Wu Eric, Hadjiiski Lubomir M, Samala Ravi K, Chan Heang-Ping, Cha Kenny H, Richter Caleb, Cohan Richard H, Caoili Elaine M, Paramagul Chintana, Alva Ajjai, Weizer Alon Z
Departments of Radiology.
Internal Medicine-Hematology/Oncology, and.
Tomography. 2019 Mar;5(1):201-208. doi: 10.18383/j.tom.2018.00036.
We compared the performance of different Deep learning-convolutional neural network (DL-CNN) models for bladder cancer treatment response assessment based on transfer learning by freezing different DL-CNN layers and varying the DL-CNN structure. Pre- and posttreatment computed tomography scans of 123 patients (cancers, 129; pre- and posttreatment cancer pairs, 158) undergoing chemotherapy were collected. After chemotherapy 33% of patients had T0 stage cancer (complete response). Regions of interest in pre- and posttreatment scans were extracted from the segmented lesions and combined into hybrid pre -post image pairs (h-ROIs). Training (pairs, 94; h-ROIs, 6209), validation (10 pairs) and test sets (54 pairs) were obtained. The DL-CNN consisted of 2 convolution (C1-C2), 2 locally connected (L3-L4), and 1 fully connected layers. The DL-CNN was trained with h-ROIs to classify cancers as fully responding (stage T0) or not fully responding to chemotherapy. Two radiologists provided lesion likelihood of being stage T0 posttreatment. The test area under the ROC curve (AUC) was 0.73 for T0 prediction by the base DL-CNN structure with randomly initialized weights. The base DL-CNN structure with pretrained weights and transfer learning (no frozen layers) achieved test AUC of 0.79. The test AUCs for 3 modified DL-CNN structures (different C1-C2 max pooling filter sizes, strides, and padding, with transfer learning) were 0.72, 0.86, and 0.69. For the base DL-CNN with (C1) frozen, (C1-C2) frozen, and (C1-C2-L3) frozen, the test AUCs were 0.81, 0.78, and 0.71, respectively. The radiologists' AUCs were 0.76 and 0.77. DL-CNN performed better with pretrained than randomly initialized weights.
我们基于迁移学习，通过冻结不同的深度学习卷积神经网络（DL-CNN）层并改变DL-CNN结构，比较了不同DL-CNN模型对膀胱癌治疗反应评估的性能。收集了123例接受化疗患者（癌症病灶129个，治疗前后癌症病灶对158对）治疗前和治疗后的计算机断层扫描图像。化疗后，33%的患者癌症处于T0期（完全缓解）。从分割后的病灶中提取治疗前和治疗后扫描图像中的感兴趣区域，并组合成混合治疗前-后图像对（h-ROIs）。获得了训练集（图像对94对，h-ROIs 6209个）、验证集（10对）和测试集（54对）。DL-CNN由2个卷积层（C1-C2）、2个局部连接层（L3-L4）和1个全连接层组成。使用h-ROIs对DL-CNN进行训练，以将癌症分类为对化疗完全缓解（T0期）或未完全缓解。两名放射科医生提供了治疗后病灶为T0期的可能性。对于权重随机初始化的基础DL-CNN结构，T0预测的ROC曲线下面积（AUC）为0.73。具有预训练权重和迁移学习（无冻结层）的基础DL-CNN结构的测试AUC为0.79。3种修改后的DL-CNN结构（不同的C1-C2最大池化滤波器大小、步长和填充，采用迁移学习）的测试AUC分别为0.72、0.86和0.69。对于冻结（C1）、冻结（C1-C2）和冻结（C1-C2-L3）的基础DL-CNN，测试AUC分别为0.81、0.78和0.71。放射科医生的AUC分别为0.76和0.77。与权重随机初始化相比，DL-CNN在使用预训练权重时表现更好。
