Abstract
<jats:p>This pilot dataset presents an organized collection of digital cervical cytology slides designed to support research, teaching, and validation of artificial intelligence models for the analysis of cervical lesions. The resource includes original digitized slides, digitally cleaned versions, subslides derived from selected regions, and patches generated from those subslides. The samples are organized according to cytological diagnostic categories: NILM, LSIL, HSIL, and CSS/SCC, while preserving a hierarchical structure that allows each derived image to be traced back to its source slide. The dataset aims to provide a resource closer to real-world whole-slide cytology scenarios, unlike highly curated databases or datasets composed mainly of ideal cellular fields. In addition, the dataset incorporates anonymization criteria, traceable file naming, and basic technical documentation to facilitate academic reuse. This resource seeks to strengthen open science, experimental reproducibility, and external evaluation of computational models in cervical cytology.</jats:p>