Inference systems for automated image analysis

The radiodiagnostic process is a complicated activity involving the integration of knowledge from low-level image features into more abstract, higher-order entities. This involves spatial and density information at a lower level and area and features at a higher level. The spatial density distribution in a radiographic image is not uniquely related to the three-dimensional structure of the object. Therefore, more information other than just first-order density characteristics of radiographs must be utilized to improve automated interpretation of the image. Prior knowledge of the size, shape and location of anatomical structures and pathognomic features is very useful for improving the process of computer-aided image analysis. Inference systems as used in expert systems can be applied to facilitate the integration of information obtained from the patient and the radiograph in the diagnostic process.