Transfer-Learning Deep Radiomics and Hand-Crafted Radiomics for Classifying Lymph Nodes from Contrast-Enhanced Computed Tomography in Lung Cancer

ABSTRACT:

Simple Summary

Positron emission tomography is currently considered the non-invasive reference standard for determining whether lung cancer also affects thoracic lymph nodes (staging). However, not all patients can undergo this diagnostic procedure due to high costs, limited availability, and additional radiation exposure. This study aimed to predict the positron emission tomography result from traditional contrast-enhanced computed tomography and test new feature extraction strategies. As input, we compared traditional (hand-crafted) imaging biomarkers (radiomics) with novel features derived from pre-trained neural networks. This hybrid approach yielded better performance than using both feature sources alone. In conclusion, both traditional radiomics features and transfer-learni

SUBMITTER: Laqua F 

PROVIDER: S-EPMC10216416 | biostudies-literature | 2023 May

REPOSITORIES: biostudies-literature