Non-invasive prediction of axillary lymph node burden in breast cancer via intratumoral–peritumoral ultrasound radiomics–deep-learning fusion

Accurate preoperative assessment of axillary lymph node (ALN) metastatic burden is crucial for guiding individualized axillary management in breast cancer, particularly for node-positive patients. This study aimed to develop and validate a combined clinical, radiomics, and deep learning (DL) feature–based model using preoperative ultrasound (US) for non-invasive prediction of ALN metastatic burden (low <3 vs. high ≥3 positive nodes). This retrospective study included 322 breast cancer patients with ALN metastasis confirmed by ALND pathology. Patients were categorized into low- and high-burden groups and randomly split into training (n=225) and test (n=97) sets. Clinical, radiomics, and DL features were extracted from intratumoral regions, combined intratumoral and peritumoral (1–3 mm) regions, and whole US images. Machine learning algorithms were used to build clinical, radiomics, DL, and combined models, which were evaluated using AUC, sensitivity (SEN), specificity (SPE), positive predictive value (PPV), and negative predictive value (NPV). Clinical models showed limited performance in predicting ALN metastatic burden. Radiomics and DL models demonstrated improved performance, with further gains after incorporating peritumoral information. The combined model showed superior performance: a random forest-based model integrating clinical data, intratumoral and 3-mm peritumoral radiomics features, and ResNet50-derived DL features achieved the highest test-set AUC (0.855), with SEN/SPE/PPV/NPV of 0.875/0.867/0.854/0.811, respectively. An ultrasound-based model that integrates clinical, radiomics, and DL features—especially those from both intratumoral and peritumoral regions—can effectively and non-invasively predict ALN metastatic burden in breast cancer patients. This approach may serve as a valuable preoperative decision-support tool for identifying patients at high risk of nodal burden and supporting individualized axillary management.

Authors

Institutions

Publication Details

Journal
Breast Cancer Research
Published
2026-11-13
DOI
https://doi.org/10.1186/s13058-026-02407-4
Primary Topic
Radiomics and Machine Learning in Medical Imaging
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Non-invasive prediction of axillary lymph node burden in breast cancer via intratumoral–peritumoral ultrasound radiomics–deep-learning fusion

Yuehui Liao, Xiaoying Qiu, Xiaobo Lai, Lang Meng et al.
4 links
Breast Cancer Research
Radiomics and Machine Learning in Medical Imaging
article

Non-invasive prediction of axillary lymph node burden in breast cancer via intratumoral–peritumoral ultrasound radiomics–deep-learning fusion

Yuehui Liao, Xiaoying Qiu, Xiaobo Lai, Lang Meng, Wu Chen, Jing Cheng, Chunhong Yan, Yun Zheng, Weiji Yang, Tian’an Jiang, Jionghui Gu, Jing Chen
article en

Abstract

Accurate preoperative assessment of axillary lymph node (ALN) metastatic burden is crucial for guiding individualized axillary management in breast cancer, particularly for node-positive patients. This study aimed to develop and validate a combined clinical, radiomics, and deep learning (DL) feature–based model using preoperative ultrasound (US) for non-invasive prediction of ALN metastatic burden (low <3 vs. high ≥3 positive nodes). This retrospective study included 322 breast cancer patients with ALN metastasis confirmed by ALND pathology. Patients were categorized into low- and high-burden groups and randomly split into training (n=225) and test (n=97) sets. Clinical, radiomics, and DL features were extracted from intratumoral regions, combined intratumoral and peritumoral (1–3 mm) regions, and whole US images. Machine learning algorithms were used to build clinical, radiomics, DL, and combined models, which were evaluated using AUC, sensitivity (SEN), specificity (SPE), positive predictive value (PPV), and negative predictive value (NPV). Clinical models showed limited performance in predicting ALN metastatic burden. Radiomics and DL models demonstrated improved performance, with further gains after incorporating peritumoral information. The combined model showed superior performance: a random forest-based model integrating clinical data, intratumoral and 3-mm peritumoral radiomics features, and ResNet50-derived DL features achieved the highest test-set AUC (0.855), with SEN/SPE/PPV/NPV of 0.875/0.867/0.854/0.811, respectively. An ultrasound-based model that integrates clinical, radiomics, and DL features—especially those from both intratumoral and peritumoral regions—can effectively and non-invasively predict ALN metastatic burden in breast cancer patients. This approach may serve as a valuable preoperative decision-support tool for identifying patients at high risk of nodal burden and supporting individualized axillary management.

Breast Cancer Research
Zhejiang Chinese Medical University (CN), Peking Union Medical College Hospital (CN), First Affiliated Hospital Zhejiang University (CN)
Good health and well-being
Radiomics and Machine Learning in Medical Imaging
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon