2 results
Comparison of rule-based and neural network models for negation detection in radiology reports
- D. Sykes, A. Grivas, C. Grover, R. Tobin, C. Sudlow, W. Whiteley, A. Mcintosh, H. Whalley, B. Alex
-
- Journal:
- Natural Language Engineering / Volume 27 / Issue 2 / March 2021
- Published online by Cambridge University Press:
- 18 November 2020, pp. 203-224
-
- Article
- Export citation
-
Using natural language processing, it is possible to extract structured information from raw text in the electronic health record (EHR) at reasonably high accuracy. However, the accurate distinction between negated and non-negated mentions of clinical terms remains a challenge. EHR text includes cases where diseases are stated not to be present or only hypothesised, meaning a disease can be mentioned in a report when it is not being reported as present. This makes tasks such as document classification and summarisation more difficult. We have developed the rule-based EdIE-R-Neg, part of an existing text mining pipeline called EdIE-R (Edinburgh Information Extraction for Radiology reports), developed to process brain imaging reports, (https://www.ltg.ed.ac.uk/software/edie-r/) and two machine learning approaches; one using a bidirectional long short-term memory network and another using a feedforward neural network. These were developed on data from the Edinburgh Stroke Study (ESS) and tested on data from routine reports from NHS Tayside (Tayside). Both datasets consist of written reports from medical scans. These models are compared with two existing rule-based models: pyConText (Harkema et al. 2009. Journal of Biomedical Informatics42(5), 839–851), a python implementation of a generalisation of NegEx, and NegBio (Peng et al. 2017. NegBio: A high-performance tool for negation and uncertainty detection in radiology reports. arXiv e-prints, p. arXiv:1712.05898), which identifies negation scopes through patterns applied to a syntactic representation of the sentence. On both the test set of the dataset from which our models were developed, as well as the largely similar Tayside test set, the neural network models and our custom-built rule-based system outperformed the existing methods. EdIE-R-Neg scored highest on F1 score, particularly on the test set of the Tayside dataset, from which no development data were used in these experiments, showing the power of custom-built rule-based systems for negation detection on datasets of this size. The performance gap of the machine learning models to EdIE-R-Neg on the Tayside test set was reduced through adding development Tayside data into the ESS training set, demonstrating the adaptability of the neural network models.
Flow and pressure drop in systems of repeatedly branching tubes
- T. J. Pedley, R. C. Schroter, M. F. Sudlow
-
- Journal:
- Journal of Fluid Mechanics / Volume 46 / Issue 2 / 29 March 1971
- Published online by Cambridge University Press:
- 29 March 2006, pp. 365-383
-
- Article
- Export citation
-
The airways of the lung form a rapidly diverging system of branched tubes, and any discussion of their mechanics requires an understanding of the effects of the bifurcations on the flow downstream of them. Experiments have been carried out in models containing up to two generations of symmetrical junctions with fixed branching angle and diameter ratio, typical of the human lung. Flow visualization studies and velocity measurements in the daughter tubes of the first junction verified that secondary motions are set up, with peak axial velocities just outside the boundary layer on the inner wall of the junction, and that they decay slowly downstream. Axial velocity profiles were measured downstream of all junctions at a range of Reynolds numbers for which the flow was laminar.
In each case these velocity profiles were used to estimate the viscous dissipation in the daughter tubes, so that the mean pressure drop associated with each junction and its daughter tubes could be inferred. The dependence of the dissipation on the dimensional variables is expected to be the same as in the early part of a simple entrance region, because most of the dissipation will occur in the boundary layers. This is supported by the experimental results, and the ratio Z of the dissipation in a tube downstream of a bifurcation to the dissipation which would exist in the same tube if Poiseuille flow were present is given by \[ Z = (C/4\surd{2})(Re\,d/L)^{\frac{1}{2}}, \] where L and d are the length and diameter of the tube, Re is the Reynolds number in it, and the constant C (equal to one for simple entry flow) is equal to 1·85 (the average value from our experiments). In general, C is expected to depend on the branching angles and diameter ratios of the junctions used. No experiments were performed in which the flow was turbulent, but it is argued that turbulence will not greatly affect the above results at Reynolds numbers less than and of the order of 10000. Many more experiments are required to consolidate this approach, but predictions based upon it agree well with the limited number of physiological experiments available.