Risk Analysis for Pathological Changes in Pulmonary Parenchyma Based on Lung Computed Tomography Images

Area of Science:

• Medical Imaging
• Radiology
• Computational Pathology

Background:

• Pulmonary parenchyma pathological changes require accurate assessment for risk analysis.
• Existing medical image retrieval systems may lack specificity in pathological analysis.
• Computed tomography (CT) imaging is crucial for visualizing lung structures.

Purpose of the Study:

• To design a content-based medical image retrieval system for pathological pulmonary parenchyma analysis.
• To utilize a latent Dirichlet allocation (LDA) model for risk assessment of lung abnormalities.
• To enhance the excavation and evaluation of pathological changes in lung CT images.

Main Methods:

• A dataset of 115 lung CT scans with pathological changes was utilized.
• Images were preprocessed and decomposed into pixel blocks (words) using morphological theory to build a vocabulary.
• A latent Dirichlet allocation (LDA) model was developed and validated using leave-one-out cross-validation, with precision and recall as performance metrics.

Main Results:

• The LDA model successfully ranked retrieval results by relevance.
• Precision for identical tissue identification reached 0.76 ± 0.031 from the top 50 retrieved images.
• Precision for individual pulmonary parenchyma attributes ranged from 0.776 ± 0.043 to 0.984 ± 0.008.

Conclusions:

• The proposed LDA model demonstrates effectiveness in lung CT image retrieval.
• The system shows reliable efficacy for risk analysis of pathological changes in the pulmonary parenchyma.
• This approach facilitates improved assessment of lung abnormalities through advanced image analysis.