N7 - Enabling Robotic Paracentesis Through Autonomous Ultrasound Ascites Recognition

Introduction: : Nationally, only 40% of patients with cirrhosis and ascites received a timely paracentesis [1]. Access to paracentesis is an important part of management and symptom relief for patients with ascites, especially in underserved areas. Sampling ascites early in admission can help rule out fatal infections and prevent complications. The objective of this study is to assess the feasibility of an algorithm that can reliably identify and localize ascites on abdominal ultrasound images and ultimately be integrated into a prototype that can perform this procedure autonomously.

Materials and

Methods: : A two-phase study was conducted involving 1) algorithm development and 2) testing on de-identified ultrasound images. The algorithm, developed using Python and the OpenCV library, delineates regions of ascites. Eighteen images of abdominal ultrasound images (acquired via publicly available images in literature) were used for validation of the effective detection of anechoic regions of ascites.

Results, Conclusions, and Discussions:: The algorithm successfully demonstrated its ability to identify regions of ascites on each acquired image, 100% of regions of ascites were identified in all twenty-four still images acquired in various quadrants (Right Upper Quadrant (RUQ), Left Upper Quadrant (LUQ), Right Lower Quadrant (RLQ), and Left Lower Quadrant (LLQ)). However, the algorithm also identified areas that should be negligible due to their minuscule size or anatomic location. This finding calls for further refinement of the algorithm.

The algorithm demonstrated reliable identification of ascitic fluid, showing potential for integration into robotic systems to perform image-guided paracentesis autonomously. Ongoing efforts focus on enhancing algorithm specificity, likely through deep learning methods, and developing a prototype for clinical application. In addition to further refinement of the algorithm, it is suggested that the proposed algorithm can be integrated into a mechanical system for the future development of an autonomous drainage system. Future goals include a prototyped guide needle identification algorithm on ultrasound for better robotic system integration. Such advancements could improve procedural access globally, particularly in underserved regions, while maintaining high accuracy.

Acknowledgements and/or References (Optional)::

References:
[1] Orman ES, Hayashi PH, Bataller R, Barritt AS 4th. Paracentesis is associated with reduced mortality in patients hospitalized with cirrhosis and ascites. Clin Gastroenterol Hepatol. 2014;12(3):496-503.e1. doi:10.1016/j.cgh.2013.08.025