Full L1-regularized Traction Force Microscopy over whole cells

BMC Bioinformatics

Research Article

Arrate Muñoz-Barrutia¹  Alejandro Suñé-Auñón¹  Alvaro Jorge-Peñas²  Hans Van Oosterwyck³  Rocío Aguilar-Cuenca⁴  Miguel Vicente-Manzanares⁴ 

[1] Bioengineering and Aerospace Engineering Department, Universidad Carlos III de Madrid, Leganés, Spain;Instituto de Investigación Sanitaria Gregorio Marañón, 28911, Madrid, Spain;Biomechanics Section, Department of Mechanical Engineering, KU Leuven, 3001, Leuven, Belgium;Biomechanics Section, Department of Mechanical Engineering, KU Leuven, 3001, Leuven, Belgium;Prometheus, Division of Skeletal Tissue Engineering, KU Leuven, Leuven, Belgium;Instituto de Investigación Sanitaria-Hospital Universitario de la Princesa, Universidad Autónoma de Madrid, School of Medicine, 28006, Madrid, Spain;

关键词: Traction Force Microscopy;    Spatial domain;    Regularization;    Spatial resolution;   

DOI  :  10.1186/s12859-017-1771-0

 received in 2017-02-28, accepted in 2017-07-30,  发布年份 2017

来源: Springer

【 摘 要 】

BackgroundTraction Force Microscopy (TFM) is a widespread technique to estimate the tractions that cells exert on the surrounding substrate. To recover the tractions, it is necessary to solve an inverse problem, which is ill-posed and needs regularization to make the solution stable. The typical regularization scheme is given by the minimization of a cost functional, which is divided in two terms: the error present in the data or data fidelity term; and the regularization or penalty term. The classical approach is to use zero-order Tikhonov or L2-regularization, which uses the L2-norm for both terms in the cost function. Recently, some studies have demonstrated an improved performance using L1-regularization (L1-norm in the penalty term) related to an increase in the spatial resolution and sensitivity of the recovered traction field. In this manuscript, we present a comparison between the previous two regularization schemes (relying in the L2-norm for the data fidelity term) and the full L1-regularization (using the L1-norm for both terms in the cost function) for synthetic and real data.ResultsOur results reveal that L1-regularizations give an improved spatial resolution (more important for full L1-regularization) and a reduction in the background noise with respect to the classical zero-order Tikhonov regularization. In addition, we present an approximation, which makes feasible the recovery of cellular tractions over whole cells on typical full-size microscope images when working in the spatial domain.ConclusionsThe proposed full L1-regularization improves the sensitivity to recover small stress footprints. Moreover, the proposed method has been validated to work on full-field microscopy images of real cells, what certainly demonstrates it is a promising tool for biological applications.

【 授权许可】

CC BY   
© The Author(s). 2017

【 图 表 】

Fig. 2

Fig. 1

Fig. 1

Fig. 3

Fig. 2

Fig. 2

12936_2016_1182_Article_IEq39.gif

12951_2017_255_Article_IEq40.gif

Fig. 1

Fig. 3

Fig. 4

Fig. 2

Fig. 6

Fig. 2

Fig. 1