Method for recognizing projection views of radiographs
A method for recognizing the projection view of radiographs comprising the steps of correcting the orientation of the input radiograph, locating a region of interesting in the radiograph, recognizing the projection view of the radiograph.
1. A method for recognizing the projection view of radiographs, comprise the steps carried out by a programmable digital computer of:
accessing an input radiograph;
preprocessing said input radiograph to extract the anatomy of the radiograph:
correcting the orientation of the radiograph;
extracting a region of interest from said anatomy of the radiograph;
computing projection view sensitive features from the region of interest: and
recognizing the projection view of the radiograph, based on said projection view sensitive features, using a set of pretrained classifiers.
2. The method of claim 1 , wherein the step of extracting the region of interest is accomplished by the steps of:
detecting a medial axis;
locating a center of the region of interest;
determining a size and shape of the region of interest; and
extracting the region of interest.
3. The method of claim 2 , wherein the step of medial axis is detected using an Euclidean distance map or by a multiscale medial analysis.
4. The method of claim 1 , wherein the step of recognizing the projection view of the radiograph is accomplished by the steps of:
classifying the radiograph using a set of pre-trained classifiers corresponding to all possible projection views; and
determining the projection view of the radiograph as the one with the highest confidence according to the classification results.
5. The method of claim 1 , wherein a set of a pre-trained classifiers is obtained by the steps of:
collecting a pre-determined number of training images with known projection view information;
locating a region of interest for each of the training images;
computing a set of features from the region of interest of each of the training images;
associating a target output specifying the known projection view of each of the training images;
computing a transformed feature set for each of the training images using principal component analysis based on all the training images; and
training a classifier with the transformed feature set and target output.
6. The method of claim 1 , wherein the step of pre-processing the radiograph comprises the steps of:
sub-sampling the radiograph;
segmenting the radiograph into a foreground region, a background region, and an anatomy region;
removing the foreground and background regions from the radiograph to generate an anatomy image; and
normalizing the anatomy image based on characteristics of the anatomy region.
7. The method of claim 1 wherein said step of computing projection view sensitive features from the region of interest comprises:
dividing the region of interest into a set of blocks; and
computing a set of low-level visual features from the blocks.
8. The method of claim 7 wherein said step of computing projection sensitive features includes studying the characteristics of the low-level visual features, and deriving the most discriminate features from the low-level visual features for classification by using normalization, principle component analysis, independent component analysis, or other feature extraction methods.
9. The method of claim 7 , wherein the low-level visual features include gray level mean, variance, edge information, texture and other image features extracted from a block.