Computed Tomography-Based Deep Learning Nomogram Can Accurately Predict Lymph Node Metastasis in Gastric Cancer

Area of Science:

• Medical Imaging
• Oncology
• Artificial Intelligence

Background:

• Computed tomography (CT) is standard for assessing lymph node status before surgery, but its accuracy is insufficient.
• Accurate preoperative lymph node staging is crucial for effective gastric cancer treatment planning.

Purpose of the Study:

• To evaluate the predictive performance of a deep learning (DL) model using CT for lymph node metastasis in gastric cancer.
• To verify the diagnostic efficacy of a CT-based DL nomogram for presurgical lymph node evaluation.

Main Methods:

• Retrospective analysis of 347 gastric cancer patients (242 training, 105 testing).
• Extraction of radiomics and DL features from enhanced CT arterial phase images.
• Development of predictive models using Support Vector Machine (SVM) or Random Forest (RF) and a clinical nomogram.

Main Results:

• A ResNet50-RF model demonstrated high classification performance (AUC=0.9803 in test cohort).
• A nomogram integrating DL features and CT findings achieved excellent discrimination (AUC=0.9914 in test cohort).
• Decision analysis confirmed the nomogram's clinical utility.

Conclusions:

• CT-based DL nomogram accurately predicts lymph node metastasis in gastric cancer preoperatively.
• This approach enhances diagnostic accuracy for lymph node status, aiding surgical decisions.