Xiang Alec, Wang Fei
Horace Greeley High School, Chappaqua, New York.
Weill Cornell Medical College, New York City, New York.
AMIA Annu Symp Proc. 2020 Mar 4;2019:1246-1255. eCollection 2019.
Skin disease is a prevalent condition all over the world. Computer vision-based technology for automatic skin lesion classification holds great promise as an effective screening tool for early diagnosis. In this paper, we propose an accurate and interpretable deep learning pipeline to achieve such a goal. Comparing with existing research, we would like to highlight the following aspects of our model. 1) Rather than a single model, our approach ensembles a set of deep learning architectures to achieve better classification accuracy; 2) Generative adversarial network (GAN) is involved in the model training to promote data scale and diversity; 3) Local interpretable model-agnostic explanation (LIME) strategy is applied to extract evidence from the skin images to support the classification results. Our experimental results on real-world skin image corpus demonstrate the effectiveness and robustness of our method. The explainability of our model further enhances its applicability in real clinical practice.
皮肤病是一种在全球普遍存在的病症。基于计算机视觉的自动皮肤病变分类技术作为一种早期诊断的有效筛查工具，具有巨大的潜力。在本文中，我们提出了一种准确且可解释的深度学习流程来实现这一目标。与现有研究相比，我们想强调我们模型的以下几个方面。1）我们的方法不是单个模型，而是集成了一组深度学习架构以实现更好的分类准确率；2）生成对抗网络（GAN）参与模型训练以提升数据规模和多样性；3）应用局部可解释模型无关解释（LIME）策略从皮肤图像中提取证据以支持分类结果。我们在真实世界皮肤图像语料库上的实验结果证明了我们方法的有效性和鲁棒性。我们模型的可解释性进一步增强了其在实际临床实践中的适用性。
