针对多形性腺瘤诊断完全依赖人工的问题,提出一种计算机辅助诊断方法.先通过采集数据并构建多形性腺瘤数据集,对当前稠密连接网络进行改进并融合通道注意力机制进行疾病组织分类特征提取,得到组织类别和概率,然后使用CART(classificatio...针对多形性腺瘤诊断完全依赖人工的问题,提出一种计算机辅助诊断方法.先通过采集数据并构建多形性腺瘤数据集,对当前稠密连接网络进行改进并融合通道注意力机制进行疾病组织分类特征提取,得到组织类别和概率,然后使用CART(classification and regression tree)进行推理学习,得到诊断结果.对难判断的类别选择进行人工辅助,进而实现对多形性腺瘤疾病的计算机辅助工作.实验结果表明,该方法在分类识别模块分类提取准确率达97.7%,决策树推理诊断准确率达100%.此外,分类识别模块在血细胞分类领域的准确率达98.6%.该方法具有一定的迁移性和有效性.展开更多
The effects of the occurrence and position of knots on modal parameter of Larix gmelinii was studied in this paper. The vibration signals are decomposed by wavelet package. The main results are as follows:the componen...The effects of the occurrence and position of knots on modal parameter of Larix gmelinii was studied in this paper. The vibration signals are decomposed by wavelet package. The main results are as follows:the component of high frequency of sample with knot was higher than that of sample without knot. With knots at different position,the component of high frequency of sample with knots at left is the maximum and that of sample with knots at right is the minimum. The knots of samples can be nondestructively tested based on vibration modal analysis.展开更多
文摘针对多形性腺瘤诊断完全依赖人工的问题,提出一种计算机辅助诊断方法.先通过采集数据并构建多形性腺瘤数据集,对当前稠密连接网络进行改进并融合通道注意力机制进行疾病组织分类特征提取,得到组织类别和概率,然后使用CART(classification and regression tree)进行推理学习,得到诊断结果.对难判断的类别选择进行人工辅助,进而实现对多形性腺瘤疾病的计算机辅助工作.实验结果表明,该方法在分类识别模块分类提取准确率达97.7%,决策树推理诊断准确率达100%.此外,分类识别模块在血细胞分类领域的准确率达98.6%.该方法具有一定的迁移性和有效性.
文摘The effects of the occurrence and position of knots on modal parameter of Larix gmelinii was studied in this paper. The vibration signals are decomposed by wavelet package. The main results are as follows:the component of high frequency of sample with knot was higher than that of sample without knot. With knots at different position,the component of high frequency of sample with knots at left is the maximum and that of sample with knots at right is the minimum. The knots of samples can be nondestructively tested based on vibration modal analysis.