Unstructured data
Information that does not fit neatly into rows and columns: free-text notes, images, waveforms, audio. Roughly 80% of healthcare data is unstructured, and it is both the richest and the hardest for AI to use.
A hospital wants to track every incidental lung nodule that radiologists recommend following up, and those recommendations exist only as sentences in free-text reports. A tool that extracts them has to handle negations such as 'no nodule seen', hedges such as 'possible nodule, likely vessel', and each radiologist's own phrasing, any of which a careless extractor can turn into a wrong entry. Before trusting its list, have a radiologist read a few hundred reports against the tool's output and count both misses and false entries.
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 →