A joint shape evolution approach to medical image segmentation using expectation-maximization algorithm.

Area of Science:

• Medical Imaging
• Computer Vision
• Neuroscience

Background:

• Accurate segmentation of deep brain structures in MRI is crucial for neurological research and clinical diagnosis.
• Existing methods often struggle with accuracy and robustness, particularly for complex structures like caudate, putamen, and thalamus.

Purpose of the Study:

• To develop and evaluate a novel expectation-maximization (EM)-based curve evolution algorithm for automated segmentation of brain structures in 3D MRI.
• To improve the accuracy and robustness of brain structure segmentation by integrating a hidden variable model with a statistical shape model.

Main Methods:

• The proposed algorithm utilizes an expectation-maximization (EM) framework for curve evolution.
• It incorporates a hidden variable model for local voxel labeling, estimated using image data and prior anatomical knowledge.
• The M-step jointly estimates structure shapes using the hidden variable model and a statistical prior model, while the E-step estimates expected observation likelihood and prior distributions.

Main Results:

• The EM-joint shape-based algorithm was applied to segment caudate, putamen, and thalamus in 3D brain MRIs of volunteers and patients.
• Experimental results demonstrated superior robustness and accuracy compared to traditional statistical shape model-based techniques.
• Performance also surpassed a current state-of-the-art region competition level set method.

Conclusions:

• The proposed EM-based curve evolution algorithm offers a robust and accurate solution for segmenting deep brain structures in MRI.
• This method holds significant potential for advancing neuroimaging analysis in both research and clinical settings.
• The integration of hidden variable and statistical shape models represents a key advancement in automated image segmentation.