A Spitzoid Tumor dataset with clinical metadata and Whole Slide Images for Deep Learning models

Abstract

Spitzoid tumors (ST) are a group of melanocytic tumors of high diagnostic complexity. Since 1948, when Sophie Spitz first described them, the diagnostic uncertainty remains until now, especially in the intermediate category known as Spitz tumor of unknown malignant potential (STUMP) or atypical Spitz tumor. Studies developing deep learning (DL) models to diagnose melanocytic tumors using whole slide imaging (WSI) are scarce, and few used ST for analysis, excluding STUMP. To address this gap, we introduce SOPHIE: the first ST dataset with WSIs, including labels as benign, malignant, and atypical tumors, along with the clinical information of each patient. Additionally, we explain two DL models implemented as validation examples using this database.

Fig. 1

Representative patches extracted from WSIs presenting ST; (a-b): SN containing large uniform melanocytic cells devoid of mitotic activity in an organized fashion; (c-d) STUMP with deep mitosis; (e-f): SM with numerous mitotic figures and pagetoid spread.

Fig. 2

Learning curves for the source and target models. (a) Accuracy and loss for the source model trained at the patch level. (b) Accuracy and loss for the target model trained at the biopsy level.

Fig. 3

Technical validation experiments. (A) Data pre-processing carried out before training models. (B) Source model: ROI identification at patch-level; (C) Target model: slide-level classification using bags of instances, leveraging the pre-trained source model. Note that the methodological core is similar in both approaches.

Fig. 4

Normalized confusion matrix for the source and target models. (a) Confusion matrix for the source model trained at the patch level. (b) Confusion matrix for the target model trained at the biopsy level.