Impact of Voxel Normalization on a Machine Learning-Based Method: A Study on Pulmonary Nodule Malignancy Diagnosis...

Area of Science:

• Medical Imaging
• Radiology
• Artificial Intelligence in Medicine

Background:

• Lung cancer (LC) is a leading cause of cancer mortality globally, making early detection crucial for survival.
• Low-dose computed tomography (LDCT) is vital for lung cancer screening, but generates large datasets that burden radiologists.
• Computer-aided diagnostic (CAD) tools can aid diagnosis, yet image voxel size variations pose challenges for model generalization and efficacy.

Purpose of the Study:

• To investigate the impact of different voxel sizes on diagnostic model performance for lung cancer detection.
• To evaluate disparities in diagnostic models trained on original versus reconstructed LDCT images with varying isotropic voxel sizes.
• To assess the ability of a support vector machine (SVM) model to differentiate between benign and malignant lung nodules using standardized voxel sizes.

Main Methods:

• LDCT images were reconstructed to achieve isotropic voxel sizes.
• A support vector machine (SVM) classifier was trained using 11 features on LDCT images with a 1.5 mm isotropic voxel size.
• The model was trained and evaluated using data from 225 patients.

Main Results:

• The SVM model achieved a high diagnostic performance.
• The model demonstrated an accuracy of 0.9596.
• The model achieved an area under the receiver operating characteristic curve (ROC/AUC) of 0.9855, indicating excellent discrimination between benign and malignant nodules.

Conclusions:

• Standardizing voxel size in LDCT images can enhance the performance of diagnostic models for lung cancer screening.
• The developed method shows promise for improving the efficacy and generalization of CAD tools.
• Further validation with multi-center LDCT data is recommended for clinical application.