Testing for Human Perceptual Categories in a Physician-in-the-loop CBIR System for Medical Imagery

We have addressed the following question in this contribution To what extent should the domain ex perts in our case physicians be believed with regard to what they claim to see in images that allows them to recognize di erent types of pathology Until recently our approach was to have a physi cian delineate the pathology bearing regions in the images We then used what could be referred to as a scattershot approach to the characterization of these regions meaning that we d extract a very large num ber of features from these regions Subsequently we d reduce the dimensionality of this feature space by us ing standard search techniques such as the Sequential Forward Selection method This contribution represents an alternative to the scattershot approach to initial feature extraction In this paper we rst describe the perceptual categories that the physicians claim to use for classifying images as belonging to di erent diseases We then describe the speci c low level features that need to be extracted to determine the presence or the absence of the vari ous perceptual categories We subsequently show the discriminatory power of the perceptual categories by presenting retrieval results obtained when a query im age is matched with the database images on the basis of the presence or the absence of the various percep tual categories