Philips · Image-Guided Therapy · AI
Sahar Yousefi
Senior Computer Vision and AI Scientist at Philips
I’m a Senior Computer Vision and AI Scientist at Philips, where I help lead AI development for DeviceGuide, an image-guided therapy solution that supports clinicians during heart valve procedures. I hold a PhD in Artificial Intelligence. Previously, I was Lead Computer Vision Engineer at Prime Vision, working on document understanding with large language models. At AutoFill Technologies, I was Head of AI Development, working on machine-vision technology to automate inspections in the automotive industry. Earlier, I was a postdoctoral researcher at Leiden University Medical Center.
Selected projects
- DeviceGuide: AI-enabled device guidance for minimally invasive heart valve repair, built on Philips EchoNavigator.
- Donut-hole: A distilled, sparsified version of the Donut document-understanding model that reduces model density by 54% while preserving performance.
- fastMRI: I contributed to the Philips–LUMC team that won the most challenging part of the 2019 fastMRI challenge, reconstructing near-intact knee MRI images with one-eighth of the usual data.
- DDAUnet: An attention-based deep learning method for automatic esophageal tumor segmentation in CT images, evaluated on 792 scans from 288 patients.
- ASL-to-PET Translation: A semi-supervised model that translates ASL MRI into PET images using paired and unpaired data.
- Hadamard-te-ASL Reconstruction: Deep learning reconstruction of dynamic perfusion and angiography from time-encoded ASL data.
- Motion Detector: A 3D wavelet-based method for detecting motion under severe illumination changes.