Object based image retrieval
A computer-aided image comparison, evaluation and retrieval system compares objects and object clusters, or images. User controlled or automatic filtering to enhance object features may be performed prior to object definition/detection. The query image may be substantially continuously displayed during the image filtering and object definition processes. Scoring to suspected biological, medical, chemical, physical or clinical condition may be performed based on retrieved objects or images and their relative similarities to the unknown.
1. A method of analyzing digital images produced by a data source having an output defined by a collection of digital values forming a single dimensional or multi-dimensional image so as to verify a suspected biological, medical, chemical, physical or clinical condition of a patient comprising:
processing a first image of a portion of a first human body having a suspected biological, medical, chemical, physical or clinical condition with one or more filter functions so as to modify at least some pixel intensity values and so as to define at least one separated portion of said first image, wherein said separated portion comprises a first set of object pixels distinguished from background pixels, and wherein said first set of object pixels is associated with a physical structure in said first human body potentially related to said suspected biological, medical, chemical, physical or clinical condition;
parameterizing said separated portion of said first image to produce a corresponding parameter set associated with said separated portion of said first image, said parameter set including at least one parameter selected from area, perimeter, maximum diameter, minimum diameter, form factor, equivalent circular diameter, aspect ratio, optical density, integrated density, relief, and homogeneity;
processing a second image of a portion of a second human body having a known biological, medical, chemical, physical or clinical condition with substantially the same one or more filter functions so as to define at least one separated portion of said second image, wherein said separated portion comprises a second set of object pixels distinguished from background pixels, and wherein said second set of object pixels is associated with a physical structure in said second human body related to said known biological, medical, chemical, physical or clinical condition;
parameterizing said separated portion of said second image to produce a corresponding parameter set associated with said separated portion of said second image, said parameter set including at least one parameter selected from area, perimeter, maximum diameter, minimum diameter, form factor, equivalent circular diameter, aspect ratio, optical density, integrated density, relief, and homogeneity;
comparing the content of a parameter set associated with at least one separated portion of said first image with the content of a parameter set associated with at least one separated portion of said second image so as to produce a measure of similarity between the structure in the first human body having the suspected biological, medical, chemical, physical or clinical condition and the structure in the second human body having a known biological, medical, chemical, physical or clinical condition; and
alerting a user that said suspected biological, medical, chemical, physical or clinical condition is likely actually present in said first human body.
2. The method of claim 1 , wherein said alerting comprises producing a numerical score indicative of the likelihood that said suspected biological, medical, chemical, physical or clinical condition is actually present in said first human body.
3. The method of claim 2 , wherein said numerical score is computed at least in part by selecting a number of closest relative similarity matches to the unknown object, and determining whether the majority of the closest matches were diagnosed as exhibiting said suspected biological, medical, chemical, physical or clinical condition.
4. The method of claim 2 , wherein said numerical score is computed at least in part by selecting a number of closest relative similarity matches to the unknown object, and averaging a numerical score indicative of said suspected biological, medical, chemical, physical or clinical condition previously assigned to each of said closest relative similarity matches.
5. The method of claim 4 , wherein the average is weighted to matches having higher relative similarity with the unknown object.