https://github.com/ada-site-jml/medical-imaging
Science Score: 13.0%
This score indicates how likely this project is to be science-related based on various indicators:
-
○CITATION.cff file
-
✓codemeta.json file
Found codemeta.json file -
○.zenodo.json file
-
○DOI references
-
○Academic publication links
-
○Academic email domains
-
○Institutional organization owner
-
○JOSS paper metadata
-
○Scientific vocabulary similarity
Low similarity (7.5%) to scientific vocabulary
Repository
Basic Info
- Host: GitHub
- Owner: ADA-SITE-JML
- Language: Jupyter Notebook
- Default Branch: main
- Size: 1.14 MB
Statistics
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
- Releases: 0
Metadata Files
README.md
Classification of the breast cancer images using the keypoint regions stable against the image transformations
Abstract
Breast cancer remains a significant health concern, necessitating advanced and robust diagnostic methodologies. The unconventional representation of medical images (different formats with a dominant dark background) challenges the application of conventional approaches and pre-trained models designed for the smaller input size (usually not larger than 512x512), and the irrelevant input format (medical images typically use grayscale instead of expected tricolor model). This paper introduces a novel approach for the classification of breast cancer images, which solves two problems in the mentioned domain: the usage of the interest regions that are stable against transformations and the quality of the images provided to the machine learning models. A custom 2-dimensional convolutional neural network model adapted to multi-channel input accepts the top interest regions identified by a classical keypoint detection algorithm. To filter out arbitrary keypoints and retain only essential ones, we have considered the ranking of each keypoint based on its stability to image rotations as a decisive criterion. The rectangular regions with the highest number of keypoints are taken as channels of the input data. With the mentioned approach, the model preserves the image resolution, eliminating the risk of information loss while using the traditional image resizing. The preprocessing phase selects the most informative regions from the mammogram image and feeds the model with the part of the original image, which affects the efficient use of the computational resources. Experimental evaluations on image dataset demonstrate the effectiveness of the proposed approach in achieving good classification accuracy, even in the presence of high-variety of the provided images.
Owner
- Name: ADA-SITE-JML
- Login: ADA-SITE-JML
- Kind: organization
- Repositories: 2
- Profile: https://github.com/ADA-SITE-JML
GitHub Events
Total
- Push event: 1
Last Year
- Push event: 1