Automated quantification of MART1-verified Ki-67 indices: useful diagnostic aid in melanocytic lesions.

Abstract

The MART1-verified Ki-67 proliferation index is a valuable aid to distinguish melanomas from nevi. Because such indices are quantifiable by image analysis, they may provide a novel automated diagnostic aid. This study aimed to validate the diagnostic performance of automated dermal Ki-67 indices and to explore the diagnostic capability of epidermal Ki-67 in lesions both with and without a dermal component. In addition, we investigated the automated indices' ability to predict sentinel lymph node (SLN) status. Paraffin-embedded tissues from 84 primary cutaneous melanomas (35 with SLN biopsy), 22 melanoma in situ, and 270 nevi were included consecutively. Whole slide images were captured from Ki-67/MART1 double stains, and image analysis computed Ki-67 indices for epidermis and dermis. In lesions with a dermal component, the area under the receiver operating characteristic (ROC) curve was 0.79 (95% confidence interval [CI], 0.72-0.86) for dermal indices. By excluding lesions with few melanocytic cells, this area increased to 0.93 (95% CI, 0.88-0.98). A simultaneous analysis of epidermis and dermis yielded an ROC area of 0.94 (95% CI, 0.91-0.96) for lesions with a dermal component and 0.98 (95% CI, 0.97-1.0) for lesions with a considerable dermal component. For all lesions, the ROC area of the simultaneous analysis was 0.89 (95% CI, 0.85-0.92). SLN-positive patients generally had a higher index than SLN-negative patients (P ≤ .003). Conclusively, an automated diagnostic aid seems feasible in melanocytic pathology. The dermal Ki-67 index was inferior to a combined epidermal and dermal index in diagnosis but valuable for predicting the SLN status of our melanoma patients.