Machine Learning Prediction of Lymph Node Metastasis in Breast Cancer: Performance of a Multi-institutional MRI-based...

Area of Science:

• Radiology and Medical Imaging
• Oncology
• Artificial Intelligence in Medicine

Background:

• Accurate prediction of breast cancer nodal metastasis is crucial for treatment planning and patient outcomes.
• Current methods for assessing lymph node status can be invasive or lack sufficient accuracy.
• Dynamic contrast-enhanced (DCE) breast MRI offers rich temporal and spatial information for tumor characterization.

Purpose of the Study:

• To develop and validate a custom deep convolutional neural network (CNN) for noninvasive prediction of breast cancer nodal metastasis.
• To integrate clinicopathologic data with image features from DCE-MRI for enhanced prognostic prediction.
• To assess the performance of the developed model in differentiating between node-negative and node-positive disease.

Main Methods:

• A retrospective study analyzed data from 350 female patients with primary invasive breast cancer.
• A four-dimensional (4D) CNN model was developed, incorporating temporal information from DCE-MR image sets.
• The model combined learned image features with clinicopathologic data (age, ER/HER2 status, Ki-67, tumor grade) to predict nodal status (cN0/cN+ and pN0/pN+).
• Performance was evaluated using the area under the receiver operating characteristic curve (AUC) with nested cross-validation.

Main Results:

• The 4D hybrid CNN model achieved an AUC of 0.87 for predicting pathologic nodal status (pN0 vs pN+).
• For clinical nodal status (cN0 vs cN+), the model achieved an AUC of 0.79.
• The model demonstrated high sensitivity in identifying breast cancer lymph node metastasis, with values of 89% for pN+ and 80% for cN+.

Conclusions:

• The proposed deep learning model effectively utilizes DCE-MR images for predicting breast cancer nodal metastasis.
• The model shows significant promise as a noninvasive clinical decision support tool for oncologists.
• Further validation and integration into clinical workflows could enhance breast cancer management and patient care.