| 摘要 |
Data from a plurality of sensors representing a patient's condition, including the measurement signals and also secondary parameters derived from the measurement signals, are displayed in a simple way by mapping them from the multidimensional measurement space to a two-dimensional visualisation space. This can be achieved using a mapping which preserves the topography of the data points, for instance by ensuring that the inter-point distances in the visualisation space match as closely as possible the corresponding inter-point distances in the measurement space. Such a mapping, for instance Sammon's mapping is achieved by a suitably trained artificial neural network. The parameters are normalised before the mapping process and the normalisation and mapping are such that mapped points from a patient whose condition is normal appear in the centre of the visualisation space, whereas points from a patient whose condition is abnormal appear at the edge of the visualisation space. The artificial neural network may be trained using data points from a single patient or from a group of patients, and data may be thinned out using a pre-clustering algorithm. <IMAGE> |
