Training data
The examples an AI model learns from. A model can only recognize patterns that were present in its training data, which is why gaps and bias in that data carry straight into the model.
An emergency department adds a fracture detection tool to its X-ray workflow, and the department also sees children. If the training data held mostly adult films, the model has rarely seen open growth plates or the buckle and greenstick fractures typical of young bones, and it may flag the first and miss the second. Check which ages the tool's intended use covers, ask what share of training images came from patients under 18, and keep pediatric films out of scope unless the intended use includes children and the tool holds up on your own pediatric cases.
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 →