Application of Machine Learning Algorithms to the Discretization Problem in Wearable Electrical Tomography Imaging...

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Artificial Intelligence

Background:

• Urinary tract monitoring requires precise discretization for accurate imaging.
• Electrical Impedance Tomography (EIT) offers a non-invasive method for physiological monitoring.
• Existing EIT methods need robust discretization for effective bladder tracking.

Purpose of the Study:

• To implement and compare artificial intelligence (AI) algorithms for discretization in EIT for urinary tract monitoring.
• To develop a finite element mesh (FEM) classifier for separating bladder (inclusion) from background.
• To adapt supervised learning methods for tracking urinary bladder shape and position changes.

Main Methods:

• Adaptation of supervised learning algorithms: logistic regression, decision trees, linear and quadratic discriminant analysis.
• Development of a finite element mesh (FEM) classifier for image discretization.
• Implementation of AI algorithms to supplement inverse problem solutions in EIT.
• Design and demonstration of a novel EIT device for urinary tract monitoring.

Main Results:

• Demonstrated successful adaptation of AI algorithms for EIT-based bladder discretization.
• Showcased the capability of developed algorithms to track bladder placement and shape changes.
• Validated the effectiveness of the developed EIT device and accompanying IT solutions.
• Presented a robust measurement solution using sensors and statistical methods for bladder monitoring.

Conclusions:

• AI-driven discretization significantly enhances EIT capabilities for urinary bladder tracking.
• The developed EIT system provides effective information on bladder dynamics for rehabilitation.
• This approach addresses a strong market demand for advanced urinary tract monitoring solutions.