[论文解读] DONet: Dual Objective Networks for Skin Lesion Segmentation

Yaxiong Wang, Yunchao Wei+3 more|arXiv (Cornell University)|Aug 19, 2020

Cutaneous Melanoma Detection and Management参考文献 50被引用 18

一句话总结

该论文提出DONet，一种用于皮肤病变分割的双目标网络，通过使用两个对称解码器并分别优化不同损失函数，生成互补的概率图，从而提高分割的可靠性。该方法通过整合不同优化目标的预测结果，并采用循环上下文编码模块（RCEM）实现多尺度上下文特征学习，在ISIC 2018数据集上达到0.881的Dice分数，在PH²数据集上达到0.931的SOTA性能。

ABSTRACT

Skin lesion segmentation is a crucial step in the computer-aided diagnosis of dermoscopic images. In the last few years, deep learning based semantic segmentation methods have significantly advanced the skin lesion segmentation results. However, the current performance is still unsatisfactory due to some challenging factors such as large variety of lesion scale and ambiguous difference between lesion region and background. In this paper, we propose a simple yet effective framework, named Dual Objective Networks (DONet), to improve the skin lesion segmentation. Our DONet adopts two symmetric decoders to produce different predictions for approaching different objectives. Concretely, the two objectives are actually defined by different loss functions. In this way, the two decoders are encouraged to produce differentiated probability maps to match different optimization targets, resulting in complementary predictions accordingly. The complementary information learned by these two objectives are further aggregated together to make the final prediction, by which the uncertainty existing in segmentation maps can be significantly alleviated. Besides, to address the challenge of large variety of lesion scales and shapes in dermoscopic images, we additionally propose a recurrent context encoding module (RCEM) to model the complex correlation among skin lesions, where the features with different scale contexts are efficiently integrated to form a more robust representation. Extensive experiments on two popular benchmarks well demonstrate the effectiveness of the proposed DONet. In particular, our DONet achieves 0.881 and 0.931 dice score on ISIC 2018 and $ ext{PH}^2$, respectively. Code will be made public available.

研究动机与目标

• 解决由于病变尺度可变及病变-背景边界模糊导致的皮肤病变分割中高不确定性和模糊性问题。
• 通过利用不同优化目标生成的互补预测，而非依赖单一损失函数，提升分割的可靠性。
• 通过新颖的循环上下文编码模块（RCEM）增强多尺度上下文表征能力，以适应形状和大小各异的病变。
• 通过双解码器架构联合优化多个分割目标，在基准数据集上实现SOTA性能。

提出的方法

• DONet采用共享编码器，后接两个对称解码器，每个解码器使用不同的损失函数（如Dice损失和Focal Tversky损失）进行优化，实现不同的优化目标。
• 两个解码器分别生成独立的概率图，二者具有互补性，因为各自偏向于优化其对应的损失目标。
• 通过联合决策机制聚合两个概率图，生成最终更可靠的分割图，降低不确定性。
• 循环上下文编码模块（RCEM）利用卷积LSTM，通过扩张卷积（扩张率分别为1、2、4和8）迭代累积多尺度上下文特征。
• RCEM通过循环机制整合不同感受野的特征，提升对病变尺度变化的鲁棒性。
• 多尺度跳跃连接将RCEM中的上下文知识传递至解码路径，增强解码器中的特征表示能力。

实验结果

研究问题

• RQ1 通过使用不同损失函数训练两个解码器，能否通过生成互补预测来提升分割可靠性？
• RQ2 如何有效聚合多尺度上下文特征，以提升对病变尺度和形状变化的鲁棒性？
• RQ3 与简单的特征拼接相比，用于捕捉皮肤镜图像中长距离依赖关系的循环机制是否更具优势？
• RQ4 双目标架构与RCEM在分割性能上的贡献程度如何，二者是独立作用还是存在协同效应？

主要发现

• 在ISIC 2018数据集上，DONet的Dice分数达到0.881，比基线模型提升0.027分。
• 在PH²数据集上，DONet的Dice分数达到0.931，展现出优异的跨数据集泛化能力。
• 消融实验表明，双目标架构（DOA）和RCEM均具有显著贡献，其中DOA使Dice分数提升0.017，RCEM提升0.019。
• DOA与RCEM联合使用时性能最优，ISIC 2018数据集Dice分数达0.881，PH²数据集达0.931，表明二者存在协同效应。
• RCEM在四个时间步（扩张率1、2、4、8）时性能最佳，超过四步后收益递减。
• 可视化消融结果表明，DOA与RCEM均能提升边界贴合度并减少假阳性，尤其在小尺寸或不规则病变的挑战性病例中表现更优。