Morphological Descriptors as Tool for Characterization of Nuclear Pleomorphism in Breast Cancer

Abstract

In recent years, the advancement in molecular pathology and genetic analysis of cancerous tissues has significantly remarked an increase in objective and measurable data. However, the traditional approach of morphological analysis in pathology diagnosis remains subjective in comparison, despite the introduction of digital pathology aiding computer-aided diagnosis. Certain pathological grading features, such as nuclear pleomorphism in breast cancer, still depend on a pathologist's expertise and are largely non-quantitative. This study aimed to investigate morphological descriptors as key elements to characterize the qualitative description of nuclear pleomorphism in breast cancer, in line with the Nottingham Histopathology Grading (NHG) system. Four morphological descriptors were extracted from segmented nuclear cells, including area, minimum ferret diameter, minor axis length and perimeter and used to assign scores of 1 to 3 to characterize pleomorphic nuclei. The proposed method was validated using the support vector machine (SVM), achieving promising results with 95.0% and 92.0% in accuracy (Acc) and F1 score (F1), respectively. This study serves as a pilot investigation for the quantitative measurement of nuclear pleomorphism in breast cancer.