A delegation of researchers from the MAIBAI project presented 5 posters and 1 oral contribution at the 47th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, held in Copenhagen (DK), on 14-17 July 2025.
Emir Ahmed (Department of Data Science and AI, National Physical Laboratory (NPL), UK)
Style transfer as data augmentation: evaluating unpaired image-to-image translation models in mammography
The presented works focuses on evaluating image-to-image translation models, specifically CycleGAN and SynDiff, to improve cross-domain generalization in mammography through style transfer. It compares commonly used metrics and their limitations, highlighting the importance of using multiple complementary metrics to evaluate style-transfer methods in medical imaging reliably.
Authors: Emir Ahmed1, Spencer A. Thomas1 and Ciaran Bench1
- Department of Data Science and AI, National Physical Laboratory (NPL), UK
Ph.D Student Martina Oria (Istituto Nazionale di Ricerca Metrologica (INRiM), Torino, Italy)
Evaluation of Mammogram Vendor-Style Transfer with Generative AI
The presented work focuses on generating synthetic mammograms with varying vendor characteristics using a CycleGAN-based neural style transfer approach. It includes an extended parametric analysis to assess performance quantitatively, examining how well the method matches target styles while preserving the original image content.
Authors: Martina Oria1,2, Riccardo Ferrero1, and Alessandra Manzin1
- Istituto Nazionale di Ricerca Metrologica (INRiM), Torino, Italy
- Politecnico di Torino, Torino, Italy
Ph.D Student Martina Oria (Istituto Nazionale di Ricerca Metrologica (INRiM), Torino, Italy)
Generation of Digital Breast Phantoms and Synthetic Mammograms
The work focuses on generating synthetic mammograms by reconstructing breast digital phantoms from CT images using trivariate tensor-product B-splines. The phantoms are further modified through geometric transformations that mimic typical breast shape variations and by adding artificial lesions. Following biomechanical compression, an x-ray–phantom interaction model is applied to produce realistic mammographic images.
Authors: Martina Oria1,2, Riccardo Ferrero1, Chiara Andreis3, Marta Vicentini1, Ruben van Engen4, Carlijn Roozemond4, Paola Lamberti3, Sara Remogna3, and Alessandra Manzin1
- Istituto Nazionale di Ricerca Metrologica (INRiM), Torino, Italy
- Politecnico di Torino, Torino, Italy
- Dipartimento di Matematica “Giuseppe Peano”, Università degli Studi di Torino, Torino, Italy
- Dutch Expert Center for Screening (LRCB), Nijmegen, Netherlands
Dr. Alessandra Manzin (Istituto Nazionale di Ricerca Metrologica (INRiM), Torino, Italy)
Convolutional Neural Network Application for Cancer Detection in Mammography Images
The work focuses on developing and evaluating a convolutional neural network for detecting cancerous lesions in mammograms acquired from heterogeneous imaging systems. The algorithm is tested
on images with different vendor styles, which have been previously standardised using a neural style transfer method.
Authors: Eva Raina1,2, Martina Oria1,2, Riccardo Ferrero1, Mirko Polato3, and Alessandra Manzin1
- Istituto Nazionale di Ricerca Metrologica (INRiM), Torino, Italy
- Politecnico di Torino, Torino, Italy
- Dipartimento di Informatica, Università degli Studi di Torino, Torino, Italy
Dr. Marta Vicentini (Istituto Nazionale di Ricerca Metrologica (INRiM), Torino, Italy)
Comparison of mass segmentation methods for mammography
The work focuses on evaluating semi-automatic segmentation techniques for detecting breast cancer lesions in mammograms, addressing the challenge posed by the lack of standardised criteria. It compares different methods and examines how image preprocessing can optimise their performance, highlighting the need for consistent and reliable segmentation practices.
Authors: Marta Vicentini1, Martina Oria1,2, Caterina Fracchia3, Paola Lamberti3, Sara Remogna3, and Alessandra Manzin1
- Istituto Nazionale di Ricerca Metrologica (INRiM), Torino, Italy
- Politecnico di Torino, Torino, Italy
- Dipartimento di Matematica “Giuseppe Peano”, Università degli Studi di Torino, Torino, Italy
Alistair J. Hickman (Department of Physics, University of Surrey, Guildford, UK)
Visual Explanation of Generalisation between Mammography Manufacturers
Authors: Alistair J. Hickman1,2,3
- Department of Physics, University of Surrey, Guildford, United Kingdom
- Department of Scientific Computing and National Co-ordinating Centre for the Physics of Mammography, Royal Surrey NHS Foundation Trust, Guildford, United Kingdom,
- National Physical Laboratory, Teddington, United Kingdom