Stanciu Stefan G, Xu Shuoyu, Peng Qiwen, Yan Jie, Stanciu George A, Welsch Roy E, So Peter T C, Csucs Gabor, Yu Hanry
1] Center for Microscopy-Microanalysis and Information Processing, University Politehnica of Bucharest, Romania [2] Light Microscopy and Screening Center, ETH Zurich, Switzerland.
1] Computation and System Biology Program, Singapore MIT Alliance, Singapore, Singapore [2] Biosystems and Micromechanics IRG, Singapore MIT Alliance for Research and Technology, Singapore, Singapore [3] Institute of Bioengineering and Nanotechnology, Singapore, Singapore.
Sci Rep. 2014 Apr 10;4:4636. doi: 10.1038/srep04636.
The accurate staging of liver fibrosis is of paramount importance to determine the state of disease progression, therapy responses, and to optimize disease treatment strategies. Non-linear optical microscopy techniques such as two-photon excitation fluorescence (TPEF) and second harmonic generation (SHG) can image the endogenous signals of tissue structures and can be used for fibrosis assessment on non-stained tissue samples. While image analysis of collagen in SHG images was consistently addressed until now, cellular and tissue information included in TPEF images, such as inflammatory and hepatic cell damage, equally important as collagen deposition imaged by SHG, remain poorly exploited to date. We address this situation by experimenting liver fibrosis quantification and scoring using a combined approach based on TPEF liver surface imaging on a Thioacetamide-induced rat model and a gradient based Bag-of-Features (BoF) image classification strategy. We report the assessed performance results and discuss the influence of specific BoF parameters to the performance of the fibrosis scoring framework.
肝纤维化的准确分期对于确定疾病进展状态、治疗反应以及优化疾病治疗策略至关重要。非线性光学显微镜技术，如双光子激发荧光（TPEF）和二次谐波产生（SHG），可以对组织结构的内源性信号进行成像，并可用于对未染色组织样本进行纤维化评估。虽然迄今为止一直致力于对SHG图像中的胶原蛋白进行图像分析，但TPEF图像中包含的细胞和组织信息，如炎症和肝细胞损伤，与SHG成像的胶原蛋白沉积同样重要，迄今为止仍未得到充分利用。我们通过在硫代乙酰胺诱导的大鼠模型上基于TPEF肝脏表面成像的联合方法和基于梯度的特征袋（BoF）图像分类策略进行肝纤维化定量和评分实验来解决这一情况。我们报告评估的性能结果，并讨论特定BoF参数对纤维化评分框架性能的影响。
