Cervical whole slide images and annotations for automated reporting of cervical biopsies using artificial intelligence

The dataset comprises 2540 whole slide images of cervical cancer biopsies scanned using a Phillips Ultra Fast Scanner (UFS) with resolution equivalent to 40x or more specifically 0.25 microns/pixel, and stored in the isyntax file format. The dataset also contains detailed annotations for each slide specifying areas as malignant, high grade, low grade or normal/inflammation. Finally there is an index of slides, with the overall classification and sub -classifications for each slide and if they are part of training, validation or testing sets.

Other
This study uses archived samples from the NHS Greater Glasgow and Clyde Biorepository and Pathology Tissue Resource. Patients gave informed consent for surplus tissue to be stored and used for medical research, this consent was recorded in an electronic Surplus Tissue Authorisation form. All data was de-identified within the biorepository and as provided is fully anonymized.

Ethics approval for the study was granted by NHS Greater Glasgow and Clyde Biorepository and Pathology Tissue Resource (REC reference 16/WS/0207) on 4th April 2019.
Biorepository approval was obtained (application number 511)