Skip to main content
. Author manuscript; available in PMC: 2020 Sep 4.
Published in final edited form as: Phys Med Biol. 2019 Sep 4;64(17):175012. doi: 10.1088/1361-6560/ab326a

Fig. 1.

Fig. 1.

Illustration of the proposed SS-OLHF algorithm by fusing the highest level CNN representation learned at the output layer (OL) of a 3D CNN into the handcrafted features (HF). In the CNN architecture, Stage I mainly includes convolutional layers and pooling layers with tensor output; Stage Ⅱ mainly includes hidden fully connected layers with vector output; Stage Ⅲ is the output layer with vector output.