Transfer learning
Reusing what a model learned on one task as the starting point for another, by loading the pre-trained weights into a downstream model. It is how the knowledge from a large pre-training run is carried into a specific clinical task with limited labelled data.
A screening program wants to grade diabetic retinopathy on photos from a new handheld fundus camera, and has only a few hundred labeled images from it. The team starts from a model already trained on a large set of desk-mounted camera images, so it begins knowing what vessels, hemorrhages and exudates look like, and the small set only has to adapt it to the new device. Ask how the handheld images differ from the source set in field of view and quality, and ask for results on held-out handheld images graded by your own readers.
Terms like this come up in real clinical scenarios across the HelloAI courses: bite-sized modules with verifiable certificates. An account takes one minute, no password needed.
Sign in →