Skip to main content

Table 3 The architecture of the self-defined dSPIC network

From: dSPIC: a deep SPECT image classification network for automated multi-disease, multi-lesion diagnosis

Layer

Configuration

Convolution

11 × 11, 96, S = 4, P  = 2

Pooling

MaxPool, 3 × 3, S = 2

Added layer

Attention/residual module

Convolution

5 × 5, 256, S = 1, P  = 2

Pooling

MaxPool, 3 × 3, S = 2

Convolution

3 × 3, 384, S = 1, P  = 3

Convolution

3 × 3, 384, S = 1,   P = 3

Convolution

3 × 3, 256, S = 1,   P = 1

Fully connected

4096

Fully connected

4096

SoftMax

3

  1. S stride, P padding, MaxPool max pooling