The dataset was created to support reproducible artificial intelligence (AI) model development and benchmarking in papillary thyroid carcinoma (PTC) histopathology research. This dataset was generated as part of a broader research project focused on building and evaluating deep learning models for the classification of PTC. It addresses a specific gap in the availability of high-quality, image-level annotated histopathology datasets for thyroid cancer, especially those derived from underrepresented populations in Southeast Asia. The dataset is designed to be reusable for various AI downstream tasks, including classification, domain adaptation, stain normalization, and model interpretability in thyroid pathology. The PathoTiroid Dataset – Indonesian Collection (PTIC) dataset is a curated and expert-validated dataset of PTC histopathology images acquired from Indonesia’s national referral hospital. The dataset comprises 1,006 high-resolution images extracted from 46 whole-slide images (WSIs). A rigorous multistage validation process was employed to minimize annotation noise and enhance the dataset’s utility for supervised deep learning applications. The dataset is structured into two main classes corresponding to the PTC-like and non-PTC-like categories. This dataset aims to enhance the development and evaluation of AI-based models for PTC identification, serving as a valuable benchmark for comparative studies in computational pathology.
Shabrina et al. (Mon,) studied this question.