Foundation models for breast imaging
I am a PhD student at the University of Girona. I study how large pretrained vision and vision-language models can detect lesions and generate clinical text from mammography.
Research
Lesion detection
Adapting large pretrained vision backbones to localise masses, calcifications and other findings, with detection heads that transfer across views, scanners and datasets.
Text generation
Vision-language models that ground findings in the image and generate structured, clinically faithful radiology text, evaluated beyond n-gram overlap.
Foundation models for the breast
Self-supervised and multimodal pretraining on mammography, parameter-efficient fine-tuning, domain adaptation and zero- or few-shot transfer to new clinical tasks.
Interactive demo: mammography preprocessing
Models only perform as well as their inputs. This demo follows a 14-bit DICOM mammogram through windowing, breast cropping and resizing, and shows how annotations stay aligned at every step. It uses a synthetic image, with no patient data.
Publications
- 2024Application of Deep Learning Models in Image Segmentation
Enric Sena Alvarez, Antonio Lozano Bagén, · UAB Digital Document Repository · PDF
Education
- 2026– 2030 PhD · Universitat de Girona
Foundation models for breast imaging. - 2024-2025MSc in Big Data & Visual Analytics
- 2019-2024BSc in Applied Statistics · Universitat Autònoma de Barcelona
Contact
GitHub · Google Scholar · LinkedIn · ORCID