Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-1792
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-10-18T06:54:38.842659Z
status: published
lens: trace
sector: health
headline: Artificial Intelligence for Prediction of Breast Cancer-Related Lymphedema: A Systematic Review
dek: Background Breast cancer-related lymphedema (BCRL) impairs quality of life and function, and early detection is crucial for preventive intervention. AI models, particularly machine learning and deep learning, hold promise for improving BCRL prediction and personalized risk stratification. This systematic review evaluates AI applications for predicting BCRL and, where available, summarizes how their reported performance compared with traditional diagnostic or non-AI methods. Methods This systematic review followe…
gain_title: Across five studies reporting classification results, AI models predicted breast cancer-related lymphedema with a mean accuracy of 87.2%, mean sensitivity of 85.5%, and mean specificity of 87.6%.
problem_title: The same BCRL prediction models have small samples, heterogeneous data, no external validation, and inconsistent calibration, so they are not yet clinically established.
trace_subject: AI prediction models for breast cancer-related lymphedema
gain_reading: Across five studies reporting classification results, AI models predicted breast cancer-related lymphedema with a mean accuracy of 87.2%, mean sensitivity of 85.5%, and mean specificity of 87.6%.
gain_evidence: AI models achieved accuracies of 81%-93.75% (mean 87.2%), mean sensitivity 85.5%, and mean specificity 87.6%
problem_reading: The same BCRL prediction models have small samples, heterogeneous data, no external validation, and inconsistent calibration, so they are not yet clinically established.
problem_evidence: Small sample sizes, data heterogeneity, absent external validation, and inconsistent calibration | current models remain exploratory rather than clinically established
quick_read: A systematic review searched major medical databases in July 2024 and included seven studies published between 2018 and 2024, covering 5284 patients. It examined how machine learning and deep learning models predict breast cancer-related lymphedema, a condition that harms function and quality of life.

The findings suggest AI could support earlier detection and more personalized risk stratification, which matters because early detection is key to preventive intervention. However, the evidence is not strong enough for clinical use, and larger multi-institutional datasets and prospective validation are still needed.
limitation: The authors state that current models are exploratory, limited by small samples, heterogeneous data, absent external validation, and inconsistent calibration, so prospective validation is needed before clinical use.
tag: Dual reading
key_points: Five of the seven studies reported classification performance, with accuracies ranging from 81% to 93.75%. | Mean sensitivity across those studies was 85.5% and mean specificity was 87.6%. | Commonly reported predictors were higher body mass index, greater lymph node burden, and chemotherapy or radiotherapy. | Younger age and neoadjuvant chemotherapy were linked to higher predicted risk in some studies.
rundown: The review followed PRISMA guidelines and searched MEDLINE, EMBASE, the Cochrane Library, and Google Scholar. Studies were chosen using predefined inclusion and exclusion criteria, and accuracy, sensitivity, and specificity were extracted.

Where available, the authors summarized how reported AI performance compared with traditional diagnostic or non-AI methods. The abstract does not give the comparison results.

The authors conclude that the models are promising but exploratory, and call for standardized evaluation and prospective validation before they are integrated into clinical care.
sources:
- peer_reviewed | Plastic and Reconstructive Surgery - Global Open | https://doi.org/10.1097/gox.0000000000008132 | 2026-10-06
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

c4c96278481d762c06a4ceab4c6ad9860fa338c3ed169c425bc2d6492262bea3

previous

0000000000000000000000000000000000000000000000000000000000000000