Background: The status of sentinel lymph nodes is crucial for prognosis and treatment decisions in breast cancer patients. This study aimed to evaluate the predictive value of combined methylene blue and indocyanine green for sentinel lymph node metastasis in breast cancer.
Methods: This prospective cohort study enrolled 90 clinically node-negative breast cancer patients.
Deep learning based methods for medical images can be easily compromised by adversarial examples (AEs), posing a great security flaw in clinical decision-making. It has been discovered that conventional adversarial attacks like PGD which optimize the classification logits, are easy to distinguish in the feature space, resulting in accurate reactive defenses. To better understand this phenomenon and reassess the reliability of the reactive defenses for medical AEs, we thoroughly investigate the characteristic of conventional medical AEs.
View Article and Find Full Text PDF