Prognostic Utility of a Deep Learning Radiomics Nomogram Integrating Ultrasound and Multi-Sequence MRI in...

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Triple-negative breast cancer (TNBC) presents unique challenges in predicting treatment response and patient outcomes.
• Accurate prognostic tools are crucial for guiding neoadjuvant chemotherapy (NAC) and tailoring patient management.
• Integrating diverse imaging data with advanced computational models can enhance predictive accuracy.

Purpose of the Study:

• To evaluate a novel nomogram combining clinical data with deep learning radiomics (DLRN) features from ultrasound and MRI.
• To assess the prognostic performance of this DLRN model in predicting survival, recurrence, and metastasis in TNBC patients receiving NAC.
• To compare the DLRN model's efficacy against other predictive approaches.

Main Methods:

• A retrospective, multicenter study involving 103 TNBC patients.
• Extraction of radiomics features from ultrasound and multi-sequence MRI images.
• Development and validation of a DLRN model using concordance index (C-index) for performance evaluation.
• Risk stratification based on the DLRN model to compare recurrence and metastasis rates.

Main Results:

• The DLRN model demonstrated strong predictive capability for disease-free survival (DFS) (C-index: 0.859-0.887) and moderate performance for overall survival (OS) (C-index: 0.800-0.811).
• The DLRN model outperformed other models for DFS prediction.
• Patients classified as low-risk exhibited significantly lower 3-year recurrence and metastasis rates compared to the high-risk group.

Conclusions:

• The preoperative DLRN model integrating ultrasound and MRI shows significant potential as a prognostic tool for TNBC patients undergoing NAC.
• The model aids in predicting recurrence, metastasis, and survival outcomes.
• The DLRN-derived risk score can facilitate individualized prognostic evaluation and preoperative risk stratification in clinical practice.