Evaluation of Deep Learning Architectures for Complex Immunofluorescence Nuclear Image Segmentation
Evaluation of Deep Learning Architectures for Complex Immunofluorescence Nuclear Image Segmentation
复制标题
DOI：
10.1109/tmi.2021.3069558
复制
发表时间：
2021-07-01
影响因子：
10.6
通讯作者：
Taschner-Mandl, Sabine
中科院分区：
文献类型：
--
作者：
Kromp, Florian;Fischer, Lukas;Taschner-Mandl, Sabine
分离和标记每个核实例（实例感知分割）是核图像分割的关键挑战。深度卷积神经网络已被证明可以解决不同成像模式的核图像分割任务，但尚未对复杂的免疫荧光图像进行系统比较。基于深度学习的分割需要带注释的数据集进行训练，但带注释的荧光核图像数据集很少，而且大小和复杂性有限。在这项工作中，我们评估和比较了多种深度学习架构（U-Net、U-Net ResNet、Cellpose、Mask R-CNN、KG 实例分割）和两种传统算法（基于迭代 h-min 的分水岭、属性关系图）在各种类型的复杂荧光核图像上的分割效果。我们提出并评估了一种创建人工图像以扩展训练集的新颖策略。结果表明，实例感知分割架构和 Cellpose 在 F1 分数方面优于 U-Net 架构和复杂图像上的传统方法，而 U-Net 架构总体上实现了更高的平均 Dice 分数。使用额外的人工生成图像进行训练可以提高复杂图像的召回率和 F1 分数，从而使五分之三的样本制备类型获得最高的 F1 分数。在人工图像上训练的 Mask R-CNN 在与训练集图像条件相似的复杂图像上实现了总体最高的 F1 分数，而 Cellpose 在新成像条件的复杂图像上实现了总体最高的 F1 分数。我们提供的定量结果表明，本科生注释的图像足以训练实例感知的分割架构，以有效地分割复杂的荧光核图像。
Separating and labeling each nuclear instance (instance-aware segmentation) is the key challenge in nuclear image segmentation. Deep Convolutional Neural Networks have been demonstrated to solve nuclear image segmentation tasks across different imaging modalities, but a systematic comparison on complex immunofluorescence images has not been performed. Deep learning based segmentation requires annotated datasets for training, but annotated fluorescence nuclear image datasets are rare and of limited size and complexity. In this work, we evaluate and compare the segmentation effectiveness of multiple deep learning architectures (U-Net, U-Net ResNet, Cellpose, Mask R-CNN, KG instance segmentation) and two conventional algorithms (Iterative h-min based watershed, Attributed relational graphs) on complex fluorescence nuclear images of various types. We propose and evaluate a novel strategy to create artificial images to extend the training set. Results show that instance-aware segmentation architectures and Cellpose outperform the U-Net architectures and conventional methods on complex images in terms of F1 scores, while the U-Net architectures achieve overall higher mean Dice scores. Training with additional artificially generated images improves recall and F1 scores for complex images, thereby leading to top F1 scores for three out of five sample preparation types. Mask R-CNN trained on artificial images achieves the overall highest F1 score on complex images of similar conditions to the training set images while Cellpose achieves the overall highest F1 score on complex images of new imaging conditions. We provide quantitative results demonstrating that images annotated by under-graduates are sufficient for training instance-aware segmentation architectures to efficiently segment complex fluorescence nuclear images.