CT-Based Habitat Radiomics Combining Multi-Instance Learning for Early Prediction of Post-Neoadjuvant Lymph Node...

Area of Science:

• Oncology
• Radiology
• Medical Imaging

Background:

• Accurate prediction of lymph node metastasis (LNM) after neoadjuvant therapy (NAT) is vital for esophageal squamous cell carcinoma (ESCC) treatment optimization.
• Current methods may lack precision in predicting LNM post-NAT.

Purpose of the Study:

• To develop and validate a computed tomography (CT)-based radiomic model for predicting LNM in ESCC patients post-NAT.
• To assess the model's performance in guiding individualized therapy decisions.

Main Methods:

• Retrospective analysis of 469 ESCC patients, divided into training (n=328) and testing (n=141) cohorts.
• Development of three signatures: tumor-habitat-based (Habitat_Rad), multiple instance learning-based (MIL_Rad), and clinicoradiological (Clinic).
• Construction of a combined radiomic nomogram integrating the developed signatures.

Main Results:

• The combined radiomic nomogram achieved high predictive accuracy, with AUCs of 0.929 in the training cohort and 0.852 in the test cohort.
• The nomogram demonstrated superior performance compared to individual signatures.
• Decision curve analysis confirmed a significant net clinical benefit, indicating clinical utility.

Conclusions:

• The developed CT-based radiomic nomogram is a promising tool for predicting LNM in ESCC patients after NAT.
• This model can aid in timely treatment adjustments and personalized therapeutic strategies for ESCC.