Electrical Impedance Tomography (EIT) is a non-invasive imaging technique, which has the potential to expedite the differentiation of ischaemic or haemorrhagic stroke, decreasing the time to treatment. Whilst demonstrated in simulation, there are currently no suitable imaging or classification methods which can be successfully applied to human stroke data. Development of these complex methods is hindered by a lack of quality Multi-Frequency EIT (MFEIT) data. To address this, MFEIT data were collected from 23 stroke patients, and 10 healthy volunteers, as part of a clinical trial in collaboration with the Hyper Acute Stroke Unit (HASU) at University College London Hospital (UCLH). Data were collected at 17 frequencies between 5 Hz and 2 kHz, with 31 current injections, yielding 930 measurements at each frequency. This dataset is the most comprehensive of its kind and enables combined analysis of MFEIT, Electroencephalography (EEG) and Computed Tomography (CT) or Magnetic Resonance Imaging (MRI) data in stroke patients, which can form the basis of future research into stroke classification.

A highly versatile Electrical Impedance Tomography (EIT) system, nicknamed the ScouseTom, has been developed. The system allows control over current amplitude, frequency, number of electrodes, injection protocol and data processing. Current is injected using a Keithley 6221 current source, and voltages are recorded with a 24-bit EEG system with minimum bandwidth of 3.2 kHz. Custom PCBs interface with a PC to control the measurement process, electrode addressing and triggering of external stimuli. The performance of the system was characterised using resistor phantoms to represent human scalp recordings, with an SNR of 77.5 dB, stable across a four hour recording and 20 Hz to 20 kHz. In studies of both haeomorrhage using scalp electrodes, and evoked activity using epicortical electrode mats in rats, it was possible to reconstruct images matching established literature at known areas of onset. Data collected using scalp electrode in humans matched known tissue impedance spectra and was stable over frequency. The experimental procedure is software controlled and is readily adaptable to new paradigms. Where possible, commercial or open-source components were used, to minimise the complexity in reproduction. The hardware designs and software for the system have been released under an open source licence, encouraging contributions and allowing for rapid replication.

No abstract

We describe a practical method to find near-optimal solutions for the area-optimal simple polygonization problem: Given a set of points S in the plane, the objective is to find a simple polygon of minimum or maximum area defined by S . Our approach is based on the celebrated metaheuristic Simulated Annealing. The method consists of a modular pipeline of steps, where each step can be implemented in various ways and with several parameters controlling it. We have implemented several different algorithms and created an application that computes a polygon with minimal (or maximal) area. We experimented with the various algorithmic options and with the controlling parameters of each algorithm to tune up the pipeline. Then, we executed the application on each of the benchmark instances, exploiting a grid of servers, to obtain near optimal results.

We introduce bindings that enable the convenient, efficient, and reliable use of software modules of Cgal (Computational Geometry Algorithm Library), which are written in C++, from within code written in Python. There are different tools that facilitate the creation of such bindings. We present a short study that compares three main tools, which leads to the tool of choice. The implementation of algorithms and data structures in computational geometry presents tremendous difficulties, such as obtaining robust software despite the use of (inexact) floating point arithmetic, found in standard hardware, and meticulous handling of all degenerate cases, which typically are in abundance. The code of Cgal extensively uses function and class templates in order to handle these difficulties, which implies that the programmer has to make many choices that are resolved during compile time (of the C++ modules). While bindings take effect at run time (of the Python code), the type of the C++ objects that are bound must be known when the bindings are generated, that is, when they are compiled. The types of the bound objects are instances (instantiated types) of C++ function and class templates. The number of object types that can potentially be bound, in implementation of generic computational-geometry algorithms, is enormous; thus, the generation of the bindings for all these types in advance is practically impossible. For example, the programmer needs to choose among a dozen types of curves (e.g., line segments, circular arcs, geodesic arcs on a sphere, or polycurves of any curve type) to yield a desired arrangement type; often there are several choices to make, resulting in a prohibitively large number of combinations. We present a system that rapidly generates bindings for desired object types according to user prescriptions, which enables the convenient use of any subset of bound object types concurrently. After many years, in which the usage of these packages was restricted to C++ experts, the introduction of the bindings made them easily accessible to newcomers and practitioners in non-computing fields, as we report in the paper.