A feasibility study of an algorithm that identifies ascites on image-guided paracentesis

Area of Science:

• Medical imaging analysis
• Artificial intelligence in healthcare
• Abdominal diagnostics

Background:

• Ascites management via image-guided paracentesis is crucial for patient therapy and infection detection.
• Limited availability of trained providers for image-guided procedures presents a challenge.
• Autonomous detection of ascites on ultrasound could improve accessibility and timeliness of care.

Purpose of the Study:

• To develop an autonomous algorithm for reliable identification and localization of ascites in abdominal ultrasound images.
• To create a software tool that assists in the prompt management of ascites.
• To address the need for accessible diagnostic tools in resource-limited settings.

Main Methods:

• Development of an algorithm using the OpenCV image processing library in Python.
• Validation of the algorithm using 16 random abdominal ultrasound images with multiple views.
• Quantitative analysis of the algorithm's performance in identifying ascitic regions.

Main Results:

• The developed algorithm demonstrated 100% accuracy in identifying targeted ascitic regions.
• The software successfully localized ascites across diverse abdominal ultrasound images.
• The algorithm proved to be a reliable tool for ascites detection.

Conclusions:

• The proposed algorithm is a reliable and accurate tool for autonomous ascites detection on abdominal ultrasound.
• This AI-driven approach can support prompt ascites management and infection identification.
• The technology holds potential for enhancing diagnostic capabilities in various medical settings.