Quantitative analysis of cell morphology based on the contourlet transform

Yali Huang,Xuefang Hu,Lei Hao,Yuehua Gao,Zhiwen Liu,Peiguang Wang

DOI: https://doi.org/10.1049/iet-ipr.2019.0909

IF: 2.3

2020-01-01

IET Image Processing

Abstract:Cellular morphology analysis has been widely used to detect abnormalities in biological processes. Clinicians have observed that lymphocytes become highly deformable under special conditions, particularly when graft rejection occurs. The characterisation of lymphocyte boundary deformation provides important quantitative parameters to assist clinical rejection diagnosis. To evaluate the dynamic features of the boundaries of target lymphocyte when a graft rejection occurs, a contourlet transform-based method is proposed to extract the characteristics of cell boundary variation. First, the lymphocyte is segmented and tracked to obtain their edge-to-centroid distance signals. Subsequently, a contourlet transform is performed on these signals, during which the edge-to-centroid distance signals of the lymphocyte is decomposed at multiple scales using the Laplacian pyramid; a multi-directional decomposition is then performed using a direction filter to merge the singularities distributed along the same direction and obtain the contourlet transform coefficients. Finally, statistical parameters of the cell dynamic boundaries are calculated, then fed into the support vector machine for classification of the cell deformation. Our findings demonstrate that contourlet transform has better performance in representing image features such as cell boundaries than wavelet transform for its prosperities of multi-scale and multi-directional decomposition for cell images.