ImViA Laboratory, University of Burgundy, Dijon, France.
Radiation Oncology Department, CGFL, Batiment I3M, 64b rue sully, 21000, Dijon, France.
Int J Comput Assist Radiol Surg. 2020 Sep;15(9):1437-1444. doi: 10.1007/s11548-020-02223-x. Epub 2020 Jul 11.
To achieve accurate image segmentation, which is the first critical step in medical image analysis and interventions, using deep neural networks seems a promising approach provided sufficiently large and diverse annotated data from experts. However, annotated datasets are often limited because it is prone to variations in acquisition parameters and require high-level expert's knowledge, and manually labeling targets by tracing their contour is often laborious. Developing fast, interactive, and weakly supervised deep learning methods is thus highly desirable.
We propose a new efficient deep learning method to accurately segment targets from images while generating an annotated dataset for deep learning methods. It involves a generative neural network-based prior-knowledge prediction from pseudo-contour landmarks. The predicted prior knowledge (i.e., contour proposal) is then refined using a convolutional neural network that leverages the information from the predicted prior knowledge and the raw input image. Our method was evaluated on a clinical database of 145 intraoperative ultrasound and 78 postoperative CT images of image-guided prostate brachytherapy. It was also evaluated on a cardiac multi-structure segmentation from 450 2D echocardiographic images.
Experimental results show that our model can segment the prostate clinical target volume in 0.499 s (i.e., 7.79 milliseconds per image) with an average Dice coefficient of 96.9 ± 0.9% and 95.4 ± 0.9%, 3D Hausdorff distance of 4.25 ± 4.58 and 5.17 ± 1.41 mm, and volumetric overlap ratio of 93.9 ± 1.80% and 91.3 ± 1.70 from TRUS and CT images, respectively. It also yielded an average Dice coefficient of 96.3 ± 1.3% on echocardiographic images.
We proposed and evaluated a fast, interactive deep learning method for accurate medical image segmentation. Moreover, our approach has the potential to solve the bottleneck of deep learning methods in adapting to inter-clinical variations and speed up the annotation processes.
在医学图像分析和干预中，实现精确的图像分割是至关重要的第一步，使用深度神经网络似乎是一种很有前途的方法，前提是有足够大和多样化的专家标注数据。然而，标注数据集通常是有限的，因为它容易受到采集参数的变化的影响，并且需要高级专家的知识，而手动跟踪目标轮廓进行标注通常是很繁琐的。因此，开发快速、交互式和弱监督的深度学习方法是非常需要的。
我们提出了一种新的有效的深度学习方法，可以在生成深度学习方法的标注数据集的同时，准确地从图像中分割目标。它涉及到基于生成神经网络的伪轮廓标记的先验知识预测。然后，使用卷积神经网络对预测的先验知识进行细化，该网络利用预测的先验知识和原始输入图像的信息。我们的方法在 145 例术中超声和 78 例图像引导前列腺近距离放射治疗术后 CT 图像的临床数据库上进行了评估。它还在 450 张 2D 超声心动图的心脏多结构分割上进行了评估。
实验结果表明，我们的模型可以在 0.499 秒内（即每张图像 7.79 毫秒）分割前列腺临床靶区，平均 Dice 系数为 96.9 ± 0.9%和 95.4 ± 0.9%，3D Hausdorff 距离为 4.25 ± 4.58 和 5.17 ± 1.41 毫米，体积重叠比为 93.9 ± 1.80%和 91.3 ± 1.70%，分别来自 TRUS 和 CT 图像。它还在超声心动图上获得了平均 Dice 系数为 96.3 ± 1.3%。
我们提出并评估了一种快速、交互式的医学图像分割深度学习方法。此外，我们的方法有可能解决深度学习方法适应临床间变化的瓶颈问题，并加速标注过程。
