Abstract
Gastric cancer, a significant global health concern, exhibits high morbidity and mortality, especially in advanced stages. Timely diagnosis and intervention are crucial for improving patient outcomes, with Endoscopic Submucosal Dissection (ESD) playing a pivotal role in precise, minimally invasive early-stage treatments. Despite its importance, challenges include significant interobserver variability among pathologists and the intensive labor required for detailed pathological analysis of ESD specimens impede optimal outcomes. Artificial Intelligence (AI) offers promising solutions to these challenges, yet its advancement is limited by the scarcity of comprehensive, annotated pathological datasets. In this paper, we curate a fully annotated pathology slide dataset for ESD specimen examination. This dataset not only poses a challenging task for the computational pathology field but also enables precise detection of precancerous stages and accurate quantification of lesion distribution in patients with early-stage gastric lesions. Furthermore, it enhances the correlation between endoscopic findings and pathological interpretations, thereby advancing precision medicine strategies in the prevention and treatment of early gastric cancer.