Brightfield Microscopy Image Dataset for Candida Species

OVERVIEW Invasive fungal infections present a critical nosocomial challenge, with global mortality burdens exceeding 1.5 million cases annually. This repository provides the high-fidelity microscopic dataset introduced in "Explainable Vision Transformers for Candida Species Identification from Brightfield Microscopy" (WCCI 2026). The collection focuses explicitly on the taxonomic differentiation of the dimorphic pathogen Candida albicans from the strictly blastoconidial, azole-resistant pathogen Candida glabrata directly from raw brightfield field-of-view (FOV) acquisitions. DATASET SPECIFICATIONS & TOTAL VOLUME The complete raw cohort archived in this download comprises N = 61 unique biological slides yielding a total of 3,763 non-overlapping microscopic patches: Candida albicans: 34 unique biological slides (2,088 total patches). Candida glabrata: 27 unique biological slides (1,675 total patches). The physical slide serves as the primary independent biological replicate to capture realistic inter-sample confounders, while individual image patches operate as dependent observational units. DATA PROVENANCE & PAPER REPLICATION NOTE Please note a minor structural distinction between this full raw archive and the experimental cohort reported in the WCCI 2026 publication (which notes 60 slides and 3,731 patches): During final pipeline execution, a single slide assembly (comprising 32 patches under the Candida albicans class) was filtered out to ensure a strict, clean intersection between parallel image formatting modalities (JPG/TIFF). To perfectly reproduce the exact experimental configurations, baseline accuracies (92.86% slide-level balanced accuracy), and explainable AI (XAI) manifold topologies detailed in the paper, users do not need to alter these raw directories manually. Instead, simply initialize the pipeline using the canonical split_indices.json document located within the project's official GitHub repository. This file automatically isolates the exact 60-slide cohort used during model auditing. ACQUISITION AND CULTURE PROTOCOL Culture Conditions: Reference strains (ATCC) were cultivated in Synthetic Complete (SC) medium (pH 6.5) at 35°C and standardized to an optical density (OD600) of 3.0. Cell Concentrations: Equal biomass measurements yielded standardized concentrations of 3.6 × 10^7 cells/mL for Candida albicans and 6 × 10^7 cells/mL for Candida glabrata. Mounting: Aliquots of 5 µL were fixed using a 1.7% agarose pad. Optical Setup: Imaging was performed using brightfield illumination (10% LED) on a Zeiss Axio Observer microscope equipped with a Plan-Apochromat 63x/1.4 Oil DIC objective. Sensor & Resolution: A Prime 95B monochrome camera containing a 12-bit sensor captured random fields of view under a fixed focus configuration, establishing an initial pixel spacing of 0.175 µm. FUNDING ACKNOWLEDGMENTS This work is funded by national funds through the Portuguese Recovery and Resilience Plan (PRR) through project C645008882-00000055, Center for Responsible AI. It was also supported by the FCT - Foundation for Science and Technology, I.P., within the scope of the research unit UID/00326 - Centre for Informatics and Systems of the University of Coimbra, and through the Research Grant with reference 2024.01388.BD.