Staining Pattern Classification in Antinuclear Autoantibodies Testing

In Indirect Immunofluorescence (IIF) the use of Computer-Aided Diagnosis (CAD) tools can support physicians' estimation of both fluorescence intensity and staining pattern. This paper reports our experiences in the staining pattern recognition of IIF wells. Since several cells constitute each well, we have developed a Multiple Expert System (MES) based on the one-per-class approach devised to classify the pattern of individual cells. As a novelly, we introduce an aggregation rule based on the estimation of the reliability of each composing experts. Then, the whole well staining pattern is computed using the reliability of its cells classification. The approach has been successfully tested on an annotated set of IIF images.