Abstract: In this paper we review a PDM method for defining simple and flexible model of shape variations of the lung and heart anatomy in chest radiographs. Boundaries of these organs are represented by a set of landmark points and modes of variations are determined from ways which the landmark points tend to deform from a mean shape. The method has linear computational complexity. This model can be used as initial hypothesis in automatic image search, segmentation, and cancer detection algorithms. This method is shown to be an efficient method for refining initial hypothesis and has much less dimensions as compared to the original shape data.