Ultrasound Image Interpretation
Ultrasound imaging generates rich, dynamic information about tissue motion, structure, and function. Transforming these raw signals into clinically meaningful insight and interpretable results requires advanced analysis tools. Within PULS/e, we develop data鈥慸riven and physics鈥慽nformed algorithms to extract quantitative biomarkers from ultrasound images, reduce operator dependence, and ultimately support clinicians in making earlier, more precise decisions.
Our research spans the full chain from automated image interpretation to model鈥慴ased assessment of disease progression. By combining ultrasound physics, signal and image processing, and AI鈥慸riven analysis, we aim to make ultrasound more robust, reproducible, and informative across clinical settings.
These methods not only support high鈥憅uality imaging systems but are also essential in emerging point鈥憃f鈥慶are ultrasound applications, where consistent and automated interpretation is key for broader clinical adoption.
Ultrasound Image Analysis & AI
In this line of research, we develop methods that 鈥渓ook beyond the image鈥�, transforming complex ultrasound data into reliable markers of tissue structure, function, and dynamics. This includes robust segmentation of organs and vascular structures, motion and strain estimation, elastography, flow analysis, and multimodal image registration.
These algorithms support clinicians in quantifying tissue health, identifying pathological changes, and interpreting ultrasound data consistently, even in challenging imaging environments.
We also design AI鈥慸riven workflows that improve interpretability and reduce operator dependence, such as autonomous probe positioning, real鈥憈ime quality feedback, and deep鈥憀earning鈥揵ased reconstruction and noise reduction. By merging physics鈥慴ased models with data-driven learning, our goal is to enhance image clarity, expand the diagnostic value of ultrasound, and create tools suitable for both expert and point鈥憃f鈥慶are users.
Ultrasound鈥慽nformed Digital Twins
Within this research line, we integrate ultrasound-derived measurements with biomechanical modeling to create digital replicas of vascular structures, such as abdominal aortic aneurysms, stenotic arteries, or plaques.
These digital twins allow us to:
鈥� assess mechanical stress, rupture risk, and tissue vulnerability;
鈥� simulate disease progression over time;
鈥� predict the effect of interventions;
鈥� and support personalized clinical decision-making.
Our approach combines advanced image analysis, finite element and fluid鈥搒tructure interaction modeling, and in vivo validation with clinical partners. This framework has already demonstrated strong potential for vascular risk stratification and is now extending toward other organ systems and ultrasound-guided therapies.