DAN: A Segmentation-Free Document Attention Network for Handwritten Document Recognition.

Unconstrained handwritten text recognition is a challenging computer vision task. It is traditionally handled by a two-step approach, combining line segmentation followed by text line recognition. For the first time, we propose an end-to-end segmentation-free architecture for the task of handwritten document recognition: the Document Attention Network. In addition to text recognition, the model is trained to label text parts using begin and end tags in an XML-like fashion. This model is made up of an FCN encoder for feature extraction and a stack of transformer decoder layers for a recurrent token-by-token prediction process. It takes whole text documents as input and sequentially outputs characters, as well as logical layout tokens. Contrary to the existing segmentation-based approaches, the model is trained without using any segmentation label. We achieve competitive results on the READ 2016 dataset at page level, as well as double-page level with a CER of 3.43% and 3.70%, respectively. We also provide results for the RIMES 2009 dataset at page level, reaching 4.54% of CER. We provide all source code and pre-trained model weights at https://github.com/FactoDeepLearning/DAN.

End-to-End Handwritten Paragraph Text Recognition Using a Vertical Attention Network.

Unconstrained handwritten text recognition remains challenging for computer vision systems. Paragraph text recognition is traditionally achieved by two models: the first one for line segmentation and the second one for text line recognition. We propose a unified end-to-end model using hybrid attention to tackle this task. This model is designed to iteratively process a paragraph image line by line. It can be split into three modules. An encoder generates feature maps from the whole paragraph image. Then, an attention module recurrently generates a vertical weighted mask enabling to focus on the current text line features. This way, it performs a kind of implicit line segmentation. For each text line features, a decoder module recognizes the character sequence associated, leading to the recognition of a whole paragraph. We achieve state-of-the-art character error rate at paragraph level on three popular datasets: 1.91% for RIMES, 4.45% for IAM and 3.59% for READ 2016. Our code and trained model weights are available at https://github.com/FactoDeepLearning/VerticalAttentionOCR.

Combining Structure and Parameter Adaptation of HMMs for Printed Text Recognition.

We present two algorithms that extend existing HMM parameter adaptation algorithms (MAP and MLLR) by adapting the HMM structure. This improvement relies on a smart combination of MAP and MLLR with a structure optimization procedure. Our algorithms are semi-supervised: to adapt a given HMM model on new data, they require little labeled data for parameter adaptation and a moderate amount of unlabeled data to estimate the criteria used for HMM structure optimization. Structure optimization is based on state splitting and state merging operations and proceeds so as to optimize either the likelihood or a heuristic criterion. Our algorithms are successfully applied to the recognition of printed characters by adapting the HMM character models of a polyfont printed text recognizer to new fonts. Our experiments involve a total of 1,120,000 real and 3,100,000 synthetic character images and concern a set of 89 HMM models. A comparison of our results with those of state-of-the-art adaptation algorithms (MAP and MLLR) shows a significant increase in the accuracy of character recognition.

Phase I study investigating everolimus combined with sorafenib in patients with advanced hepatocellular carcinoma.

BACKGROUND & AIMS: Sorafenib is the only therapy shown to improve overall survival in advanced hepatocellular carcinoma (HCC). Combination therapy targeting multiple signaling pathways may improve outcomes. This phase I study was designed to determine the maximum tolerated dose (MTD) of everolimus given with sorafenib 400mg twice daily in patients with advanced HCC of Child-Pugh class A liver function who were naive to systemic therapy. METHODS: Everolimus was initiated at 2.5mg once daily and increased per a Bayesian sequential dose-escalation scheme based on the dose-limiting toxicities experienced within the first 28 days of treatment. Adverse events were assessed continuously. Efficacy was evaluated using the best overall response rate per RECIST. RESULTS: Thirty patients were enrolled; 25 were evaluable for MTD determination. One out of 12 patients treated with everolimus 2.5mg once daily and 6 out of 13 patients treated with everolimus 5.0mg once daily experienced a dose-limiting toxicity, most commonly thrombocytopenia (n=5). All patients experienced 1 adverse event, most commonly diarrhea (66.7%), hand-foot skin reaction (66.7%), and thrombocytopenia (50.0%). Best overall response was stable disease (62.5% and 42.9% in the 2.5-mg and 5.0-mg cohorts, respectively). Median time to progression and overall survival in the 2.5-mg cohort were 4.5 months and 7.4 months, respectively, and 1.8 months and 11.7 months, respectively, in the 5.0-mg cohort. CONCLUSIONS: In patients with advanced HCC, the everolimus MTD in combination with standard-dose sorafenib was 2.5mg once daily. The inability to achieve a biologically effective everolimus concentration at the MTD precluded phase II study of this combination.

Effects of hepatic impairment on the pharmacokinetics of everolimus: a single-dose, open-label, parallel-group study.

BACKGROUND: Although the pharmacokinetics of everolimus, an oral mammalian target of rapamycin inhibitor, have been characterized in patients with moderate hepatic impairment, they have not been assessed in those with mild or severe hepatic impairment. OBJECTIVE: The goal of this study was to assess the pharmacokinetics and safety of everolimus in healthy volunteers with normal hepatic function and patients with mild (Child-Pugh class A), moderate (Child-Pugh class B), and severe (Child-Pugh class C) hepatic impairment in otherwise good health to inform dosing in the clinical setting. METHODS: A multicenter, open-label, Phase I study in which all enrollees received a single, 10-mg, oral everolimus dose was conducted. Blood samples for pharmacokinetic assessment were collected at predetermined time points up to 168 hours postdosing. Safety was also assessed. Proposed dose recommendations based on Child-Pugh status at baseline and day 8 were calculated based on AUC0-∞ geometric mean ratios and their associated 90% CIs. Post hoc analysis of the relationship between pharmacokinetic parameters and markers of hepatic function was also performed to identify thresholds for dose adjustment. RESULTS: Thirteen subjects with normal hepatic function and 7 patients with mild, 8 patients with moderate, and 6 patients with severe hepatic impairment were enrolled. Compared with normal subjects, everolimus AUC0-∞ for patients with mild, moderate, and severe hepatic impairment increased by 1.60-, 3.26-, and 3.64-fold, respectively. Based on Child-Pugh classification at day 8, the everolimus doses required to adjust the exposure of patients with mild, moderate, and severe hepatic impairment to that of normal subjects were 6.25, 3.07, and 2.75 mg, respectively. Thresholds for 2-fold everolimus dose reduction were 15.0 µmol/L for bilirubin, 43.1 g/L for albumin, and 1.1 for the international normalized ratio; using these thresholds could lead to underdosing or overdosing in some patients. Most adverse events were of grade 1 severity, ≤1 day in duration, and not everolimus related. CONCLUSIONS: Everolimus exposure after a single 10-mg dose was influenced by the degree of hepatic impairment. Child-Pugh classification was found to be the most conservative means of guiding dose adjustment in patients with hepatic impairment. Based on these data, as well as previously reported data for patients with moderate hepatic impairment, everolimus once-daily dosing should be 7.5 mg and 5 mg in patients with mild and moderate impairment, respectively. Everolimus is not recommended in patients with severe hepatic impairment unless benefits outweigh risks; in that case, 2.5 mg once daily should not be exceeded. ClinicalTrials.gov identifier: NCT00968591.