A maximum likelihood method for estimating the parameters of a search model

During case interpretation the radiologist searches a patient's image for possible lesions and marks and rates any perceived suspicious regions. A model is described for the mark-rating pairs generated in such interpretations, which is closely paralleled by the free-response receiver operating characteristic (FROC) paradigm. The search model has parameters quantifying perceived lesion signal-to-noise-ratio, the observer's ability to avoid making non-lesion localizations, and the observer's ability to find lesions. A method for estimating the model parameters from the observer's FROC or receiver operating characteristic (ROC) data and a preliminary validation of the procedure are described. Since search is a fundamental aspect of many activities in both human and machine vision, the ability to model and estimate the parameters of the search model from observer data may have considerable significance.