Paper 11960-79
Paper 11960-79
Development of a superpixel spectral unmixing framework for the volumetric photoacoustic imaging: A Machine Learning approach
25 January 2022 • 4:45 PM - 5:00 PM PST | Room 211 (Level 2 South)
Abstract
The volumetric assessment of molecular tissue components is crucial in the diagnosis and treatment of a large number of disease. Here we propose a novel framework that enables the automated assessment of volumetric tissue composition from 3-D spectral photoacoustic imaging. The framework, based on unsupervised ML algorithms, has been tested and validated with tissue mimicking phantoms and in vivo preclinical studies. The initial results show that the proposed image-based framework outperforms the conventional unmixing approaches, with enhanced sensitivity and specificity to differentiate the less prominent absorbers and improved quantification measures in deep tissues.
Presenter
FUJIFILM VisualSonics, Inc. (Netherlands), Christian-Albrechts-Univ. zu Kiel (Germany)
Valeria Grasso received her Master’s degree in Biomedical Engineering from the University of Pisa, Pisa, Italy, in 2018. During her Master’s thesis project she worked on the unsupervised image-based assessment of mechanical properties of cardiovascular tissues from Phase Contrast MRI, at the Great Ormond Street Hospital for Children in London, UK, and at BioCardioLab in Massa, Italy. She is currently a PhD student within the Marie Curie network called MgSAFE and she is based at FUJIFILM VisualSonics, Amsterdam, The Netherlands. Her main research interest centers on optimizing the high-resolution Ultrasound and Photoacoustic (US-PA) imaging to non-invasively monitor the molecular changes in the soft tissues surrounding biodegradable bone implants made of Magnesium. Her current work focuses on the development of advanced US-PA image processing algorithms and optimal hardware solutions to adapt the US-PA technology for human patients treated with Mg implants in case of bone injuries.