Nonstationary Mapping of Spatial Uncertainty for Medical Image Classification

Automated classification of medical images is very useful for physicians and surgeons in the diagnoses of complex diseases. Computerized medical pattern recognition tools can capture subtle image properties of various pathological patterns and therefore narrow down the gap of reproducible results for reliable decision making under uncertainty. In this paper, a nonstationary mapping of spatial uncertainty in medical images is introduced for feature extraction, which can be effectively applied for diagnostic pattern classification. Experimental results obtained from using abdominal computed tomography imaging and comparisons with other feature extraction methods demonstrate the usefulness of the proposed mapping model.