Fig. 1
Dataset collection and preview of randomly selected images. The dataset was collected following Hologic’s recommended guidelines for ThinPrep® Pap slide preparation. In particular, the minimally invasive broom and brush methods were used to collect cells from the cervix, which were then placed in Hologic’s PreservCyt solution. Slide preparation was automated using a ThinPrep® 5000 instrument. Slides were initially imaged and a diagnosis was made by 2 expert board-certified cytopathologists who were third parties to this work. The slides were subsequently deidentified by removing the barcode to be used as teaching slides. Another expert board certified cytopathologist then imaged each slide and captured multicellular images, seen in the “Dataset preview” section below. Finally, 2 separate board-certified cytopathologists then reviewed each image. Images were only included if all 3 experts agreed.
