谷歌浏览器插件
订阅小程序
在清言上使用

Bladder Tumor Grading and Staging Prediction of Magnetic Resonance Imaging Based on Transfer Learning

Proceedings of the 2019 2nd International Conference on Algorithms, Computing and Artificial Intelligence(2019)

引用 0|浏览0
暂无评分
摘要
Bladder cancer ranks fourth in male malignant tumors and eighth in mortality. Early detection of bladder tumors is important for preventing bladder cancer, reducing mortality, and improving patients' quality of life. In this paper, we used the deep learning model such as Resnet_v2_50 and Inceptionv3 to predict Bladder tumor grading and staging. When the model was trained and tested, the experiment data set coming from the Chinese University Computer Design Competition - Big Data and Artificial Intelligence Challenge, was used. The test set which accounted for 10% of the data set was randomly disrupted. First, the method of transfer learning was used to train the corresponding network where the convolutional layer parameters were fixed and the fully connected layer was retrained. Then for this data set, we compared the impact of three improved strategies on its classification accuracy. The three improved strategies were that: 1. Doing data augmentation by randomly changing brightness and contrast of images and then added the changed images to the training set. 2. Features extracted by Resnet_v2_50 and Inception_v3 were stitched before they were sent into the fully connected layer. 3. Combining the previous two methods. The experiments results showed that for the grading prediction, the inception_v3 model with data augmentation had the highest classification accuracy of 95.5%; for the staging prediction, stitching models based on the inception_v3 and Resnet_v2_50 without data augmentation had the highest classification accuracy of 92.8%. It implied the effectiveness of these two improved strategies (feature stitching and data augmentation) for different classification tasks. However, it may not work well if the two methods are used at the same time.
更多
查看译文
关键词
bladder tumor grading,magnetic resonance imaging,staging prediction,transfer
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要