Multi-Resolution Framework for Spitzoid Neoplasm Classification Using Histological Data

Spitzoid melanocytic tumors (SMTs) are a group of neoplasms that represent a formidable diagnostic challenge for experts. In daily practice, dermatopathologists examine tissue biopsies manually, which is very time-consuming. To perform an effective diagnosis, pathologists visualize the histological features of these tumors at different resolution levels, as the lesion contours (visualized at low level) and the morphology of the cells (at high level) are decisive. Aiming to mimic the daily practice of a dermatopathologist, in this paper, we present a multi-resolution framework to automatically assess morphological features at different resolution levels and combine them to provide a more accurate diagnosis. The experiments carried out demonstrate that the proposed method outperforms single-resolution frameworks in SMT classification.