This application focuses on the implementation of the pilot application on Medical Imaging Biomarkers. This radiomics approach includes a processing pipeline to extract frames from videos, classify them, select those frames with significative data, filter them and extract image features using first- and second-order texture analysis and image computing. Finally, that pipeline concludes a classification (normal, definite or borderline RHD).
- main: Main pipeline
- video_frame: Read video frame and detect doppler ones
- view_classification: Classify video frames into the different view classes:
- 0: 4 chamber
- 1: Parasternal Short Axis
- 2: Parasternal Long Axis
- doppler_segmentation: Apply a color-based segmentation to extract the colors from the doppler images using a k-means clustering
- texture_analysis: Perform first- and second-order texture analysis for image characterization and extraction of maximum blood velocities
- texture_classification: Conclude a classification (normal, borderline or definite RHD) according to the image features
- classifiers: includes files needed for the view classifier and for the textures classification
main.py -f videosfolder