Improve the diagnostic, prognostic, and predictive accuracy of medical images in clinical decision-making. Medical images are mineable data. We apply computer vision and pattern recognition algorithms to medical images to support the detection, the quantitative description, and the recognition (classification) of image area(s) of diagnostic and/or prognostic interest with a focus on:
- Feature extraction (texture analysis and morphometry), deep feature extraction, and fusion
- Feature selection (wrapped and embedded methods)
- Linear and non-linear dimensionality reduction (PCA, autoencoders)
- Classification models: statistical models, support vector machines, convolutional networks, deep knowledge transfer, and fine-tuning