Choosing a Reference Standard for Pathology AI
How many pathologists a reference standard needs depends on the intended use of the model, the expected disagreement between readers and the consequences of a wrong label.
Read the articleNotes from our work on independent central review, dataset quality and expert evaluation. The argument for why pathology expertise, not volume, is what makes clinical AI reliable.
How many pathologists a reference standard needs depends on the intended use of the model, the expected disagreement between readers and the consequences of a wrong label.
Read the articleIn a TCGA-derived oncology database, the specific surgical procedure was recovered from pathology reports for 1,930 of 9,202 cases where the public data had none or only a generic entry.
Read the articleIf you would like to hear about future insights, please submit your email address to enquiries@pathologynucleus.com.