No abstract

Highly expressive models such as deep neural networks (DNNs) have been widely applied to various applications. However, recent studies show that DNNs are vulnerable to adversarial examples, which are carefully crafted inputs aiming to mislead the predictions. Currently, the majority of these studies have focused on perturbation added to image pixels, while such manipulation is not physically realistic. Some works have tried to overcome this limitation by attaching printable 2D patches or painting patterns onto surfaces, but can be potentially defended because 3D shape features are intact. In this paper, we propose meshAdv to generate "adversarial 3D meshes" from objects that have rich shape features but minimal textural variation. To manipulate the shape or texture of the objects, we make use of a differentiable renderer to compute accurate shading on the shape and propagate the gradient. Extensive experiments show that the generated 3D meshes are effective in attacking both classifiers and object detectors. We evaluate the attack under different viewpoints. In addition, we design a pipeline to perform black-box attack on a photorealistic renderer with unknown rendering parameters.

No abstract

Cancer therapy is unsatisfactory as it typically has serious side effects, because normal cells in healthy organs are destroyed along with the tumor. Thus, researchers have tried to develop effective therapies with minimal side effects. One such method is to use nanotechnology to carry the drugs or therapeutic agents to the tumor region by secure encapsulation without leakage. Once the nanomedicine enters the target tumor site, it can release therapeutic agents in an effective manner. Accordingly, various nanomedicines have been developed to enhance the efficiency of cancer therapy and minimize the systematic toxicity. Here, we provide an overview and discuss the different types of responsive nanomedicines including physically, chemically, biologically, dual, and multi‐responsive nanomedicines, for the in situ release of cargos in recent years. We propose critical considerations that must be considered for the design of excellent stimuli‐nanomedicine. Furthermore, the possible directions for the development of successful stimuli‐responsive (smart) nanomedicine are highlighted. With the development of responsive nanomedicines, precise and personalized nanomedicine will be realized with great promise in the future. This article is categorized under: Therapeutic Approaches and Drug Discovery > Nanomedicine for Oncologic Disease Nanotechnology Approaches to Biology > Nanoscale Systems in Biology

Background The Liver Imaging Reporting and Data System (LI‐RADS) is widely adopted for noninvasive diagnosis of hepatocellular carcinoma (HCC). It's updated to version 2018 recently, with some major changes compared with v2017. However, the diagnostic performance of LI‐RADS v2018 and its difference with v2017 are yet to be validated. Purpose To compare the diagnostic performances of LI‐RADS on MR for diagnosing HCC between v2017 and v2018. Study Type Retrospective. Subjects In all, 181 patients with 217 hepatic observations (146 HCCs, 16 non‐HCC malignancies and 55 benign lesions) with liver MRI and pathological or follow‐up imaging diagnoses. Field Strength/Sequence 1.5 T or 3 T MRI. Dual‐echo T1WI, T2WI, diffusion‐weighted imaging, and a liver acquisition with volume acceleration. AssessmentImages were independently interpreted by three radiologists, and then in consensus for observations with different LR categories, according to LI‐RADS v2017 and v2018, separately. Statistical Tests Sensitivity, specificity, accuracy, positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (+LR), and Youden index. Results When adopting LR‐5 as a predictor of HCC, the sensitivity (80.8% vs. 71.2%), NPV (69.6% vs. 60.7%), and accuracy (83.9% vs. 77.9%) were all increased for LI‐RADS v2018 compared with v2017, with a greater Youden index (0.709 vs. 0.627). However, the diagnostic performances of MRI for diagnosing HCC were not changed while adopting LR‐4/5 as a predictor. The threshold growths of 76% (19/25) observations in v2017 were revised to subthreshold growth in v2018, and 16 LR‐4 observations in v2017 were changed to LR‐5 based on v2018. Data Conclusion The diagnostic performance of LI‐RADS v2018 for diagnosing HCC is superior to v2017, with a greater sensitivity, NPV, and accuracy. The revisions in v2018 mainly affect the categorization when adopting LR‐5 as a predictor of HCC. Level of Evidence: 4 Technical Efficacy Stage: 2 J. Magn. Reson. Imaging 2019;50:746–755.