Scholarlink

<p>Leishmaniases is a parasitic disease caused by the <i>Leishmania</i> parasite, transmitted by sandflies, affecting millions worldwide. Microscopic examination remains the standard method for detecting and quantifying intracellular parasite burden in leishmaniases research and Drug Discovery. This process is time-consuming and requires specific expertise. While Artificial Intelligence shows promise in automating this task, progress is limited by the lack of annotated datasets. To address this gap, we present AIR-LEISH, a dataset of 180 Giemsa-stained microscopic images with expert annotations containing 8,140 <i>Leishmania</i> amastigotes and 1511 macrophages. Images corresponded to samples from two infection models. The dataset was annotated to facilitate AI-based object detection and image segmentation tasks. We further demonstrated the potential of this dataset through training and testing two state-of-the-art architectures, namely YOLOv8 and U-Net. Both models demonstrated promising performance for automatic classification, detection and counting of amastigotes. The dataset is freely available on the Zenodo platform to accelerate the development of AI-based tools, facilitate advances in leishmaniases research and support collaborative initiatives for public health.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

AIR-LEISH: A Dataset of Giemsa-Stained Microscopy Images for AI-based Leishmania amastigotes Detection

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月2日

查看全文
  • Rafeh Oualha,
  • Nesrine Fekih-Romdhane,
  • Donia Driss,
  • Yosser Zina Abdelkrim,
  • Ikram Guizani,
  • Emna Harigua-Souiai

摘要

Leishmaniases is a parasitic disease caused by the Leishmania parasite, transmitted by sandflies, affecting millions worldwide. Microscopic examination remains the standard method for detecting and quantifying intracellular parasite burden in leishmaniases research and Drug Discovery. This process is time-consuming and requires specific expertise. While Artificial Intelligence shows promise in automating this task, progress is limited by the lack of annotated datasets. To address this gap, we present AIR-LEISH, a dataset of 180 Giemsa-stained microscopic images with expert annotations containing 8,140 Leishmania amastigotes and 1511 macrophages. Images corresponded to samples from two infection models. The dataset was annotated to facilitate AI-based object detection and image segmentation tasks. We further demonstrated the potential of this dataset through training and testing two state-of-the-art architectures, namely YOLOv8 and U-Net. Both models demonstrated promising performance for automatic classification, detection and counting of amastigotes. The dataset is freely available on the Zenodo platform to accelerate the development of AI-based tools, facilitate advances in leishmaniases research and support collaborative initiatives for public health.