Cervical cell lift: A novel triage method for the spatial mapping and grading of precancerous cervical lesions

Abstract

Background: Primary HPV screening, due to its low specificity, requires an additional liquid-based cytology (LBC) triage test. However, even with LBC triage there has been a near doubling in the number of patients referred for colposcopy in recent years, the majority having low-grade disease.

Methods: To counter this, a triage test that generates a spatial map of the cervical surface at a molecular level has been developed which removes the subjectivity associated with LBC by facilitating identification of lesions in their entirety. 50 patients attending colposcopy were recruited to participate in a pilot study to evaluate the test. For each patient, cells were lifted from the cervix onto a membrane (cervical cell lift, CCL) and immunostained with a biomarker of precancerous cells, generating molecular maps of the cervical surface. These maps were analysed to detect high-grade lesions, and the results compared to the final histological diagnosis.

Findings: We demonstrated that spatial molecular mapping of the cervix has a sensitivity of 90% (95% CI 69-98) (positive predictive value 81% (95% CI 60-92)) for the detection of high-grade disease, and that AI-based analysis could aid disease detection through automated flagging of biomarker-positive cells.

Interpretation: Spatial molecular mapping of the CCL improved the rate of detection of high-grade disease in comparison to LBC, suggesting that this method has the potential to decisively identify patients with clinically relevant disease that requires excisional treatment.

Funding: CRUK Early Detection Project award, Jordan-Singer BSCCP award, Addenbrooke's Charitable Trust, UK-MRC, Janssen Pharmaceuticals/Advanced Sterilisation Products, and NWO.

Keywords: Biomarker for cervical screening; CIN; Cervical cancer; Cytology; HPV; Methodology for cervical screening; Non-invasive sampling; Spatial mapping of lesion; Triage test.

Figure 1

Scanned image of the cervical cell lift membrane. Cell lift membrane stained with DAPI. Near-uniform coverage of surface cervical keratinocytes. (b) Higher magnification image of cell lift membrane, highlighting cervical keratinocyte nuclei (white arrows) and multi-lobed neutrophil nuclei (red arrows). (c) H&E staining of normal keratinocytes (white arrow) and neutrophils (red arrow) on the membrane.

Figure 2

Immunostaining of the cervical epithelium with keratin markers to assess cell type. Cross section of the cervical epithelium stained with K13 (green) and K19 (red). (b) Colposcopy photograph of the cervix following application of cell lift membrane. Acetowhite solution has been applied, revealing the large transformation zone (inside the green dotted line). (c) Demarcation of ectocervix (left) and transformation zone (right) boundary, K13 (green) and K19 (red) positive cells. (d) Cervical transformation zone. K13 (green) and K19 (red) positive cells. White arrows indicate K13/K19 dual positive cells.

Figure 3

MCM biomarker staining of cervical cell lifts. (a and b) Cell lift membrane from a patient with a confirmed CIN3 lesion, stained with MCM (green) and DAPI (blue). An entire cervical cell lift is shown (a). Three MCM positive HSIL/CIN3 lesions (red insets) are shown in (b). (c) Heat maps representing patient cell lift samples, showing MCM positive image fields (pink) and MCM negative image fields (green). The samples are three confirmed CIN3 cases.

Figure 4

Cervical cell lift pilot study results. Scatter plot of the number of MCM positive image fields in the HSIL vs. LSIL/negative cases (n=50). (b) ROC curve demonstrating an overall accuracy of this method of 88%, with a sensitivity of 89% to HSIL.

Figure 5

Evaluation of machine learning for MCM positive cells in patient samples. (a) and (b) CNN analysis of an MCM positive image field. The scanned image (a) was manually annotated for MCM positive cells and analysed by the trained CNN (b). The CNN output was then compared to the manual annotation. The CNN detected most of the annotated cells (264 matches vs. 89 non-matches), and no abnormal cells were missed. The non-matches were artefacts that the CNN deemed positive. (c) The algorithm was further evaluated on a set of 10 new image fields. The graph compares CNN to manual annotation of MCM positive cells. A linear fit of the data points results in a line with a slope of 1.1, indicating close similarity between manual and CNN annotation.

Figure 6

Evaluation of machine learning for cell lift image analysis. (a) and (b) Cell lift membrane stained with MCM and DAPI (a) assessed by the algorithm (b). (c-e) Scanned image fields (red insets) and corresponding CNN output image showing a lesion (b-d) and non-specific staining (d and e).