Yang Qi, Yu Xin, Lee Ho Hin, Tang Yucheng, Bao Shunxing, Gravenstein Kristofer S, Moore Ann Zenobia, Makrogiannis Sokratis, Ferrucci Luigi, Landman Bennett A
Vanderbilt University, Department of Computer Science, Nashville, Tennessee, United States.
Vanderbilt University, Department of Electrical and Computer Engineering, Nashville, Tennessee, United States.
J Med Imaging (Bellingham). 2022 Sep;9(5):052405. doi: 10.1117/1.JMI.9.5.052405. Epub 2022 May 19.
: Muscle, bone, and fat segmentation from thigh images is essential for quantifying body composition. Voxelwise image segmentation enables quantification of tissue properties including area, intensity, and texture. Deep learning approaches have had substantial success in medical image segmentation, but they typically require a significant amount of data. Due to the high cost of manual annotation, training deep learning models with limited human label data is desirable, but it is a challenging problem. Inspired by transfer learning, we proposed a two-stage deep learning pipeline to address the thigh and lower leg segmentation issue. We studied three datasets, 3022 thigh slices and 8939 lower leg slices from the BLSA dataset and 121 thigh slices from the GESTALT study. First, we generated pseudo labels for thigh based on approximate handcrafted approaches using CT intensity and anatomical morphology. Then, those pseudo labels were fed into deep neural networks to train models from scratch. Finally, the first stage model was loaded as the initialization and fine-tuned with a more limited set of expert human labels of the thigh. : We evaluated the performance of this framework on 73 thigh CT images and obtained an average Dice similarity coefficient (DSC) of 0.927 across muscle, internal bone, cortical bone, subcutaneous fat, and intermuscular fat. To test the generalizability of the proposed framework, we applied the model on lower leg images and obtained an average DSC of 0.823. Approximated handcrafted pseudo labels can build a good initialization for deep neural networks, which can help to reduce the need for, and make full use of, human expert labeled data.
从大腿图像中进行肌肉、骨骼和脂肪分割对于量化身体成分至关重要。逐体素图像分割能够对包括面积、强度和纹理在内的组织特性进行量化。深度学习方法在医学图像分割方面取得了显著成功，但通常需要大量数据。由于手动标注成本高昂，使用有限的人工标注数据训练深度学习模型是可取的，但这是一个具有挑战性的问题。受迁移学习的启发，我们提出了一种两阶段深度学习管道来解决大腿和小腿分割问题。我们研究了三个数据集，来自BLSA数据集的3022个大腿切片和8939个小腿切片以及来自GESTALT研究的121个大腿切片。首先，我们基于使用CT强度和解剖形态的近似手工方法为大腿生成伪标签。然后，将这些伪标签输入深度神经网络从头开始训练模型。最后，加载第一阶段模型作为初始化，并使用更有限的一组大腿专家人工标签进行微调。我们在73张大腿CT图像上评估了该框架的性能，在肌肉、内部骨骼、皮质骨、皮下脂肪和肌间脂肪上获得的平均骰子相似系数（DSC）为0.927。为了测试所提出框架的通用性，我们将该模型应用于小腿图像，获得的平均DSC为0.823。近似的手工伪标签可以为深度神经网络构建良好的初始化，这有助于减少对人工专家标注数据的需求并充分利用这些数据。
