No abstract

No abstract

SummaryMany kinds of neuroscience data are being acquired regarding the dynamic behaviour and phenotypic diversity of nerve cells. But as the size, complexity and numbers of 3D neuroanatomical datasets grow ever larger, the need for automated detection and analysis of individual neurons takes on greater importance. We describe here a method that detects and identifies neurons within confocal image stacks acquired from the zebrafish brainstem. The first step is to create a template that incorporates the location of all known neurons within a population -in this case the population of reticulospinal cells. Once created, the template is used in conjunction with a sequence of algorithms to determine the 3D location and identity of all fluorescent neurons in each confocal dataset. After an image registration step, neurons are segmented within the confocal image stack and subsequently localized to specific locations within the brainstem template -in many instances identifying neurons as specific, individual reticulospinal cells. This image-processing sequence is fully automated except for the initial selection of three registration points on a maximum projection image. In analysing confocal image stacks that ranged considerably in image quality, we found that this method correctly identified on average ∼80% of the neurons (if we assume that manual detection by experts constitutes 'ground truth'). Because this identification can be generated approximately 100 times faster than manual identification, it offers a considerable time savings for the investigation of zebrafish reticulospinal neurons. In addition to its cell identification function, this protocol might also be integrated with stereological techniques to enhance quantification of neurons in larger databases. Our focus has been on zebrafish brainstem systems, but the Correspondence to: Donald M. O'Malley.

When performing handwriting recognition on natural language text, the use of a word-level language model (LM) is known to significantly improve recognition accuracy. The most common type of language model, the n-gram model, decomposes sentences into short, overlapping chunks.In this paper, we propose a new type of language model which we use in addition to the standard n-gram LM. Our new model uses the likelihood score from a statistical machine translation system as a reranking feature. In general terms, we automatically translate each OCR hypothesis into another language, and then create a feature score based on how "difficult" it was to perform the translation. Intuitively, the difficulty of translation correlates with how well-formed the input sentence is. In an Arabic handwriting recognition task, we were able to obtain an 0.4% absolute improvement to word error rate (WER) on top of a powerful 5-gram LM.

Offline handwriting recognition (OHR) is an extremely challenging task because of many factors including variations in writing style, writing device and material, and noise in the scanning and collection process. Due to the diverse nature of the above challenges, it is highly unlikely that a single recognition technique can address all the characteristics of real-world handwritten documents. Therefore, one must consider designing different systems, each addressing specific challenges in the handwritten corpus, and then combining the hypotheses from these diverse systems. To that end, we present an innovative approach for combining hypotheses from multiple handwriting recognition systems. Our approach is based on generating a consensus network using hypotheses from a diverse set of handwriting recognition systems. Next, we decode the consensus network for producing the best possible hypothesis given an error criterion. Experimental results on an Arabic OHR task show that our combination algorithm outperforms the NIST ROVER technique and results in a 7% relative reduction in the word error rate over the single best OHR system.