A dataset of microscopic spirometra mansoni for medical image segmentation.

Accurate diagnosis of adult Spirometra mansoni infections remains challenging, due to the limited sensitivity and high cost associated with immunodiagnostic methods. Advances in computer vision suggest deep learning-based etiological image analysis could overcome these limitations. A barrier persists, however, in current computational parasitology research: the shortage of available datasets for training convolutional neural networks (CNNs) in adult S. mansoni identification. We present a microscopic imaging dataset specifically optimized for developing and validating semantic segmentation CNNs in S. mansoni detection to address this issue. This resource comprises 1420 microscopic images derived from fecal suspensions of confirmed infection cases, captured through standardized protocols using a Nikon BM-1000 optical microscope. All specimens were expert-annotated using LabelMe to generate segmentation masks delineating egg boundaries and distinctive morphological features. Furthermore, we conducted quantitative analyses on this dataset, including assessments of background entropy and the morphological entropy of individual S. mansoni eggs, along with investigation into their correlation. We evaluated segmentation architectures using this dataset to establish performance benchmarks, providing baselines for future methodological improvements. Spirometra Mansoni Dataset (SMD) is publicly available at https://www.heywhale.com/mw/dataset/684ff79484f1ba451ba76660/content under a CC BY 4.0 license.