Portable Ultrasound Research System for Use in Automated Bladder Monitoring with Machine-Learning-Based Segmentation.

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Ultrasound Technology

Background:

• Current bladder monitoring methods can be invasive or require frequent user interaction.
• There is a need for automated, non-invasive solutions for long-term bladder assessment.

Purpose of the Study:

• To develop and evaluate a mobile ultrasound device for automated, long-term bladder monitoring.
• To assess the performance of machine learning-based bladder segmentation algorithms.

Main Methods:

• A 32-element phased array transducer and mobile device architecture were utilized.
• Real-time image reconstruction was achieved using GPU-accelerated plane wave compounding.
• Machine learning (CNN) models were trained and evaluated for bladder segmentation accuracy.

Main Results:

• The ultrasound system demonstrated good imaging performance (resolution, SNR, CNR) on phantoms.
• Machine learning algorithms achieved reliable segmentation of human bladders across various filling levels.
• Quantitative and qualitative segmentation results were compared against manual segmentation ground truth.

Conclusions:

• The developed mobile ultrasound system facilitates automated and non-invasive bladder monitoring.
• ML-based segmentation shows promise for accurate bladder volume assessment.
• This technology has the potential to improve patient care and diagnostic efficiency.