Conference Paper

Learning from uncertain image data using granular fuzzy sets and bio-mimetic applicability functions.

Conference: Proceedings of the Joint 4th Conference of the European Society for Fuzzy Logic and Technology and the 11th Rencontres Francophones sur la Logique Floue et ses Applications, Barcelona, Spain, September 7-9, 2005
Source: DBLP

ABSTRACT In this paper we present a new method for ma- chine learning from images using uncertain informa- tion granules which has been inspired by bio-mimetic study of human perceptual processing. We present a method for generating labelled image data using the rapid application of rough labelled regions to the im- age under study. Over each region is defined an ap- plicability function which acts as a centre of focus for the uncertain information contained within the re- gion. We present a number of alternative applicabil- ity functions inspired by the human visual system and particularly by the centre-weighted effect of the fovea within the retina. We also show how this uncertain data can be used to directly train a granular fuzzy ma- chine learner.

  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: We present Cartesian granule feature, a new multidimensional feature formed over the cross product of fuzzy partition labels. Traditional fuzzy modelling approaches, mainly use flat features (one dimensional features) and, consequently suffer from decomposition error when modelling systems where there are dependencies between the input variables. Cartesian granule features help reduce (if not eliminate) the error due to the decompositional usage of features. In the approach taken here, we label the (fuzzy) subsets which partition the various universes and incorporate these labels in the form of Cartesian granules into our modelling process. Fuzzy sets defined in terms of these Cartesian granules, are extracted automatically from statistical data using the theory of mass assignments, and are incorporated into fuzzy rules. Consequently we not only compute with words, we also model with words. Due to the interpolative nature of fuzzy sets, this approach can be used to model both classification and prediction problems. Overall Cartesian granule features incorporated into fuzzy rules yield glass-box models and when demonstrated on the ellipse classification problem yields a classification accuracy of 98%, outperforming standard modelling approaches such as neural networks and the data browser
    Fuzzy Systems, 1997., Proceedings of the Sixth IEEE International Conference on; 08/1997
  • [Show abstract] [Hide abstract]
    ABSTRACT: We present a new method for tagging image regions using uncertain information granules. This tagging forms an efficient route for the elicitation of knowledge from domain experts with respect to images. We then use this uncertain granular information to train a fuzzy machine learner and then to classify unseen images. This method is particularly suited to applications where an expert input into the classification process is essential but where the expert's time is in extremely short supply. Results are presented within the example domain of detecting the lung disease from computed tomography scans.
    Fuzzy Systems, 2004. Proceedings. 2004 IEEE International Conference on; 08/2004
  • [Show abstract] [Hide abstract]
    ABSTRACT: To use the eye position recordings of observers as they read chest images to advance understanding of the mechanism of "satisfaction of search," a phenomenon in which the detection of one abnormality interferes with the detection of other abnormalities. Eight radiologists examined chest images that contained simulated pulmonary nodules and native abnormalities. Accuracy in detecting nodules and native abnormalities and eye position, gaze duration, and total search time were recorded. Nodule detectability was lower on native abnormality-containing images than it was on normal images (P < .005). Native-abnormality detectability was not negatively affected by the nodules. Most missed nodules were fixated; only those on images without native abnormalities, however, received prolonged visual attention. The satisfaction of search phenomenon is an important source of error in the detection of subtle abnormalities but not of obvious abnormalities. Obvious abnormalities capture visual attention and decrease vigilance for more subtle abnormalities.
    Radiology 04/1995; 194(3):895-902. · 6.34 Impact Factor

Full-text (2 Sources)

Available from
May 19, 2014