Can AI Non-Invasively Predict Lymph Node Metastasis?

Accurate axillary staging is highly essential for defining the prognosis and optimal treatment of breast cancer patients in India. Currently, sentinel lymph node biopsy remains the standard of care, but this invasive procedure carries risks such as lymphedema. Consequently, a new meta-analysis evaluates how artificial intelligence can non-invasively predict axillary lymph node metastasis. Specifically, the researchers compared magnetic resonance imaging (MRI) and ultrasound (US) models.

Predicting Axillary Lymph Node Metastasis with Radiomics and Deep Learning

To solve the limitations of manual imaging interpretation, researchers are developing computer-aided tools. In particular, deep learning (DL) and hand-crafted radiomics (HCR) are the two primary methodologies in clinical research today. HCR extracts quantitative texture features from medical images, whereas DL automatically learns hierarchical features directly from raw data. However, previous studies showed highly variable diagnostic performance across different cohorts. Therefore, Kiani et al. conducted a systematic review to pool the diagnostic accuracy of these models. In total, they analyzed 41 studies involving histopathological confirmation as the reference standard.

Key Diagnostic Findings from the Meta-Analysis

Overall, the meta-analysis demonstrated moderate to good diagnostic accuracy for predicting nodal status. During internal validation, models achieved a sensitivity of 0.79 and specificity of 0.78 (AUC = 0.84). In contrast, external validation showed a slightly lower sensitivity of 0.78 and specificity of 0.74 (AUC = 0.82). Furthermore, likelihood ratio analysis yielded a positive likelihood ratio of 3.0 and a negative likelihood ratio of 0.33. Consequently, these results suggest that AI models currently have limited clinical utility as standalone tools. Nonetheless, they can still provide helpful adjunctive risk stratification.

The Role of Combined Intra- and Peritumoral Regions

Interestingly, the meta-analysis revealed that model architecture and region-of-interest selection significantly affect diagnostic performance. For instance, ensemble approaches combining both DL and HCR showed much higher diagnostic performance than individual methods alone. Specifically, these hybrid models achieved an AUC of 0.88 in MRI and 0.92 in ultrasound. Additionally, models incorporating both intratumoral and peritumoral regions yielded higher AUCs than those using intratumoral features alone. Because of this, analyzing the tissue microenvironment surrounding the tumor appears essential for accurate predictions.

Clinical Relevance for Indian Practitioners

Currently, breast cancer represents the most common cancer among Indian women, often presenting at locally advanced stages. Therefore, rapid and accurate axillary staging is a critical component of treatment planning. Implementing AI models can optimize preoperative planning and potentially minimize unnecessary sentinel lymph node biopsies. However, before integrating these tools into daily practice, Indian clinicians must demand methodological standardization. Furthermore, developers must validate these algorithms on diverse Indian patient populations to ensure diagnostic generalizability. Ultimately, AI will likely serve as a powerful assistant rather than a replacement for skilled radiologists.

Frequently Asked Questions

Q1: Can AI models completely replace sentinel lymph node biopsies?

No. Currently, these AI models demonstrate moderate diagnostic accuracy and should serve as adjunctive tools rather than standalone replacements.

Q2: Why do ensemble AI models perform better in predicting metastasis?

Ensemble models combine deep learning features with hand-crafted radiomics, capturing both raw spatial relationships and engineered textures for superior diagnostic accuracy.

Q3: How does including the peritumoral region improve prediction accuracy?

The peritumoral region contains critical microenvironmental clues, such as lymphatic invasion and inflammation, which strongly correlate with metastasis.

References

1. Kiani I et al. Assessing the performance of deep learning and hand-crafted radiomics models using MRI and ultrasound in predicting axillary lymph node status in breast cancer: a systematic review and meta-analysis. Eur Radiol. 2026 Jun 19. doi: 10.1007/s00330-026-12682-6. PMID: 42319409.
2. Wang SR et al. Machine learning-driven ultrasound radiomics for assessing axillary lymph node burden in breast cancer. Frontiers. 2025 Feb 1. doi: 10.3389/fendo.2025.1548888.
3. Unal S et al. Deep learning-based non-invasive prediction of axillary lymph node metastasis in breast cancer: performance of the YOLO-v11 object detection algorithm. BMC Med Imaging. 2026 Mar 16. doi: 10.1186/s12880-026-01524-1.