Transformer model
A neural network architecture that reads a whole sequence at once instead of one step at a time. It uses attention to weigh how much every part of the input matters to every other part, which lets it train in parallel and handle long-range context that older recurrent networks lost track of. Transformers are the architecture behind modern large language models and most current medical-imaging foundation models.
Two vendors describe their products as transformer-based. One reads chest X-rays and returns findings from a fixed list, the other drafts the free-text report, and they fail in different ways: the first can miss or mislabel a finding, while the second can also write a fluent sentence about something absent from the image. Ask what the product actually outputs, whether a label, a score, a contour or free text, and design the local validation around that output and its typical failures.
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 →