3D gray density coding feature for benign-malignant pulmonary nodule classification on chest CT.

Area of Science:

• Medical Imaging
• Radiology
• Computer-Aided Diagnosis

Background:

• Early lung cancer detection is crucial for reducing mortality.
• Computer-aided detection systems (CADs) assist radiologists in early diagnosis.
• Accurate classification of pulmonary nodules as benign or malignant is essential.

Purpose of the Study:

• To propose a novel 3D gray density coding feature (3D GDC) for pulmonary nodule classification.
• To fuse 3D GDC with geometric features for improved diagnostic accuracy.
• To evaluate the performance of a random forest classifier using the fused features on Chest CT scans.

Main Methods:

• A dictionary model was created to generate a codebook for feature extraction.
• 3D GDC features were obtained through histogram statistics on feature descriptors derived from the codebook.
• Geometric features were extracted and fused with 3D GDC features.
• A random forest classifier was employed for benign-malignant pulmonary nodule classification.

Main Results:

• The proposed method achieved high accuracy (93.17 ± 1.94%) and AUC (97.53 ± 1.62%) on the LIDC-IDRI dataset.
• On the private ZSHD dataset, the method yielded 90.0% accuracy and 93.15% AUC.
• Results were superior to other state-of-the-art methods on the LIDC-IDRI dataset.

Conclusions:

• The developed method accurately classifies benign and malignant pulmonary nodules, aiding in auxiliary diagnosis.
• The approach offers greater interpretability compared to 3D CNN methods.
• The system provides valuable auxiliary information for medical professionals.