| 主权项 |
1. A method for the computerised classification of a digital medical diagnostic image of a lung or a part thereof, comprising applying to said image a trained statistical classifier which has been trained by supervised learning on a training set of methodologically similar lung images each of which images has been labelled by metadata being data which is not obtained from the image or images and which is indicative of the likelihood of the respective image relating to a lung characterised by a lung disease, or the degree of such disease, or propensity to develop such disease, wherein in said training of the classifier, for each image in the training set a number of regions of interest (ROIs) were defined, and textural information relating to the intensities of locations within each ROI was obtained, and combinations of features of said textural information were found which suitably classify said training set images according to said metadata, and wherein, in applying said trained statistical classifier to said image, in a computer a number of regions of interest (ROIs) are defined in said image, and textural information relating to the intensities of locations within each ROI of the kind used in training the classifier is obtained, and features of said textural information for the locations within the ROIs of the image are combined as learnt in the training of the classifier to calculate probabilities of said locations within the ROIs belonging to a healthy lung as against a lung characterised by a said disease or degree of disease or propensity to develop such disease and said probabilities are combined to obtain for the image a quantitative fused probability of the image belonging to a healthy lung as against a lung characterised by said disease, or a degree thereof, or a propensity to develop said disease. |
