Thursday, December 4 | 1:50 p.m.-2:00 p.m. | R6-SSBR11-3 | Room S406A
Here, attendees will find out how a multimodal ultrasound-based deep-learning (DL) model can detect metastatic axillary lymph nodes in breast cancer patients.
The model improved diagnostic specificity and could reduce unnecessary biopsies, according to the prospective study being presented by Manali Saini, PhD, from the Mayo Clinic in Rochester, MN.
The model used tri-modal ultrasound imaging that consisted of conventional ultrasound B-mode images with shear-wave elastography (SWE) and color Doppler images. The team used these images to train a customized DL network with the goal of classifying axillary lymph node status. It validated the method using in-vivo datasets comprising tri-modal ultrasound images of axillary lymph nodes from 174 women with breast cancer.
The model achieved high marks, including a classification accuracy of 93%, 100% specificity, an area under the curve (AUC) of 0.99, 90% sensitivity, an F1 score of 0.95, and an area under the precision-recall curve of 0.99. The model delivered a higher classification accuracy than other DL models the team compared its model with, which included ranges of 67% to 88%. This also outperformed B-mode images alone (81%), SWE alone (88%), and color Doppler ultrasound alone (84%).
“The finding shows its ability to accurately identify all metastatic [axillary lymph node] cases, reducing unnecessary biopsies,” the team highlighted in its abstract.
Find out more about the model by attending this session.
