Juxta-vascular nodule segmentation based on flow entropy and geodesic distance

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Radiology

Background:

• Accurate lung nodule segmentation is essential for quantitative analysis in computed-aided diagnosis (CAD) of lung CT data.
• Distinguishing malignant from benign nodules relies on indicators like growth rate, necessitating precise segmentation.
• Juxta-vascular nodules present segmentation challenges due to similar gray levels between nodules and adjacent blood vessels.

Purpose of the Study:

• To develop and validate an automated method for segmenting juxta-vascular lung nodules.
• To improve the accuracy of nodule segmentation by distinguishing nodule regions from adjacent blood vessels.
• To enhance the reliability of nodule growth rate calculation for cancer diagnosis.

Main Methods:

• Introduction of a 'flowing direction feature' (normal vector direction) to differentiate vessel and nodule pixels.
• Utilizing flow entropy and geodesic distance to k-means clustering for segmentation.
• Validation on juxta-vascular nodules from the Chinalung-CT screening trial and the Lung Image Database Consortium (LIDC) dataset.

Main Results:

• High accuracies achieved in fully automated segmentation: 92.9% (Chinalung-CT), 87.5% (LIDC dataset 1), and 94.9% (LIDC dataset 2).
• The proposed method effectively distinguishes nodule regions from adjacent blood vessels.
• Demonstrated low time complexity and high accuracy in segmenting challenging juxta-vascular nodules.

Conclusions:

• The proposed method offers an accurate and efficient solution for segmenting juxta-vascular lung nodules.
• This technique can significantly aid in the quantitative analysis of lung nodules for improved cancer diagnosis.
• The flow entropy and geodesic distance approach provides a robust tool for medical image analysis in radiology.