Artificial Intelligence Resource, Molecular Imaging Branch, National Cancer Institute, Bethesda, Maryland, USA.
Radiology and Imaging Sciences, Clinical Center, National Institutes of Health, Bethesda, Maryland, USA.
Med Phys. 2023 Aug;50(8):5020-5029. doi: 10.1002/mp.16303. Epub 2023 Mar 13.
von Hippel-Lindau syndrome (VHL) is an autosomal dominant hereditary syndrome with an increased predisposition of developing numerous cysts and tumors, almost exclusively clear cell renal cell carcinoma (ccRCC). Considering the lifelong surveillance in such patients to monitor the disease, patients with VHL are preferentially imaged using MRI to eliminate radiation exposure.
Segmentation of kidney and tumor structures on MRI in VHL patients is useful in lesion characterization (e.g., cyst vs. tumor), volumetric lesion analysis, and tumor growth prediction. However, automated tasks such as ccRCC segmentation on MRI is sparsely studied. We develop segmentation methodology for ccRCC on T1 weighted precontrast, corticomedullary, nephrogenic, and excretory contrast phase MRI.
We applied a new neural network approache using a novel differentiable decision forest, called hinge forest (HF), to segment kidney parenchyma, cyst, and ccRCC tumors in 117 images from 115 patients. This data set represented an unprecedented 504 ccRCCs with 1171 cystic lesions obtained at five different MRI scanners. The HF architecture was compared with U-Net on 10 randomized splits with 75% used for training and 25% used for testing. Both methods were trained with Adam using default parameters ( ) over 1000 epochs. We further demonstrated some interpretability of our HF method by exploiting decision tree structure.
The HF achieved an average kidney, cyst, and tumor Dice similarity coefficient (DSC) of 0.75 ± 0.03, 0.44 ± 0.05, 0.53 ± 0.04, respectively, while U-Net achieved an average kidney, cyst, and tumor DSC of 0.78 ± 0.02, 0.41 ± 0.04, 0.46 ± 0.05, respectively. The HF significantly outperformed U-Net on tumors while U-Net significantly outperformed HF when segmenting kidney parenchymas ( ).
For the task of ccRCC segmentation, the HF can offer better segmentation performance compared to the traditional U-Net architecture. The leaf maps can glean hints about deep learning features that might prove to be useful in other automated tasks such as tumor characterization.
von Hippel-Lindau 综合征（VHL）是一种常染色体显性遗传综合征，易发生多个囊肿和肿瘤，几乎全部为透明细胞肾细胞癌（ccRCC）。鉴于此类患者需要终生监测以监测疾病，因此优先使用 MRI 对 VHL 患者进行成像以避免辐射暴露。
在 VHL 患者的 MRI 上对肾脏和肿瘤结构进行分割对于病变特征（例如囊肿与肿瘤）、容积病变分析和肿瘤生长预测均具有重要意义。然而，MRI 上的 ccRCC 自动分割等自动化任务研究甚少。我们开发了用于 T1 加权对比前、皮质髓质、肾源性和排泄期 MRI 的 ccRCC 分割方法。
我们应用了一种新的神经网络方法，即一种新的可微分决策森林，称为铰链森林（HF），对 115 名患者的 117 张图像中的肾脏实质、囊肿和 ccRCC 肿瘤进行分割。该数据集前所未有地包含了在 5 台不同的 MRI 扫描仪上获得的 504 个 ccRCC 和 1171 个囊性病变。在 10 次随机拆分中，HF 架构与 U-Net 进行了比较，其中 75%用于训练，25%用于测试。两种方法均使用 Adam 默认参数（ ）在 1000 个 epoch 上进行训练。我们还通过利用决策树结构进一步证明了我们的 HF 方法的一些可解释性。
HF 在肾脏、囊肿和肿瘤的平均 Dice 相似系数（DSC）分别为 0.75±0.03、0.44±0.05、0.53±0.04，而 U-Net 分别为 0.78±0.02、0.41±0.04、0.46±0.05。HF 在肿瘤分割方面显著优于 U-Net，而 U-Net 在肾脏实质分割方面显著优于 HF（ ）。
在 ccRCC 分割任务中，HF 可提供比传统 U-Net 架构更好的分割性能。叶图可以提供有关深度学习特征的提示，这些特征可能在肿瘤特征等其他自动化任务中证明有用。
