【摘要】 肝部病理组织切片传统的分类方法都是通过提取图像特征来进行识别和预测,由于图像特征不明显,且需要人工提取,受外在因素影响大,识别率较低,因此提出利用深度学习的卷积神经网络来进行识别分类。对轻量级模型进行改进,将样本图像直接作为输入数据,通过卷积神经网络训练验证即可得到实验结果,省去繁琐的特征提取环节。结果表明,改进后的轻量级神经网络验证准确率高达99.57%,明显高于当前的传统方法,且训练时间减少了十余个小时,方便快捷。更多还原

【Abstract】 The traditional classification of liver pathological tissue sections is to identify and predict by extracting image features.Because of the need to manually extract features,extraction has a certain degree of complexity and it will lead to the recognition rate of liver tissue sections is low.Therefore,the convolution neural network with deep learning is proposed to identify the classification.The lightweight model is improved,the sample image is directly used as input data,and the experimental results can be obtained by convolution neural network.There is no complicated feature extraction part.The results showed that the accuracy of the improved lightweight neural network was 99.57%,which was significantly higher than the current traditional method,and the training time was reduced by more than ten hours,which was convenient and quick.更多还原