Machine learning4 min read

Deep learning and neural networks

Deep neural networks, inspired by the brain, power today's imaging AI — and convolutional networks were designed for images.

A neural network is a stack of simple computing units, loosely inspired by neurons. Deep learning means many layers of these units, each learning progressively more abstract features — from edges and textures in early layers to organs and lesions in deep layers.

Convolutional neural networks (CNNs) were designed for images: they sweep across pixels looking for local patterns — edges, shapes, textures — in a translation-invariant, remarkably efficient way. CNNs and their successors are the workhorses of mammogram-reading AI.

Why deep learning suits images

  • Learns features automatically, from edges to organs
  • Handles the enormous dimensionality of images
  • Scales with data and computing power
  • Reaches expert-level performance across many medical tasks

In brief

Deep learning means many stacked neural layers.
CNNs were created specifically for image understanding.
Deep models reach expert-level accuracy in imaging.

Keep learning

What is machine learning?AI HubHow machine learning learnsAI HubMachine learning vs deep learningAI HubHow AI reads mammogramsAI Hub