In medical image processing, the most important information is often located on small parts of the image. Patch-based approaches aim at using only the most relevant parts of the image. Finding ways to automatically select the patches is a challenge. In this paper, we investigate two criteria to choose patches: entropy and a spectral similarity criterion. We perform experiments at different levels of patch size. We train a Convolutional Neural Network on the subsets of patches and analyze the training time. We find that, in addition to requiring less preprocessing time, the classifiers trained on the datasets of patches selected based on entropy converge faster than on those selected based on the spectral similarity criterion and, furthermore, lead to higher accuracy. Moreover, patches of high entropy lead to faster convergence and better accuracy than patches of low entropy.

翻译：在医学图像处理中,最重要的信息往往位于图像的一小部分。基于补丁法只着眼于使用图像中最相关的部分。 寻找自动选择补丁的方法是一项挑战。 在本文中,我们调查了选择补丁的两个标准: 酶和光谱相似性标准。 我们在不同的补丁大小水平上进行实验。 我们在补丁子子组上培训进化神经网络并分析培训时间。 我们发现,除了需要较少的预处理时间外, 分类人员还接受过关于根据酶聚合标准选择的补丁数据集的培训,比根据光谱相似性标准选择的补丁要快, 并且还导致更高的准确性。 此外, 高酶导的补丁比低酶的补丁要更快和准确性。