||
本文为美国范德比尔特大学(作者:Justin Paul)的硕士论文,共29页。
为了学习复杂的模式,深度学习已经被成功地应用于有监督的分类任务中。本研究的目的是应用这一机器学习技术来分类不同类型肿瘤的大脑图像:脑膜瘤、神经胶质瘤和垂体。图像数据集包含233名患者,其中的3064幅脑图像为脑膜瘤、胶质瘤或垂体瘤。图像为轴面(横面)、冠状面(正面)或矢状面(侧面)的T1加权对比增强MRI(CE-MRI)图像。本研究以轴面图像为研究对象,并将此数据集扩展为无肿瘤大脑轴面图像的添加集合,以增加提供给神经网络训练的图像数量。通过这些数据对神经网络进行训练的结果,已经证明其分类是准确的,平均5倍交叉验证率为91.43%。
Deep learning has been used successfully insupervised classification tasks in order to learn complex patterns. The purposeof the study is to apply this machine learning technique to classifying imagesof brains with different types of tumors: meningioma, glioma, and pituitary.The image dataset contains 233 patients with a total of 3064 brain images witheither meningioma, glioma, or pituitary tumors. The images are T1- weightedcontrast enhanced MRI (CE-MRI) images of either axial (transverse plane),coronal (frontal plane), or sagittal (lateral plane) planes. This researchfocuses on the axial images, and expands upon this dataset with the addition ofaxial images of brains without tumors in order to increase the number of imagesprovided to the neural network. Training neural networks over this data hasproven to be accurate in its classifications an average five-fold crossvalidation of 91.43%.
更多精彩文章请关注公众号:
Archiver|手机版|科学网 ( 京ICP备07017567号-12 )
GMT+8, 2024-5-18 22:59
Powered by ScienceNet.cn
Copyright © 2007- 中国科学报社