The sudden deterioration of patients with novel coronavirus disease 2019 (COVID-19) into critical illness is of major concern. It is imperative to identify these patients early. We show that a deep learning-based survival model can predict the risk of COVID-19 patients developing critical illness based on clinical characteristics at admission. We develop this model using a cohort of 1590 patients from 575 medical centers, with internal validation performance of concordance index 0.894 We further validate the model on three separate cohorts from Wuhan, Hubei and Guangdong provinces consisting of 1393 patients with concordance indexes of 0.890, 0.852 and 0.967 respectively. This model is used to create an online calculation tool designed for patient triage at admission to identify patients at risk of severe illness, ensuring that patients at greatest risk of severe illness receive appropriate care as early as possible and allow for effective allocation of health resources.

No abstract

Microsatellite instability (MSI) has been approved as a pan-cancer biomarker for immune checkpoint blockade (ICB) therapy. However, current MSI identification methods are not available for all patients. We proposed an ensemble multiple instance deep learning model to predict microsatellite status based on histopathology images, and interpreted the pathomics-based model with multi-omics correlation. Methods: Two cohorts of patients were collected, including 429 from The Cancer Genome Atlas (TCGA-COAD) and 785 from an Asian colorectal cancer (CRC) cohort (Asian-CRC). We established the pathomics model, named Ensembled Patch Likelihood Aggregation (EPLA), based on two consecutive stages: patch-level prediction and WSI-level prediction. The initial model was developed and validated in TCGA-COAD, and then generalized in Asian-CRC through transfer learning. The pathological signatures extracted from the model were analyzed with genomic and transcriptomic profiles for model interpretation. Results: The EPLA model achieved an area-under-the-curve (AUC) of 0.8848 (95% CI: 0.8185-0.9512) in the TCGA-COAD test set and an AUC of 0.8504 (95% CI: 0.7591-0.9323) in the external validation set Asian-CRC after transfer learning. Notably, EPLA captured the relationship between pathological phenotype of poor differentiation and MSI ( P < 0.001). Furthermore, the five pathological imaging signatures identified from the EPLA model were associated with mutation burden and DNA damage repair related genotype in the genomic profiles, and antitumor immunity activated pathway in the transcriptomic profiles. Conclusions: Our pathomics-based deep learning model can effectively predict MSI from histopathology images and is transferable to a new patient cohort. The interpretability of our model by association with pathological, genomic and transcriptomic phenotypes lays the foundation for prospective clinical trials of the application of this artificial intelligence (AI) platform in ICB therapy.

PURPOSE Camrelizumab is an antibody against programmed death protein 1. We assessed the activity and safety of camrelizumab plus apatinib, a tyrosine kinase inhibitor of vascular endothelial growth factor receptor-2, in patients with advanced cervical cancer. METHODS This multicenter, open-label, single-arm, phase II study enrolled patients with advanced cervical cancer who progressed after at least one line of systemic therapy. Patients received camrelizumab 200 mg every 2 weeks and apatinib 250 mg once per day. The primary end point was objective response rate (ORR) assessed by investigators per RECIST version 1.1. Key secondary end points were progression-free survival (PFS), overall survival (OS), duration of response, and safety. RESULTS Forty-five patients were enrolled and received treatment. Median age was 51.0 years (range, 33-67 years), and 57.8% of patients had previously received two or more lines of chemotherapy for recurrent or metastatic disease. Ten patients (22.2%) had received bevacizumab. Median follow-up was 11.3 months (range, 1.0-15.5 months). ORR was 55.6% (95% CI, 40.0% to 70.4%), with two complete and 23 partial responses. Median PFS was 8.8 months (95% CI, 5.6 months to not estimable). Median duration of response and median OS were not reached. Treatment-related grade 3 or 4 adverse events (AEs) occurred in 71.1% of patients, and the most common AEs were hypertension (24.4%), anemia (20.0%), and fatigue (15.6%). The most common potential immune-related AEs included grade 1-2 hypothyroidism (22.2%) and reactive cutaneous capillary endothelial proliferation (8.9%). CONCLUSION Camrelizumab plus apatinib had promising antitumor activity and manageable toxicities in patients with advanced cervical cancer. Larger randomized controlled trials are warranted to validate our findings.

No abstract

No abstract

In this paper, we propose a novel end-to-end approach for scalable visual search infrastructure. We discuss the challenges we faced for a massive volatile inventory like at eBay and present our solution to overcome those. We harness the availability of large image collection of eBay listings and state-of-the-art deep learning techniques to perform visual search at scale. Supervised approach for optimized search limited to top predicted categories and also for compact binary signature are key to scale up without compromising accuracy and precision. Both use a common deep neural network requiring only a single forward inference. The system architecture is presented with in-depth discussions of its basic components and optimizations for a trade-off between search relevance and latency. This solution is currently deployed in a distributed cloud infrastructure and fuels visual search in eBay ShopBot and Close5. We show benchmark on ImageNet dataset on which our approach is faster and more accurate than several unsupervised baselines. We share our learnings with the hope that visual search becomes a first class citizen for all large scale search engines rather than an afterthought.Comment: To appear in 23rd SIGKDD Conference on Knowledge Discovery and Data Mining (KDD), 2017. A demonstration video can be found at https://youtu.be/iYtjs32vh4

Satirical news is considered to be entertainment, but it is potentially deceptive and harmful. Despite the embedded genre in the article, not everyone can recognize the satirical cues and therefore believe the news as true news. We observe that satirical cues are often reflected in certain paragraphs rather than the whole document. Existing works only consider documentlevel features to detect the satire, which could be limited. We consider paragraphlevel linguistic features to unveil the satire by incorporating neural network and attention mechanism. We investigate the difference between paragraph-level features and document-level features, and analyze them on a large satirical news dataset. The evaluation shows that the proposed model detects satirical news effectively and reveals what features are important at which level.