Capillary stalling and slowed transit time occur in neuroinflammation, Alzheimer's, and stroke, but identifying where these changes happen has been impossible. For the first time, we can map this across the entire brain non-invasively, pinpointing which regions and which specific capillaries show dysfunction—critical for understanding disease mechanisms and monitoring treatments.

The blood vessels contained within the cardiac wall cannot be imaged directly and individually, while their dysfunction account for more than 50% of patients presenting with stable angina. We were able to show the feasibility of mapping those vessels in the beating heart and are working toward clinical translation. 

Manipulating the vasculature using ultrasound to achieve bioeffects such as opening the blood brain barrier is gaining momentum, but monitoring methods in small animals remain mostly limited to magnetic resonance imaging. We are working toward the development of ultrasound-based solutions to facilitate research on larger animal populations. 

Ultrasound can propagate deep in tissue but is limited by bones in general and the skull in particular, which often require craniotomies in small animal studies and limit translation in humans. We have recently developed several new approaches for aberration correction that are robust and reproducible.

Dynamic ultrasound localization microscopy requires teh tracking of a large number of microbubbles. We are developing novel approaches to track more microbubbles, in presence of higher concentration and for longer duration to enable the development of novel biomarkers.

3D ultrafast ultrasound imaging is a challenging engineering problem as the required number of piezoelectric elements (and thus cost and datasets) grow quadratically. We are working on several approaches based on novel probe architectures to address this problem. 

A important advantage of ultrasound imaging is that methods developed for small animal imaging are readily translatable to humans, often using the same electronics. When moving to the clinic, we need to adapt our methods; we have developed new approaches to drastically improve image quality in the human heart and to cross the thicker skulls for brain imaging.

We are developing highly realistic simulation frameworks to predict how the behavior of microbubbles in the microvasculature affect hundreds of thousands of ultrasound images and how these images can be used to recover said behavior. Such simulations involve large computations typically run on massively parallel architectures.

We are developing deep learning methods to process our images at various stages from aberration correction to microbubble detections and labelling. We have been pionneers in using several novel components in these networks such as sparse or complex-valued networks.