An Approach for Asbestos-related Pleural Plaque Detection.

Area of Science:

• Medical Imaging
• Computational Pathology
• Toxicology

Background:

• Asbestos exposure can lead to pleural plaques, which may signal underlying lung dysfunction.
• Accurate detection of these plaques is crucial for assessing asbestos-related health risks.
• Current diagnostic methods may be limited by data availability and algorithmic approaches.

Purpose of the Study:

• To develop and evaluate a novel pipeline for automated detection of asbestos-related pleural plaques in CT images.
• To address the challenge of limited annotated datasets for pleural plaque detection.
• To propose a convolutional neural network (CNN) architecture learning approach for improved performance with small datasets.

Main Methods:

• A pipeline involving lung segmentation, 3D patch selection, CNN feature extraction, and support vector machine (SVM) classification was developed.
• A unique CNN approach utilizing random weights from Principal Component Analysis (PCA) was employed, bypassing traditional backpropagation.
• The method was designed to overcome the scarcity of publicly available annotated pleural plaque datasets.

Main Results:

• The proposed CNN architecture learning approach demonstrated superior performance compared to traditional backpropagation methods when trained on small datasets.
• The pipeline effectively identifies asbestos-related pleural plaques in CT scans of the human thorax.
• The study validates the efficacy of PCA-based random weights for CNNs in medical image analysis.

Conclusions:

• The developed pipeline offers a promising solution for asbestos-related pleural plaque detection, particularly in data-scarce scenarios.
• The novel CNN architecture learning strategy is effective for medical image analysis tasks with limited training data.
• This approach can aid in the early identification of potential asbestos exposure consequences.