基于多模态MRI与深度学习的乳腺病变良恶性鉴别
杨一风,祁章璇,聂生东

Differentiation of Benign and Malignant Breast Lesions Based on Multimodal MRI and Deep Learning
Yi-feng YANG,Zhang-xuan QI,Sheng-dong NIE
表3 AC_Ulsam_CNN网络结构参数设置
Table 3 The parameter setting of the proposed AC_Ulsam_CNN structure
#输入层输入尺寸卷积核尺寸通道数步长输出尺寸
(1)Conv1+BN+LeakyReLU150 × 150 × 33 × 3161150 × 150 × 16
(2)Conv2+BN+LeakyReLU150 × 150 × 161 × 181150 × 150 × 8
(3)Conv3+BN+LeakyReLU150 × 150 × 83 × 381150 × 150 × 8
(4)Conv4+BN+LeakyReLU150 × 150 × 81 × 1161150 × 150 × 16
(5)Conv5+BN+LeakyReLU150 × 150 × 161 × 1161150 × 150 × 16
(6)Add1150 × 150 × 16---150 × 150 × 16
(7)Pool1150 × 150 × 163 × 3-350 × 50 × 16
(8)Conv650 × 50 × 161 × 18150 × 50 × 8
(9)AC Block150 × 50 × 8-8-50 × 50 × 16
(10)Conv750 × 50 × 161 × 132150 × 50 × 32
(11)Conv850 × 50 × 161 × 132150 × 50 × 32
(12)Add250 × 50 × 32---50 × 50 × 32
(13)Pool250 × 50 × 323 × 3-316 × 16 × 32
(14)Conv916 × 16 × 321 × 132116 × 16 × 32
(15)AC Block216 × 16 × 32-32-16 × 16 × 64
(16)Conv1016 × 16 × 641 × 1128116 × 16 × 128
(17)Conv1150 × 50 × 323 × 316316 × 16 × 16
(18)Conv1216 × 16 × 321 × 1128116 × 16 × 128
(19)Add316 × 16 × 128---16 × 16 × 128
(20)Concatenate116 × 16 × 128
16 × 16 × 16
---16 × 16 × 144
(21)Ulsam Block116 × 16 × 144---16 × 16 × 144
(22)Ulsam Block216 × 16 × 144---16 × 16 × 144
(23)Concatenate216 × 16 × 144
16 × 16 × 144
---16 × 16 × 288
(24)Conv13+BN+LeakyReLU16 × 16 × 2881 × 1128116 × 16 × 128
-GAP16 × 16 × 128---4096
-Dense4096---16
-Sigmoid16---1