Nomogram Utilizing ABVS Radiomics and Clinical Factors for Predicting ≤ 3 Positive Axillary Lymph Nodes in HR+ /HER2- Breast Cancer with 1-2 Positive Sentinel Nodes

Bin Hu; Yanjun Xu; Huiling Gong; Lang Tang; Lihong Wang; Hongchang Li

doi:10.1016/j.acra.2024.01.026

Nomogram Utilizing ABVS Radiomics and Clinical Factors for Predicting ≤ 3 Positive Axillary Lymph Nodes in HR+ /HER2- Breast Cancer with 1-2 Positive Sentinel Nodes

Acad Radiol. 2024 Feb 20:S1076-6332(24)00049-7. doi: 10.1016/j.acra.2024.01.026. Online ahead of print.

Authors

Bin Hu¹, Yanjun Xu², Huiling Gong³, Lang Tang³, Lihong Wang³, Hongchang Li⁴

Affiliations

¹ Department of Ultrasound, Minhang Hospital, Fudan University, 170 Xinsong Rd, Shanghai 201199, China. Electronic address: hubin_10945@fudan.edu.cn.
² Department of Ultrasonography, Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai, China; Shanghai Institute of Ultrasound in Medicine, Shanghai, China.
³ Department of Ultrasound, Minhang Hospital, Fudan University, 170 Xinsong Rd, Shanghai 201199, China.
⁴ Department of General Surgery, Institute of Fudan-Minhang Academic Health System, Minhang Hospital, Fudan University, Shanghai, China.

PMID: 38383259
DOI: 10.1016/j.acra.2024.01.026

Abstract

Background: In HR+ /HER2- breast cancer patients with ≤ 3 positive axillary lymph nodes (ALNs), genomic tests can streamline chemotherapy decisions. Current studies, centered on tumor metrics, miss broader patient insights. Automated Breast Volume Scanning (ABVS) provides advanced 3D imaging, and its potential synergy with radiomics for ALN evaluation is untapped.

Objective: This study sought to combine ABVS radiomics and clinical characteristics in a nomogram to predict ≤ 3 positive ALNs in HR+ /HER2- breast cancer patients with 1-2 positive sentinel lymph nodes (SLNs), guiding clinicians in genetic test candidate selection.

Methods: We enrolled 511 early-stage breast cancer patients: 362 from A Hospital for training and 149 from B Hospital for validation. Using LASSO logistic regression, primary features were identified. A clinical-radiomics nomogram was developed to predict the likelihood of ≤ 3 positive ALNs in HR+ /HER2- patients with 1-2 positive SLNs. We assessed the discriminative capability of the nomogram using the ROC curve. The model's calibration was confirmed through a calibration curve, while its fit was evaluated using the Hosmer-Lemeshow (HL) test. To determine the clinical net benefits, we employed the Decision Curve Analysis (DCA).

Results: In the training group, 81.2% patients had ≤ 3 metastatic ALNs, and 83.2% in the validation group. We developed a clinical-radiomics nomogram by analyzing clinical characteristics and rad-scores. Factors like positive SLNs (OR=0.077), absence of negative SLNs (OR=11.138), lymphovascular invasion (OR=0.248), and rad-score (OR=0.003) significantly correlated with ≤ 3 positive ALNs. The clinical-radiomics nomogram, with an AUC of 0.910 in training and 0.882 in validation, outperformed the rad-score-free clinical nomogram (AUCs of 0.796 and 0.782). Calibration curves and the HL test (P values 0.688 and 0.691) confirmed its robustness. DCA showed the clinical-radiomics nomogram provided superior net benefits in predicting ALN burden across specific threshold probabilities.

Conclusion: We developed a clinical-radiomics nomogram that integrated radiomics from ABVS images and clinical data to predict the presence of ≤ 3 positive ALNs in HR+ /HER2- patients with 1-2 positive SLNs, aiding oncologists in identifying candidates for genomic tests, bypassing ALND. In the era of precision medicine, combining genomic tests with SLN biopsy refines both surgical and systemic patient treatments.

Keywords: Axillary lymph node dissection; Axillary lymph nodes; Breast cancer; Nomogram; Radiomics.