QSQichen Su
Qichen Su
XDXin Dai
Xin Dai
LDLinqiang Deng
Linqiang Deng
Coronary atherosclerotic plaques are a major cause of cardiovascular disease (CVD). Although there have been significant advancements in plaque-assisted diagnosis by deep learning methods utilizing intravascular optical coherence tomography (IVOCT), images still contain redundant features, including residual blood, thrombus, and extensive black backgrounds. Furthermore, the feature fusion process produces redundant information. These redundant features interfere with plaque feature extraction, resulting in decreased performance and increased computational complexity. Current methods for addressing redundant features rely mostly on simple feature selection or reduction methods, overlooking the significance of redundant features in model training. The Redundancy Feature Diminishment Strategy (RFDS) is designed to address this issue by dynamically adjusting the importance of redundant features during the training process. Meanwhile, RFDS is introduced into spatial and channel reconstruction convolution (SCConv) to form the enhancement structure RFDS-SCConv, which refines feature extraction by adaptively regulating redundant features in both spatial and channel dimensions. Based on RFDS-SCConv, a lightweight plaque detection framework, YOLO-Plaque, is proposed. RFDS-SCConv and FasterBlock are integrated into the framework to substantially reduce the interference of redundant features during feature extraction and fusion. This study evaluates the robustness of the YOLO-Plaque model by tenfold cross-validation utilizing 2504 IVOCT images collected from a collaborative project between the School of Software at Shanxi Agricultural University and Shanxi Provincial Coal Centre Hospital. Experimental results demonstrate that YOLO-Plaque performs better than existing plaque detection models while remaining lightweight. The vital metric mAP50 has increased by 3.59% reaching 91.23%, while GFLOPs and parameters decreased to 7.2 and 2.55M, respectively.

23 December 2024 in IEEE Access
Verified
Journal Article
Open access
XDXin Dai
Xin Dai
HHHaza Nuzly Abdull Hamed
Haza Nuzly Abdull Hamed
QSQichen Su
Qichen Su
XHXue Hao
Xue Hao
Frequent itemset mining (FIM) faces significant challenges with the expansion of large-scale datasets. Traditional algorithms such as Apriori, FP-Growth, and Eclat suffer from poor scalability and low efficiency when applied to modern datasets characterized by high dimensionality and high-density features. These algorithms demand substantial memory resources and multiple database scans, which diminishes their practicality for rapid data processing. To address these challenges, this study proposes the Inverted Index Automata Frequent Itemset Mining (IA-FIM) algorithm. IA-FIM integrates the swift retrieval of an inverted index with the robust pattern recognition of finite automata, enabling efficient processing of extensive datasets. Distinct from conventional FIM algorithms, IA-FIM utilizes an inverted index automata to efficiently reduce the search space and memory footprint, eliminating repetitive database scans and multiple tree constructions. The proposed algorithm employs a single-pass scan strategy, constructing a dynamic and adjustable inverted index for a streamlined and compact representation of data. IA-FIM demonstrates superior performance in processing large sparse dataset, enhancing the processing speed of large dataset and fulfilling the demands of the big data era. At the same time, it improves the efficiency and practicality of FIM by reducing repeated scans and large memory dependencies, making it more feasible when processing large dataset.