IndraLab

Statements



sparser
"Specifically, we present a new method called PET-CAM for learning segmentation at the pixel level with class labels (at the image level) using the class response map, to make a CNN capable of segmenting the tumor at the pixel level but without pixel labels."

sparser
"PET-CAM tends to focus on the region of interest while CAM without prior knowledge (constraint) focuses also on other regions."

sparser
"PET-CAM model enables 3D lung and esophageal lesions segmentation from only two 2D MIP images."