2023 World Conference on Lung Cancer (Posters)-EP08.02. Lymphovascular Invasion Prediction with CT images in Patients with Non-Small Cell Lung Cancer (NSCLC)

EP08.02. Lymphovascular Invasion Prediction with CT images in Patients with Non-Small Cell Lung Cancer (NSCLC) - PDF(Abstract)

Back to course

Pdf Summary

This study aimed to predict lymphovascular invasion (LVI) in patients with non-small cell lung cancer (NSCLC) using pre-treatment CT images. LVI is a significant prognostic factor in NSCLC but is typically only obtainable after tumor resection for some patients. The researchers utilized CT scans, tumor annotations, and surgically proven LVI information from a radiogenomic dataset. They randomly divided 64 NSCLC patients into training and validation cohorts. Radiomic features were extracted from the CT images using the PyRadiomics Python module, including first-order, gray-level co-occurrence matrix, gray-level dependence matrix, gray-level run length matrix, gray-level size zone matrix, and neighborhood gray-tone difference matrix. Univariate analysis was conducted to select informative features and a machine-learning random forest tree model was developed using the statistically significant radiomic features. The model was evaluated using the validation cohort. In the NSCLC cohort, LVI was present in 14 patients and absent in 50 patients. Thirty radiomic features showed statistically significant differences between the two groups. The machine-learning model achieved an area under the receiver operating characteristic curve (AUC) of 0.69. The study concluded that LVI can be predicted non-invasively using CT image biomarkers and artificial intelligence (AI), providing valuable insights for selecting individualized treatment options for NSCLC patients.

Asset Subtitle

Misuk Lee

Meta Tag

Speaker Misuk Lee

Topic Local-Regional NSCLC: Multimodality Therapy

Keywords

lymphovascular invasion

LVI

non-small cell lung cancer

NSCLC

pre-treatment CT images

radiogenomic dataset

radiomic features

random forest tree model

area under the receiver operating characteristic curve