[论文解读] Deep Radiomics for Brain Tumor Detection and Classification from Multi-Sequence MRI

Subhashis Banerjee, Sushmita Mitra+2 more|arXiv (Cornell University)|Mar 21, 2019

Radiomics and Machine Learning in Medical Imaging参考文献 36被引用 40

一句话总结

该论文提出三种新颖的卷积神经网络模型（PatchNet、SliceNet、VolumeNet），从头开始训练以从多序列MRI检测并分类脑肿瘤（LGG vs HGG），并展示了对VGGNet/ResNet的迁移学习；VolumeNet在LOPO中达到97.19%的精度，在holdout数据上达到95%。

ABSTRACT

Glioma constitutes 80% of malignant primary brain tumors and is usually classified as HGG and LGG. The LGG tumors are less aggressive, with slower growth rate as compared to HGG, and are responsive to therapy. Tumor biopsy being challenging for brain tumor patients, noninvasive imaging techniques like Magnetic Resonance Imaging (MRI) have been extensively employed in diagnosing brain tumors. Therefore automated systems for the detection and prediction of the grade of tumors based on MRI data becomes necessary for assisting doctors in the framework of augmented intelligence. In this paper, we thoroughly investigate the power of Deep ConvNets for classification of brain tumors using multi-sequence MR images. We propose novel ConvNet models, which are trained from scratch, on MRI patches, slices, and multi-planar volumetric slices. The suitability of transfer learning for the task is next studied by applying two existing ConvNets models (VGGNet and ResNet) trained on ImageNet dataset, through fine-tuning of the last few layers. LOPO testing, and testing on the holdout dataset are used to evaluate the performance of the ConvNets. Results demonstrate that the proposed ConvNets achieve better accuracy in all cases where the model is trained on the multi-planar volumetric dataset. Unlike conventional models, it obtains a testing accuracy of 95% for the low/high grade glioma classification problem. A score of 97% is generated for classification of LGG with/without 1p/19q codeletion, without any additional effort towards extraction and selection of features. We study the properties of self-learned kernels/ filters in different layers, through visualization of the intermediate layer outputs. We also compare the results with that of state-of-the-art methods, demonstrating a maximum improvement of 7% on the grading performance of ConvNets and 9% on the prediction of 1p/19q codeletion status.

研究动机与目标

• 证明使用多序列MRI进行无创脑肿瘤检测与分级的深度学习的可行性，无需手动ROI/VOI分割。
• 开发三种CNN架构（PatchNet、SliceNet、VolumeNet），从头训练以处理基于patch、slice和volume的输入。
• 通过对VGGNet和ResNet在MRI数据上的微调进行迁移学习，评估性能提升。

提出的方法

• 引入三种数据表示：基于patch、基于slice，以及来自TCGA-GBM/TCGA-LGG与BraTS 2017数据集的多平面体积输入。
• 提出三种CNN架构——PatchNet、SliceNet、VolumeNet——从头训练；并与在ImageNet上微调的VGGNet和ResNet进行比较。
• 使用学习率0.001、动量0.9的SGD；数据增强（旋转、平移、翻转）； dropout（0.5）和批归一化以缓解过拟合。
• 使用LOPO（Leave-One-Patient-Out）和holdout测试进行评估；在LOPO中采用多数投票确定最终类别。
• research_questions
• （1）在无手动ROI/VOI delineation的条件下，CNNs能否从多序列MRI检测并分级胶质瘤（LGG vs HGG）？（2）将体积/多平面信息引入是否比基于patch或slice的输入能提高分类准确率？（3）在此任务中，ImageNet预训练模型的迁移学习与从头训练对MRI数据的表现相比有何差异？（4）CNNs能否从MRI预测LGG的1p/19q缺失状态？（5）CNNs如何在不同层次可视化所学特征以反映分级标准？
• key_findings
• ：[
• VolumeNet在HGG/LGG分类中实现了最高LOPO精度97.19%，且无歧义样本。
• VolumeNet在holdout数据上也达到LGG/HGG分类的95.00%准确率。
• SliceNet与PatchNet的LOPO精度分别为90.18%和84.91%，而VGGNet和ResNet在holdout数据上表现不及VolumeNet。
• 对于LGG有无1p/19q缺失状态，VolumeNet在holdout数据上达到97.00%的准确率，超过基于Mayo Clinic的参考方法（Akkus等人报道97%对比88%）。
• 微调的ImageNet模型（VGGNet、ResNet）在holdout数据上只能有限程度地达到从头训练的性能。
• 定性可视化显示所学滤波器逐步捕捉组织结构、肿瘤ROI和与分级相关的异质性特征。

实验结果

主要发现

网络	分类	错误分类	模糊	准确率
PatchNet	242	39	4	84.91%
SliceNet	257	26	2	90.18%
VolumeNet	277	8	0	97.19%
VGGNet	239	40	6	83.86%
ResNet	242	42	1	84.91%