Pancreas diabetes - wistar rats

# Pancreas-diabetes Dataset — Scientific Description ## 1. Overview The **Pancreas-diabetes** dataset is a curated collection of microscopic histological images of Wistar rats pancreatic tissue obtained at 100× magnification. Wistar Glucose intolerant rat females were compared with standard Wistar females with normal glucose tolerance (control). It is designed for supervised learning experiments in medical image analysis, with a focus on distinguishing normal tissue morphology from early signs of glucose intolerance. The dataset contains **90 RGB images**, each with a resolution of **1280 × 960 pixels**, divided into: - **58 images of normal pancreatic tissue**- **32 images of glucose-intolerant tissue** The dataset is intended for machine learning research, biomedical image classification, low-data learning studies, and educational use. It is released under the **Creative Commons Attribution 4.0 International (CC-BY-4.0)** license. --- ## 2. Scientific Motivation Pancreatic tissue morphology provides valuable insight into metabolic function and early pathological alterations associated with glucose intolerance and diabetes-related processes. Microscopy imaging, combined with machine learning, enables the identification of structural differences that may not be easily perceptible by manual inspection. This dataset supports research in: - biomedical image classification - early detection of metabolic dysregulation - analysis of tissue architecture using image-based features - transfer learning and few-shot learning in histology - evaluation of machine learning algorithms under limited sample size - explainable AI (XAI) applied to microscopy images --- ## 3. Image Acquisition and Conditions All images were acquired under controlled laboratory conditions with the following characteristics: - **Magnification:** 100× - **Microscopy:** Optical bright-field microscopy - **Image resolution:** 1280 × 960 pixels - **Color format:** RGB - **File formats:** JPEG or PNG - **Preparation:** Standard histological staining - **Digitization:** Camera-coupled microscope imaging system Each image contains a representative field of pancreatic tissue, including visible acinar structures, ducts, and surrounding morphology. Minor variations in illumination or texture reflect natural biological and technical variability. --- ## 4. Class Definition The dataset defines a **binary classification task**: | Class | Meaning | Number of Images ||--------------------|----------------------------------------------------------------------------------|-----|| `normal` | Healthy/normal Wistar rats pancreatic tissue | 58 || `intolerant` | Pancreatic tissue from glucose-intolerant Wistar rats | 32 | --- ## 5. File Structure The dataset is organized as follows: ```Pancreas-diabetes/├── images/│ ├── normal/│ └── intolerant/├── labels.csv├── DESCRIPTION.md├── README.md└── LICENSE``` ### 5.1 labels.csv The accompanying `labels.csv` file contains two columns: ```filepath,classnormal/img001.jpg,normal...intolerant/img032.jpg,intolerant``` This file defines image-level annotations, enabling immediate integration with machine learning pipelines. --- ## 6. Intended Use and Limitations ### Suitable for:- Benchmarking of supervised learning algorithms - CNNs and transfer learning experiments - Data augmentation research - Studies of low-data generalization - XAI/interpretability methods - Academic projects and coursework ### Not suitable for:- Clinical diagnosis - Human/clinical patient-level inference - Automated medical decision-making This dataset is experimental and provided solely for research and educational purposes. --- ## 7. Ethical and Legal Considerations - No personal or identifiable human data are included. - Samples originate from controlled laboratory experiments with non-human biological tissue. - The dataset complies with GDPR and related regulations. - Released under **CC-BY-4.0**, permitting reuse with appropriate attribution. --- ## 8. Authors and Institution - **Luís Rato** - **F.C. Silva** - **A.R. Costa** - **C. Antunes** **University of Évora, Portugal** --- ## 9. Citation Please cite this dataset as: > Rato L., Silva F.C., Costa A.R., Antunes C. (2025). *Pancreas-diabetes: Histology Image Dataset for Binary Classification* (Version 1.0). University of Évora. Released under CC-BY-4.0. > Rato, L., Silva, F.C., Costa, A.R., & Antunes, C.M. (2013). *Analysis of pancreas histological images for glucose intolerance identificationusing ImageJ—preliminary results*, Computational Vision and Medical Image Processing IV, VIPIMAGE 2013. CRC Press 2013, pp: 319–322.--- ## 10. Version and Availability - **Version:** 1.0 - **Release year:** 2025 - **License:** CC-BY-4.0 - **Permanent archive:** Zenodo DOI (to be added) - **Machine-learning-ready version:** OpenML dataset (to be added) ---