Multi-Modal Ultrasound-Based Prognostic Model for Diffuse Large B-Cell Lymphoma with Predominantly Superficial Lymph...

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Diffuse large B-cell lymphoma (DLBCL) is an aggressive non-Hodgkin lymphoma.
• Accurate prognostic models are crucial for guiding treatment decisions in DLBCL.
• Superficial lymph node involvement is a common presentation in DLBCL.

Purpose of the Study:

• To develop and validate a multi-modal ultrasound-based predictive model for progression-free survival (PFS) in DLBCL patients.
• To integrate radiomics, deep learning, and clinical features for enhanced prognostic accuracy.
• To assess the model's performance using C-index, AUC, and risk stratification.

Main Methods:

• Retrospective analysis of 281 DLBCL patients with superficial lymph node involvement.
• Extraction of ultrasound radiomic features and deep learning features using DenseNet121.
• Feature selection via Pearson's correlation and LASSO-Cox regression.
• Development and comparison of five prognostic models.

Main Results:

• The combined multi-modal model achieved a high C-index of 0.811 in the test set.
• The model demonstrated strong 1-, 2-, and 3-year PFS prediction with AUCs around 0.81-0.83.
• High-risk stratification group showed significantly poorer PFS (p < 0.01).

Conclusions:

• A multi-modal ultrasound model integrating radiomics, deep learning, and clinical data shows significant promise for PFS prediction in DLBCL.
• The developed nomogram offers a practical tool for individualized prognostic assessment and risk stratification.
• This approach facilitates intuitive and effective patient management in DLBCL.