Cardiac image segmentation using generalized polynomial chaos expansion and level set function.

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Computational Science

Background:

• Cardiovascular Magnetic Resonance (CMR) imaging is prone to uncertainties from measurement limitations and patient variability.
• These uncertainties can lead to inaccurate cardiac function estimations, potentially resulting in misdiagnosis and suboptimal treatment.
• Reliable cardiac chamber segmentation is crucial for accurate cardiovascular assessment.

Purpose of the Study:

• To develop a stochastic image segmentation algorithm for improved cardiac chamber separation in CMR images.
• To address noise and uncertainties inherent in CMR data.
• To enhance the reliability of cardiac function evaluation through more accurate segmentation.

Main Methods:

• A novel stochastic image segmentation algorithm integrating generalized polynomial chaos (gPC) expansion with a level set function.
• A two-step process involving deterministic segmentation to define a boundary neighborhood and stochastic segmentation for boundary evolution.
• Calibration of the gPC model using pixel values from the identified neighborhood.

Main Results:

• The algorithm successfully segments cardiac chambers from the background in CMR images.
• It provides a probabilistic description of the segmented heart boundaries, quantifying segmentation uncertainty.
• Demonstrates potential for improved accuracy in cardiac function assessment.

Conclusions:

• The proposed stochastic segmentation method enhances the reliability of CMR image analysis.
• By accounting for image uncertainties, it offers a more robust approach to cardiac chamber segmentation.
• This technique holds promise for more accurate cardiac function evaluation and clinical decision-making.