[New semi-automatic ROI setting system for brain PET images based on elastic model]

Area of Science:

• Medical Imaging
• Neuroscience
• Computer-Aided Diagnosis

Background:

• Accurate delineation of anatomical regions in brain Positron Emission Tomography (PET) images is crucial for quantitative analysis.
• Conventional manual setting of Regions of Interest (ROIs) is time-consuming and prone to inter-operator variability.
• Existing automated methods may lack the flexibility to adapt to individual anatomical variations.

Purpose of the Study:

• To develop and evaluate a semi-automatic system for efficient and reproducible ROI setting in brain PET imaging.
• To reduce the operational time and inter-operator variance associated with manual ROI definition.
• To leverage an elastic network model for accurate anatomical atlas fitting.

Main Methods:

• Development of a semi-automatic system utilizing an elastic network model to fit a standard ROI atlas to individual brain PET images.
• Operator input involves defining a midsagittal line, semi-automatic brain contour detection (SNAKES algorithm), and a few specific ROIs for precise transformation.
• Comparison of the novel system's performance against conventional manual ROI setting methods.

Main Results:

• Significant reduction in operation time, with approximately 50% decrease across most cases.
• Substantial decrease in inter-operator variance, reduced by up to one-seventh in maximum cases.
• Demonstrated feasibility and efficiency of the semi-automatic approach for brain PET ROI definition.

Conclusions:

• The developed semi-automatic ROI setting system offers a substantial improvement in efficiency and reproducibility for brain PET image analysis.
• The elastic network model provides an effective means to adapt standard anatomical atlases to individual patient data.
• This system has the potential to enhance the clinical utility of brain PET imaging by streamlining data processing and reducing measurement errors.