Abstract:
Crafting effective deep learning models for medical image analysis is a complex task, particularly in cases where the medical image dataset lacks significant inter-class variation. This challenge is further aggravated when employing such datasets to generate synthetic images using generative adversarial networks (GANs), as the output of GANs heavily relies on the input data. In this research, we propose a novel filtering algorithm called Cosine Similarity-based Image Filtering (CosSIF). We leverage CosSIF to develop two distinct filtering methods: Filtering Before GAN Training (FBGT) and Filtering After GAN Training (FAGT). FBGT involves the removal of real images that exhibit similarities to images of other classes before utilizing them as the training dataset for a GAN. On the other hand, FAGT focuses on eliminating synthetic images with less discriminative features compared to real images used for training the GAN. The experimental results reveal that the utilization of either the FAGT or FBGT method reduces low inter-class variation in clinical image classification datasets and enables GANs to generate synthetic images with greater discriminative features. Moreover, modern transformer and convolutional-based models, trained with datasets that utilize these filtering methods, lead to less bias toward the majority class, more accurate predictions of samples in the minority class, and overall better generalization capabilities. Code and implementation details are available at: https://github.com/mominul-ssv/cossif.
摘要:
为医学图像分析构建有效的深度学习模型是一项复杂的任务,特别是在医学图像数据集缺乏显著的类间变化的情况下。当使用这样的数据集来使用生成对抗网络(GAN)生成合成图像时,这一挑战进一步加剧。因为GAN的输出在很大程度上依赖于输入数据。在这项研究中,我们提出了一种新颖的滤波算法,称为基于余弦相似度的图像滤波(CosSIF)。我们利用CosSIF开发了两种不同的过滤方法:GAN训练前过滤(FBGT)和GAN训练后过滤(FAGT)。FBGT涉及在将其用作GAN的训练数据集之前去除与其他类别的图像具有相似性的真实图像。另一方面,FAGT专注于消除与用于训练GAN的真实图像相比具有较少辨别特征的合成图像。实验结果表明,FAGT或FBGT方法的利用减少了临床图像分类数据集中的低类间变化,并使GAN能够生成具有更大区别特征的合成图像。此外,现代变压器和基于卷积的模型,用利用这些过滤方法的数据集进行训练,导致对多数阶层的偏见减少,对少数民族的样本进行更准确的预测,和整体更好的泛化能力。代码和实现详细信息可在以下网站获得:https://github.com/mominul-ssv/cossif。