Prediction of Regional Lymph Node Metastasis from the Clinicopathological Features of Breast Carcinoma: Application of Deep Learning

Objective: Breast cancer is the leading cause of cancer-related mortality in women globally, with invasive ductal carcinoma (IDC) as the predominant subtype. Regional lymph node (LN) metastasis significantly impacts prognosis, staging, and treatment strategies. However, the role of deep learning in predicting LN metastasis is underexplored. To develop and evaluate a deep learning model leveraging clinicopathological features for predicting LN metastasis in IDC patients, with an aim to enhance diagnostic accuracy and reduce reliance on invasive methods.Material and Methods: A cross-sectional study was conducted on 351 IDC cases from a tertiary-care hospital. Input variables included clinicopathological features: age, tumor size, modified Bloom-Richardson grade, ER, PR, HER2 receptor status, Ki-67 index, and microvessel density (MVD). LN status was dichotomized using a cut-off ratio of 0.3. A neural network model with an input layer of 8 neurons, 3 hidden layers (50 neurons each), and ReLU activation was developed. Data were split into training (70%) and test (30%) sets. Predictive accuracy was evaluated using standard performance metrics.Results: The mean age was 46.4±11.29 years and tumor volume averaged 44.9 cm³. Low ER (35.6%) and PR (26.8%) positivity rates were observed, with HER2 positivity at 21.7%. The model achieved 78.3% accuracy in predicting LN metastasis. The F1 score of the model was 0.83.Conclusion: The study demonstrates the utility of deep learning models in predicting LN metastasis using clinicopathological data. With 78.3% accuracy, the model highlights AI’s potential in oncology diagnostics, supporting personalized treatment approaches. Further integration of imaging and molecular data could enhance model performance and clinical applicability.