Adversarial Content–Noise Complementary Learning Model for Image Denoising and Tumor Detection in Low-Quality Medical Images
Rattachement africain : Kenya, Rwanda, us. Niveau de preuve : code pays fourni par la source.
Le résumé fourni par la source
Medical imaging is crucial for disease diagnosis, but noise in CT and MRI scans can obscure critical details, making accurate diagnosis challenging. Traditional denoising methods and deep learning techniques often produce overly smooth images that lack vital diagnostic information. GAN-based approaches also struggle to balance noise removal and content preservation. Existing research has not explored tumor detection after image denoising; instead, it has concentrated on content and noise learning. To address these challenges, this study proposes the Adversarial Content–Noise Complementary Learning (ACNCL) model, which enhances image denoising and tumor detection. Unlike conventional methods focusing solely on content or noise learning, ACNCL simultaneously learns both through dual predictors, ensuring the complementary reconstruction of high-quality images. The model integrates multiple denoising techniques (DnCNN, U-Net, DenseNet, CA-AGF, and DWT) within a GAN framework, using PatchGAN as a local discriminator to preserve fine image textures. The ACNCL separates anatomical details and noise into distinct pathways, ensuring stable noise reduction while maintaining structural integrity. Evaluated on CT and MRI datasets, ACNCL demonstrated exceptional performance compared to traditional models both qualitatively and quantitatively. It exhibited strong generalization across datasets, improving medical image clarity and enabling earlier tumor detection. These findings highlight ACNCL’s potential to enhance diagnostic accuracy and support improved clinical decision-making.
Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.
Le contrôle bibliographique ouvert
DOI retrouvé dans Crossref DOI retrouvé ; titre concordant.
- Titre Crossref
- Adversarial Content–Noise Complementary Learning Model for Image Denoising and Tumor Detection in Low-Quality Medical Images
- Date Crossref
- 03/04/2025
- Éditeur
- MDPI AG
- Type
- journal-article
Ce recoupement confirme des métadonnées liées au DOI. Il ne confirme ni la méthode ni les conclusions de l’étude, et il ne compte pas comme une seconde source scientifique indépendante.
Où se fait cette recherche
-
Kisii University Kenya (code pays fourni par la source)Université ou école supérieure
-
Jomo Kenyatta University of Agriculture and Technology Kenya (code pays fourni par la source)Université ou école supérieure
-
Carnegie Mellon University Africa Department of Information and Communications Technology Kigali, Rwanda (code pays fourni par la source)Université ou école supérieure
-
Carnegie Mellon University pays non établi dans la noticeUniversité ou école supérieure
-
School of Information Science and Technology Department of Computing Sciences Kenya (pays nommé en fin d’affiliation)Université ou école supérieure
-
School of Computing and Information Technology Department of Computer Science Nairobi P.O. Box 62000-, Kenya (pays nommé en fin d’affiliation)Université ou école supérieure
Kisii University (Kenya), Jomo Kenyatta University of Agriculture and Technology (Kenya) et Department of Information and Communications Technology — Carnegie Mellon University Africa (Kigali, Rwanda), avec 3 autres affiliations. Pays d’affiliation : Kenya, Rwanda.
Une affiliation ne permet pas de déduire la nationalité d’un auteur.