We propose GeoNet, a jointly unsupervised learning framework for monocular depth, optical flow and egomotion estimation from videos. The three components are coupled by the nature of 3D scene geometry, jointly learned by our framework in an end-to-end manner. Specifically, geometric relationships are extracted over the predictions of individual modules and then combined as an image reconstruction loss, reasoning about static and dynamic scene parts separately. Furthermore, we propose an adaptive geometric consistency loss to increase robustness towards outliers and non-Lambertian regions, which resolves occlusions and texture ambiguities effectively. Experimentation on the KITTI driving dataset reveals that our scheme achieves state-of-the-art results in all of the three tasks, performing better than previously unsupervised methods and comparably with supervised ones.

Background: Hepatocellular carcinoma (HCC) (about 85–90% of primary liver cancer) is particularly prevalent in China because of the high prevalence of chronic hepatitis B infection. HCC is the fourth most common malignancy and the third leading cause of tumor-related deaths in China. It poses a significant threat to the life and health of Chinese people. Summary: This guideline presents official recommendations of the National Health and Family Planning Commission of the People’s Republic of China on the surveillance, diagnosis, staging, and treatment of HCC occurring in China. The guideline was written by more than 50 experts in the field of HCC in China (including liver surgeons, medical oncologists, hepatologists, interventional radiologists, and diagnostic radiologists) on the basis of recent evidence and expert opinions, balance of benefits and harms, cost-benefit strategies, and other clinical considerations. Key Messages: The guideline presents the Chinese staging system, and recommendations regarding patients with HCC in China to ensure optimum patient outcomes.

Despite worldwide promising clinical outcome of CD19 CART therapy, relapse after this therapy is associated with poor prognosis and has become an urgent problem to be solved. We conducted a CD22 CAR T-cell therapy in 34 relapsed or refractory (r/r) BALL pediatric and adult patients who failed from previous CD19 CAR T-cell therapy. Complete remission (CR) or CR with incomplete count recovery (CRi) was achieved in 24 of 30 patients (80%) that could be evaluated on day 30 after infusion, which accounted for 70.5% of all 34 enrolled patients. Most patients only experienced mild cytokine-release syndrome and neurotoxicity. Seven CR patients received no further treatment, and 3 of them remained in remission at 6, 6.6, and 14 months after infusion. Eleven CR patients were promptly bridged to transplantation, and 8 of them remained in remission at 4.6 to 13.3 months after transplantation, resulted in 1-year leukemia-free survival rate of 71.6% (95% CI, 44.2-99.0). CD22 antigen loss or mutation was not observed to be associated with relapsed patients. Our study demonstrated that our CD22 CAR T-cells was highly effective in inducing remission in r/r BALL patients, and also provided a precious window for subsequent transplantation to achieve durable remission.

Explicit representations of the global match distributions of pixel-wise correspondences between pairs of images are desirable for uncertainty estimation and downstream applications. However, the computation of the match density for each pixel may be prohibitively expensive due to the large number of candidates. In this paper, we propose Hierarchical Discrete Distribution Decomposition (HD 3 ), a framework suitable for learning probabilistic pixel correspondences in both optical flow and stereo matching. We decompose the full match density into multiple scales hierarchically, and estimate the local matching distributions at each scale conditioned on the matching and warping at coarser scales. The local distributions can then be composed together to form the global match density. Despite its simplicity, our probabilistic method achieves state-ofthe-art results for both optical flow and stereo matching on established benchmarks. We also find the estimated uncertainty is a good indication of the reliability of the predicted correspondences.