Invasive lobular carcinoma (ILC) is almost always classified as Nottingham histological grade 2. Despite this, prognosis is markedly varied, with some ILCs behaving more akin to grade 3 invasive ductal carcinoma (IDC). Methods to separate these aggressive ILCs are needed. Digital image analysis (DIA) of the Ki-67 biomarker has potential in this regard; thus, we sought to determine the feasibility of its use for automated evaluation of ILC. An initial pilot study demonstrated no ILC specific changes were required to our Ki-67 DIA algorithm for reproducible results. Subsequently, 42 consecutive cases of ILC were evaluated by visual mitosis counting in H&E stained sections and by DIA on Ki-67 stained sections. Ki-67 proliferative index (PI) DIA showed significant correlation with visual mitosis counting on H&E stained sections (rs=0.63; p<0.05), significant strong correlation (rs=0.78; p<0.05) and substantial agreement (κ=0.62) with manual/visual Ki-67 assessment and significant positive associations with grade, nodal status and 'pleomorphic' ILC subtype, and a wide stratification of values in classical/grade 2 ILC. In conclusion, DIA of Ki-67 PI in ILC is feasible, correlates with mitotic index, manual/visual Ki-67 PI and clinico-pathological variables. The broad stratification of Ki-67 PI in classical/grade 2 ILC supports its practicability as a biomarker with prognostic and predictive potential, although large studies with outcome data are required for validation.
Keywords: Breast neoplasms; cell proliferation; computer assisted diagnosis; image processing.
Crown Copyright © 2019. Published by Elsevier B.V. All rights reserved.