一种基于注意力的混合方法,使用CNN和BiLSTM来改进皮肤病变分类
Ayesha Shaik1,2, Shivanya Shomir Dutta3, Ishaan Milind Sawant3
1Centre for Cyber Physical Systems, Vellore Institute of Technology (VIT), Chennai, 600127, India. ayesha.sk@vit.ac.in.
Scientific reports
|May 5, 2025
概括
一个新的混合深度学习模型将卷积神经网络 (CNN) 与双向长短期记忆 (BiLSTM) 和注意力机制相结合,显著提高了皮肤病变分类的准确性.
科学领域:
- 皮肤病学 皮肤病学
- 人工智能的人工智能
- 医疗成像医学成像
背景情况:
- 皮肤病变构成了日益严重的全球健康挑战,需要提高诊断准确性以获得有效的治疗.
- 早期和精确检测皮肤病变对于改善患者的治疗结果和控制疾病进展至关重要.
研究的目的:
- 开发和评估一种混合深度学习模型,用于准确地分类皮肤病变.
- 通过先进的人工智能技术,提高各种皮肤病变的诊断能力.
主要方法:
- 卷积神经网络 (CNN) 与双向长期短期记忆 (BiLSTM) 网络的集成.
- 增强混合模型的空间,通道和时间注意力机制,以改善特征提取.
- 对各种深度学习架构进行比较分析,包括独立的CNN,InceptionV3,VGG16和Xception.
主要成果:
- 提出的具有注意力机制的CNN-BiLSTM模型实现了卓越的性能,精度为92.73%.
- 关键的性能指标包括精度 (92.84%),F1得分 (92.70%),回忆 (92.73%),贾卡德指数 (87.08%),子系数 (92.70%),MCC (91.55%).
- 混合模型的表现优于其他配置,证明了注意力增强方法的有效性.
结论:
- 开发的深度学习模型为医疗保健专业人员提供了强大的工具,提高了皮肤病变诊断的准确性.
- 这一进步支持主动管理策略和个性化治疗,有可能减轻皮肤病变的全球影响.
- 该研究有助于医疗诊断的技术进步,旨在通过及时干预来提高公共卫生的弹性.
更多相关视频
相关概念视频
01:30Skin Cancer
2.9K
Skin cancer is a type of cancer that occurs when there is an abnormal growth of skin cells, usually triggered by damage to the DNA within the skin cells. It is primarily caused by exposure to ultraviolet (UV) radiation from the sun or artificial sources like tanning beds. Skin cancer is the most common type of cancer worldwide, and its incidence continues to rise.
Basal Cell Carcinoma (BCC): BCC is the most common type of skin cancer, accounting for about 80% of cases. It typically develops in...
Basal Cell Carcinoma (BCC): BCC is the most common type of skin cancer, accounting for about 80% of cases. It typically develops in...
2.9K
01:12Renewal of Skin Epidermal Stem Cells
2.4K
The skin is divided into epidermis, dermis, and hypodermis, the skin's outermost, middle, and inner layers. The human epidermal layer regularly undergoes renewal, where old, dead cells are replaced by new cells. Epidermal stem cells or EpiSCs divide and differentiate to restore the lost cells. For the renewal process, some EpiSCs continuously self-renew. In contrast, few others differentiate into transit-amplifying cells, which later form prickle or spinous cells, followed by granular...
2.4K