Automating analysis and training in Focused Cardiac Ultrasound (AutoFoCUS)

Focused cardiac ultrasound (FoCUS) refers to the use of ultrasound imaging to evaluate cardiac structure and function at the bedside by a treating physician. In recent years, FoCUS has become an indispensable first-line diagnostic tool, complementing the traditional physical examination and accelerating patients’ evaluation in acute care settings. Nevertheless, proficiency in FoCUS requires (continuous) training in image acquisition and interpretation, with its clinical efficacy tightly dependent on the user skill. This dependency, together with its ever-growing usage in daily practice, has led medical schools to introduce bedside ultrasound in their curriculum. Since FoCUS training relies heavily on lengthy hands-on training sessions and daily practice on interpretation skills, the schools’ efforts raised one key challenge – the insufficient volume and availability of expert faculty members. Moreover, practitioners need to maintain high skill levels, as misinterpretation of suboptimal images may lead to critical complications under emergency scenarios.
Leveraging recent advances in the field of artificial intelligence (AI), this project aims to address the abovementioned challenges by pursuing three intertwined research lines: (1) create an AI-based medical training software for FoCUS; (2) develop an integrated framework for automatic diagnosis in FoCUS; and (3) explore the putative potential of FoCUS for whole-heart cardiac function quantification. Overall, this project will deliver innovative artificially intelligent systems to assist medical training and analysis in FoCUS. While the former will permit practitioners’ self-learning, the latter will provide automatic tools to help detect life-threatening and time-sensitive conditions, which together may ultimately increase the effectiveness of this life-saving diagnostic tool.

Funding Agency

FCT

Project Reference

PTDC/EMD-EMD/1140/2020

Project Members

Main Project Outcomes

S. Queirós, “Right ventricular segmentation in multi-view cardiac MRI using a unified U-net model”, in E. Puyol Antón et al. (eds) Statistical Atlases and Computational Models of the Heart. Multi-Disease, Multi-View, and Multi-Center Right Ventricular Segmentation in Cardiac MRI Challenge. STACOM 2021. Lecture Notes in Computer Science, vol 13131, pp. 287-295, Springer, Cham, 2022.

“Best Paper Award in the M&Ms-2 Challenge”, by M&Ms2 Challenge organizers and the Medical Image Computing and Computer Assisted Intervention (MICCAI) Society.

Main Project Outcomes

S. Queirós, “Right ventricular segmentation in multi-view cardiac MRI using a unified U-net model”, in E. Puyol Antón et al. (eds) Statistical Atlases and Computational Models of the Heart. Multi-Disease, Multi-View, and Multi-Center Right Ventricular Segmentation in Cardiac MRI Challenge. STACOM 2021. Lecture Notes in Computer Science, vol 13131, pp. 287-295, Springer, Cham, 2022. https://doi.org/10.1007/978-3-030-93722-5_31

“Best Paper Award in the M&Ms-2 Challenge”, by M&Ms2 Challenge organizers and the Medical Image Computing and Computer Assisted Intervention (MICCAI) Society.

“NVIDIA Academic Hardware Grant”, by NVIDIA Corporation.

Provisional Portuguese Patent (PT 117845), “Device and method for obtaining a full-surface estimate of an organ from a received partial anatomical point cloud”, S. Queirós, J. Gomes-Fonseca, J. L. Vilaça, E. Lima, J. Correia-Pinto, 2022.