4-Dimensional Model Can Predict Lymph Node Metastases in Breast Cancer

THURSDAY, May 23, 2024 (HealthDay News) -- A deep learning model using tumor dynamic contrast-enhanced breast magnetic resonance imaging (MRI) has high sensitivity for identifying lymph node metastasis, according to a study published online April 12 in Radiology: Imaging Cancer.

Dogan S. Polat, M.D., from the University of Texas Southwestern Medical Center in Dallas, and colleagues conducted a retrospective study of patients with newly diagnosed primary invasive breast cancer with known pathologic (pN) and clinical nodal (cN) status who underwent dynamic contrast-enhanced breast MRI. A four-dimensional (4D) convolutional neural network model was developed, integrating temporal information from dynamic image sets. The convolutional layers learned prognostic image features, which were then combined with clinicopathologic measures.

Data were analyzed from 350 female patients. The researchers found that for the 4D hybrid model, the area under the receiver operating characteristic curve, sensitivity, and specificity values were 0.87 and 89 and 76 percent, respectively, for differentiating pN0 from pN+, and 0.79 and 80 and 62 percent, respectively, for differentiating cN0 from cN+.

"We developed and validated a hybrid clinical and 4D MRI-based model that provides individualized prediction of axillary metastasis without the need for dedicated axillary imaging or invasive procedure," the authors write. "Our model is a safe and time-efficient tool, achieving noteworthy results compared with the existing methods."

Abstract/Full Text