Bilingual Text-to-Program Dataset for Quantitative Organoid Microscopy Analysis

This repository contains the derived dataset accompanying the manuscript entitled “A bilingual text-to-program workflow for quantitative organoid microscopy data.” The dataset was constructed based on publicly available bright-field organoid microscopy images introduced in the POST study. It contains bilingual natural-language instructions in English and Chinese, structured executable programs, instance-level morphological attributes, corresponding target masks, and dataset split information. The instructions describe quantitative organoid selection and removal tasks based on seven types of attributes: area, perimeter, circularity, ellipticity, image intensity, convexity, and spatial position. Both single-condition and multi-condition instructions are included, covering threshold comparisons, logical AND/OR combinations, keep/remove operations, and cases in which no organoid satisfies the specified conditions. The dataset is designed to support research on natural-language-guided organoid image analysis, text-to-program parsing, instruction-conditioned instance retrieval, and quantitative microscopy image segmentation. The structured programs provide an interpretable intermediate representation between natural-language instructions and image analysis operations. The original bright-field microscopy images are third-party data from the POST study and are not redistributed in this repository. Users should obtain the original images from the source study. The files provided in this repository contain the annotations, instructions, programs, masks, and related metadata generated or processed in the current study. Please cite both this dataset and the associated manuscript when using these resources.