[Structural analysis based on adaptive window for pulmonary nodule detection].

Area of Science:

• Medical Imaging
• Radiology
• Computer-Aided Diagnosis

Background:

• Pulmonary nodule detection using Hessian matrices is sensitive but prone to false positives at vascular intersections.
• Accurate identification of pulmonary nodules is crucial for early lung cancer diagnosis and treatment.

Purpose of the Study:

• To develop a novel pulmonary nodule detection method using Hessian matrix-based adaptive window structure analysis.
• To improve the accuracy of pulmonary nodule detection by reducing false positives caused by intersecting blood vessels.

Main Methods:

• Utilized Hessian matrix-based adaptive window structure analysis to differentiate nodules from vascular structures.
• Employed structure coefficients to construct a 3D adaptive window for local structure analysis.
• Applied a discrimination function for nodule detection.

Main Results:

• Achieved 100% detection sensitivity for pulmonary nodules of various sizes and types in experimental CT images.
• Significantly reduced false positive results, particularly in areas with intersecting blood vessels.
• Demonstrated the approach's effectiveness on CT images from 17 patients.

Conclusions:

• The proposed Hessian matrix-based adaptive window analysis offers a robust method for pulmonary nodule detection.
• This technique significantly enhances accuracy by minimizing false positives from vascular structures.
• The approach provides valuable assistance for subsequent nodule positioning and segmentation in clinical practice.