WLPolyp UCdb - sporedata/researchdesigneR GitHub Wiki
General description
The Wide-Light Polyp dataset is a publicly available dataset used primarily in the field of computer-aided diagnosis and medical imaging, specifically related to colorectal polyp detection and classification. It is intended to support the development of machine learning models, particularly in the context of colonoscopy image analysis through endoscopic images.
The dataset contains 1680 instances of 42 different polyps displaying colonic polyps and normal colonic mucosa recorded with a White Light (WL) video colonoscope by medical doctors from the Faculty of Medicine of the University of Coimbra, Portugal.
Data Categories
- Image data:
- Image Formats: The images are provided in formats like JPEG or PNG, which are suitable for processing by machine learning algorithms, particularly Convolutional Neural Networks (CNNs) used in computer vision tasks.
- Annotated Regions: Each image typically includes annotations marking the location of polyps, identifying their boundaries, and labeling the types of polyps (if the dataset provides polyp classification).
- Wide-Light Colonoscopy Images: The primary data in WLPolyp-UCdb consists of colonoscopy images captured under standard white-light conditions. These images show the interior of the colon and various types of polyps.
Limitations
- Imbalance of Classes: Often, there are fewer positive instances of polyps in the dataset compared to non-polyp regions. This class imbalance is a common challenge in medical imaging datasets.
- Variability in Polyps: Polyps come in various shapes, sizes, and appearances, making it difficult for a model to generalize well across all cases.
- Lighting and Image Quality: As colonoscopy images can be affected by lighting conditions, reflections, and motion, training robust models requires handling such variability.
Related publications
Data access
For more information on the WLPolyp-UCdb dataset, visit https://www.mat.uc.pt/%7Eisabelf/Polyp-UCdb/WLPolyp-UCdb.html