Juan Felipe Pérez-Juste Abascal, Ph.D.
Researcher in medical imaging, image processing, inverse problems and deep learning
Current position as Data Scientist Innovation at Groupe SEB, Lyon, France

juanabascal78 at gmail.com
Latest projects
From 2020 to 2024, I have had the opportunity to work of the following challenging problems: robust deep iterative reconstruction, high-dimensional image denoising and image restoration, motion artefact reduction, object detection, NLP and ChatGPT prompt text analysis.

Deep high-dimensional denoising
2D+frequency denoising:
-
Exploits higher-order convolutions (regularization across both space and frequency)
-
Image denoising for single-pixel hyperspectral imaging


Transfer learning
-
Learned the detector response function for spectral CT from few datasets
-
A Sim2Real transfer learning approach when a ground truth is not available
-
Material decomposition of human thorax data
-
Improved regularized approaches

Deep residual prior denoising
State-of-the art denoising models
-
ResNet, U-Net architecture, ViT
-
Specifically designed architectures
Deep learning denoising with Prior Image:
-
2D and 3D image denoising of spectral CT images
-
Exploits a Prior image (extending sparse and prior-based methods)


Sparse denoising methods
Sparse spatial and prior based denoising (SPADE) method
-
3D image denoising
-
Exploits a Prior Image
-
Improved stability and convergence
-
Reduces cartoon-like artefacts


Deep iterative reconstruction
-
Robust methods for medical imaging (ensure data fidelity)
-
Unrolled methods
-
Plug-and-play iterative methods based on state-of-the art denoising
Computer vision
-
Video object detection with TFLite on Raspberry Pi
-
Intelligent video surveillance
-
Image and video restoration
-
Semantic segmentation

Photon counting Spectral CT

-
EU-H2020 Marie Skłodowska-Curie Project "SUCCESS" on spectral computed tomography (CT)
-
Material decomposition for spectral CT
-
Regularization methods
-
Deep learning
-
Feasibility of early detection of Osteoarthritis using spectral CT
-
NLP
-
Named-entity recognition
-
Text classification
-
OCR
-
ChatGPT prompt text analysis