Bi-modal ultrasound radiomics and habitat analysis enhanced the pre-operative prediction of axillary lymph node burden in patients with early-stage breast cancer

Area of Science:

• Radiology
• Medical Imaging
• Oncology

Background:

• Accurate prediction of axillary lymph node (ALN) status is crucial for staging and treatment planning in early-stage breast cancer.
• Current methods for ALN assessment can be invasive and may not fully capture tumor heterogeneity.

Purpose of the Study:

• To evaluate the efficacy of habitat analysis-based bi-modal ultrasound radiomics in predicting ALN status in early-stage breast cancer patients.
• To identify a non-invasive and accurate method for ALN status prediction.

Main Methods:

• A total of 206 early-stage breast cancer patients were included, divided into training (165) and testing (41) cohorts.
• Radiomic features were extracted from B-mode ultrasound (BMUS) and shear wave elastography (SWE) images using habitat analysis and K-means clustering.
• Eleven machine learning models were applied, comparing clinicopathological factors, omics models, and habitat models.

Main Results:

• Habitat analysis identified distinct subregions within BMUS and SWE images.
• The NaiveBayes model, utilizing SWE habitat radiomics, demonstrated the highest prediction performance in the test cohort with an Area Under the Curve (AUC) of 0.953.
• This model outperformed clinicopathological factors and other omics/habitat models.

Conclusions:

• Habitat analysis of ultrasound images can visualize intratumoral heterogeneity in breast lesions.
• Machine learning models incorporating SWE radiomics and habitat analysis show significant potential for predicting lymph node burden in early-stage breast cancer.
• This approach offers a promising non-invasive tool for clinical decision support.