<p>With medical imaging, clinicians are given new opportunities in inspection of anatomical structures, surgical planning and diagnosing. Computer vision is often used with the aim of automating these processes. Ultrasound imaging is one of the most popular medical imaging modalities. The equipment is portable and relatively inexpensive, the procedure is non-invasive and there are few known side effects. But the acquisition of ultrasound images, for instance of the heart, is not a trivial job for the inexperienced. Five classes of standard images, or standard views, have been developed to ensure acceptable quality of ultrasound heart images. Automatic recognition of these standard views, or classification, would be a good starting point for an Ultrasound for dummies project. Recently, a new class of object recognition methods has emerged. These methods are based on matching of local features. Image content is transformed into local feature coordinates, which are ideally invariant to translation, rotation, scaling and other image parameters. In [21], David Lowe proposes the Scale Invariant Feature Transform (SIFT), which is a method for extracting distinctive invariant features from an image. He also suggests a method for using these features to recognize different images of the same object. In this thesis I suggest using the SIFT features to classify heart view images. The invariance requirements to a standard heart view recognition system are special. Therefore, in addition to using Lowes algorithm for feature extraction, a new matching algorithm specialized at the heart view classification task is proposed.</p>
| Identifer | oai:union.ndltd.org:UPSALLA/oai:DiVA.org:ntnu-9260 |
| Date | January 2005 |
| Creators | Torland, Anne Vold |
| Publisher | Norwegian University of Science and Technology, Department of Computer and Information Science, Institutt for datateknikk og informasjonsvitenskap |
| Source Sets | DiVA Archive at Upsalla University |
| Language | English |
| Detected Language | English |
| Type | Student thesis, text |
Page generated in 0.0016 seconds