Machine Learning Assisted Magnetic Resonance Imaging for the Prediction of Lymph Node Metastasis in Breast Cancer: A Systematic Review and Meta-Analysis

Purpose: Breast cancer, particularly when it involves lymph node metastasis, substantially affects patient outcomes and treatment strategies, a domain where Magnetic Resonance Imaging (MRI) is essential yet often hindered by subjective interpretations. Integrating machine learning (ML) with MRI is promising for enhancing diagnostic precision, especially in predicting lymph node metastasis. This study assesses the efficacy of ML-enhanced MRI in predicting lymph node metastasis in breast cancer patients.

Methods/Materials: A systematic search up to June 2023 in PubMed, Scopus, and Web of Science identified studies applying ML algorithms to MRI data for breast cancer lymph node metastasis prediction. Statistical analysis using R software (version 4.0.3) with the 'mada' package pooled key diagnostic measures (sensitivity, specificity, false-positive rate, diagnostic odds ratio, and Likelihood Ratios) with 95% Confidence Intervals for precise evaluation.

Results: The analysis incorporated 9 studies encompassing 1395 patients, with 553 confirmed metastasis cases. The sensitivity of ML-assisted MRI for predicting lymph node metastasis was 85.2% (95% CI: 78.1-90.3%, I2=76.8%), while specificity was 81.2% (95% CI: 78.5-83.7%, I2=0%), indicating effective identification of patients without metastasis. The false-positive rate was relatively modest at 18.8% (95% CI: 16.3-21.5%), limiting the likelihood of incorrect metastasis identification. The diagnostic odds ratio was 24.97 (95% CI: 14.94-41.71), illustrating a robust distinction capability. The positive Likelihood Ratio (LR+) was 4.54 (95% CI: 3.87-5.32), suggesting a high probability of accurate positive diagnosis, and the negative Likelihood Ratio (LR-) was 0.18 (95% CI: 0.12-0.28), indicating a low chance of false negatives.

Conclusions: ML, in conjunction with MRI, substantially boosts the precision of lymph node metastasis prediction in breast cancer. This advancement improves diagnostic reliability and facilitates a move towards more tailored and precise treatment approaches. The findings highlight the importance of advancing and standardizing ML methodologies for more comprehensive clinical implementation in cancer imaging and treatment planning.