3704 - Development and Validation of a Clinicoradiological and CT-Based Radiomics Nomogram to Predict Post-Neoadjuvant Lymph Node Metastasis in Esophageal Squamous Cell Carcinoma
Presenter(s)
Q. Peng, and C. Li; State Key Laboratory of Oncology in South China, Guangdong Key Laboratory of Nasopharyngeal Carcinoma Diagnosis and Therapy, Guangdong Provincial Clinical Research Center for Cancer, Sun Yat-sen University Cancer Center, Guangzhou, China
Purpose/Objective(s): To develop a reliable tool based on clinicoradiological features and CT images of the primary tumor to predict lymph node metastasis (LNM) status in esophageal squamous cell carcinoma (ESCC) patients undergoing neoadjuvant therapy (NAT).
Materials/Methods: A total of 469 patients with pathologically confirmed ESCC were randomly divided into a training cohort (n=328) and a test cohort (n=141). The tumor-habitat-based signature (Habitat_Rad) was derived from handcrafted radiomics features extracted from three distinct tumor subregions, determined using K-means clustering. The multiple instance learning-based radiomics signature (MIL_Rad) was constructed by integrating features from multiple 2.5-dimensional deep learning (DL) models. A five-step procedure, including reproducibility assessment, univariate analysis, Pearson's correlation analysis, the LASSO method, and evaluation with machine learning algorithms, was employed for feature selection and signature building. The clinic-radiological signature (Clinic) was built on the basis of independent clinical factors and radiological characteristics. A combined radiomics nomogram was created, integrating the Habitat_Rad, MIL_Rad, and independent clinic-radiological features. Model performance was evaluated in terms of discrimination, calibration, and clinical utility.
Results: The combined nomogram model demonstrated superior discrimination and accuracy, with an area under the curve (AUC) of 0.929 (95% CI, 0.902-0.957) in the training dataset and 0.852 (95% CI, 0.778-0.925) in the testing cohort. The calibration plot displayed a favorable calibration, and decision curve analysis substantiated its clinical significance relevance.
Conclusion: The developed radiomics nomogram exhibits encouraging performance in predicting lymph node metastasis (LNM) in ESCC patients undergoing NAT. Its application aids in guiding clinical decision-making and predicting prognosis.