基于SegNet网络的皮肤癌病变区域的自动分割
无需注册登录,支付后按照提示操作即可获取该资料.
基于SegNet网络的皮肤癌病变区域的自动分割(论文14000字,外文翻译)
摘要:随着深度学习的迅猛发展,医学影像领域也逐渐开始应用深度学习来辅助医生进行诊断和治疗。黑色素瘤是一种威胁人类健康的恶性肿瘤,但医生在皮肤镜下的肉眼观察存在主观性因素,因此实现病变区域的自动分割是后续提取特征对病症进行自动分类的重要前提,是皮肤科的迫切需要。本文利用SegNet网络实现了皮肤病变区域的自动分割,实验效果良好,测试集Dice系数到达0.84,训练集Dice系数达到0.88。
关键词:深度学习;SegNet网络;自动分割
Automatic segmentation of skin cancer lesions based on SegNet network
Abstract: With the rapid development of deep learning, the field of medical imaging has gradually begun to apply deep learning to assist doctors in diagnosis and treatment. Melanoma is a dangerous human health malignancy, but there are subjective factors observed by the doctor in the naked eye under the microscope, so to achieve automatic segmentation of the lesion area is an important prerequisite for the subsequent extraction of features to automatically classify the disease, is the Department of Dermatology urgent need. In this paper, the SegNet network is used to achieve the automatic segmentation of the skin lesion area. The experimental results are good. The Dice coefficient of the test set reaches 0.84, and the Dice coefficient of the training set reaches 0.88.
Key words: Deep Learning; SegNet Network; Automatic Segmentation
目录
1 绪论 1
1.1课题研究背景 1
1.2课题研究意义 2
1.3国内外研究现状和分析 2
1.4论文的主要目标 2
1.5论文的结构安排 3
2 深度学习与卷积神经网络 3
2.1卷积神经网络的发展 3
2.2相关基础 4
2.2.1神经网络 4
2.2.2激活函数 5
2.2.3局部感受野 5
2.2.4权值共享 6
2.2.5卷积 6
2.2.6池化 7
2.2.7损失层 7
2.2.8CNN结构 8
2.3全卷积神经网络(FCN) 8
2.4SegNet 10
2.5两种传统图像分割方法简述 11
2.5.1自适应阈值分割法 11
2.5.2区域生长分割法 12
3基于SegNet网络的病变区域自动分割 13
3.1研究动机 13
3.2实验环境 13
3.3实验数据准备 14
3.4利用SegNet进行分割 14
3.5实验结果 16
3.6利用两种传统方法进行对比实验 18
3.6.1自适应阈值方法进行分割实验 18
3.6.2区域生长方法进行分割实验 19
3.7实验评价 20
3.8实验分析 21
4总结与展望 21
4.1论文总结 21
4.2未来展望 22
参考文献 22
致谢 24