DH-OOD: A decoupled hybrid framework for robust skin lesion classification via semantic-structural fusion
Journal of X-Ray Science and Technology · 2026-07 · 34 (4) · 541-551
DOI: 10.1177/08953996261441749 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of X-Ray Science and Technology, online erschienen am 10.06.2026. Verfasst von Benyuan He und 5 weiteren Autor:innen. Beteiligte Einrichtung: State Key Laboratory of Transducer Technology, Aerospace Information Research Institute (AIR), Chinese Academy of Sciences, Beijing, China. Hauptthemen laut MEDLINE: Skin Diseases. Die Arbeit nennt 36 Literaturangaben.
Abstract (Originalsprache)
Real-world skin lesion classification faces three major challenges: severe class imbalance, high intra-class variability, and the need to reject out-of-distribution (OOD) samples. Conventional monolithic models often struggle to address these issues simultaneously. To mitigate this limitation, we propose a multi-stage decoupled hybrid framework that combines Supervised Contrastive Learning (SupCon) with structural reconstruction. First, representation learning is decoupled using SupCon. Compared with standard cross-entropy training, SupCon alleviates feature degradation under long-tailed distributions by encouraging a more balanced feature space. Second, to address open-set recognition, we integrate contrastive semantic features with structural anomaly scores derived from an independent Convolutional Autoencoder (CAE). These complementary signals—semantic confidence and reconstruction error—are fused through a linear boundary formulation to support both known-class classification and unknown-sample rejection. On the ISIC 2019 dataset, the proposed framework achieves a Balanced Accuracy of 78.5% on known classes and improves the unknown-class F1-score to 51.3%. These results indicate that semantic–structural fusion enhances robustness under long-tailed and open-set conditions.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- State Key Laboratory of Transducer Technology, Aerospace Information Research Institute (AIR), Chinese Academy of Sciences, Beijing, China.
- School of Electronic, Electrical, and Communication Engineering, University of Chinese Academy of Sciences (UCAS), Beijing, China.
- Lingang Laboratory, Shanghai, China.
- Key Laboratory of Microsystems and Microstructure Manufacturing, Ministry of Education, Harbin Institute of Technology, Harbin, China.
Zitieren
APA
He, B., Yao, L., Xue, N., Liu, C., Liu, T., & Qi, Z. (2026). DH-OOD: A decoupled hybrid framework for robust skin lesion classification via semantic-structural fusion. Journal of X-Ray Science and Technology, 34(4), 541-551. https://doi.org/10.1177/08953996261441749
Vancouver
He B, Yao L, Xue N, Liu C, Liu T, Qi Z. DH-OOD: A decoupled hybrid framework for robust skin lesion classification via semantic-structural fusion. Journal of X-Ray Science and Technology. 2026;34(4):541-551. doi:10.1177/08953996261441749
BibTeX
@article{He2026,
author = {He, Benyuan and Yao, Lei and Xue, Ning and Liu, Chunxiu and Liu, Tiezhu and Qi, Zhimei},
title = {DH-OOD: A decoupled hybrid framework for robust skin lesion classification via semantic-structural fusion},
journal = {Journal of X-Ray Science and Technology},
year = {2026},
volume = {34},
number = {4},
pages = {541--551},
doi = {10.1177/08953996261441749}
}
EUVIMED