🤖 Generated with AI assistance · Does not replace individualized medical advice.
AI for Breast Cancer Lymphedema Prediction: A Systematic Review
AI Transparency Notice: This post was generated by an AI assistant to summarize published medical literature. It is for informational purposes only and does not constitute medical advice, diagnosis, or treatment. Always consult a qualified healthcare provider for personal medical concerns.
Lead
A systematic review of seven studies (5,284 patients) evaluates AI models for predicting breast cancer-related lymphedema (BCRL). While machine learning models show promising accuracy (mean 87.2%), the field remains exploratory due to small sample sizes, lack of external validation, and data heterogeneity, necessitating larger prospective trials before clinical integration.
Key Facts
| Metric | Value |
|---|---|
| Total Patients | 5,284 PMID: 42840800 |
| Studies Included | 7 (2018–2024) PMID: 42840800 |
| Mean Accuracy | 87.2% (Range: 81%–93.75%) PMID: 42840800 |
| Mean Sensitivity | 85.5% PMID: 42840800 |
| Mean Specificity | 87.6% PMID: 42840800 |
Why It Matters / Context
Breast cancer-related lymphedema (BCRL) is a common, often debilitating complication of breast cancer treatment, significantly impairing quality of life and physical function. Early detection is critical for initiating preventive interventions such as compression therapy or manual lymphatic drainage. Traditional risk assessment relies on clinical judgment and established risk factors, but these methods may lack the granularity needed for personalized risk stratification.
Artificial Intelligence (AI), specifically machine learning (ML) and deep learning (DL), offers a potential pathway to more precise prediction by integrating complex, multivariate data. This systematic review, conducted in July 2024, synthesizes the current evidence on AI’s ability to predict BCRL, comparing reported performance against traditional methods where available. The findings suggest that while AI models are technically capable of high accuracy, they have not yet reached the maturity required for routine clinical deployment.
Details
The review identified seven studies published between 2018 and 2024, encompassing a total of 5,284 patients PMID: 42840800. Five of these studies reported specific performance metrics for BCRL classification. The aggregated performance of these AI models is summarized below.
| Performance Metric | Reported Value |
|---|---|
| Accuracy | 81%–93.75% (Mean: 87.2%) PMID: 42840800 |
| Sensitivity | Mean: 85.5% PMID: 42840800 |
| Specificity | Mean: 87.6% PMID: 42840800 |
Key Predictors Identified:
The studies consistently identified several clinical and treatment-related factors as strong predictors of BCRL risk:
- Higher Body Mass Index (BMI): A well-established risk factor, consistently highlighted in AI models PMID: 42840800.
- Lymph Node Burden: Greater extent of axillary lymph node dissection or involvement PMID: 42840800.
- Adjuvant Therapy: Receipt of chemotherapy or radiotherapy PMID: 42840800.
- Age and Neoadjuvant Therapy: Some studies noted that younger age and the use of neoadjuvant chemotherapy were associated with higher predicted risk PMID: 42840800.
The review notes that AI models generally outperformed or matched traditional diagnostic methods in the reported studies, but direct head-to-head comparisons were limited by the heterogeneity of the included studies.
Limitations
Critical Limitations for Clinical Interpretation:
* Small Sample Sizes: Individual studies within the review had limited patient numbers, reducing statistical power and generalizability PMID: 42840800.
* Data Heterogeneity: Variability in study designs, data sources, and feature sets makes direct comparison difficult PMID: 42840800.
* Lack of External Validation: Most models were trained and tested on the same or similar datasets, with absent external validation, raising concerns about overfitting and real-world applicability PMID: 42840800.
* Inconsistent Calibration: The review highlights inconsistent calibration of the models, meaning the predicted probabilities may not align well with observed outcomes PMID: 42840800.
* Exploratory Status: The authors explicitly state that current models remain exploratory rather than clinically established PMID: 42840800.
Clinical Takeaway
Clinicians should not yet use AI-based tools for routine BCRL risk stratification in practice, as the current evidence base lacks the external validation and standardized evaluation necessary for safe clinical integration.
Full Reference
Mortada H, Al-Khafaji MQM, Jefri ZE, Sultan F, Alrobaiea S, Alshomer F. Artificial Intelligence for Prediction of Breast Cancer-Related Lymphedema: A Systematic Review. Plast Reconstr Surg Glob Open. 2026. PMID: 42840800 [DOI: 10.1097/GOX.0000000000008132]