25N. Selected Speaker - Laura Nicolás-Saenz скачать в хорошем качестве

25N. Selected Speaker - Laura Nicolás-Saenz

Nonlinear image registration and pixel classification pipeline for the study of tumor heterogeneity maps. November 25th, 2020. by Laura Nicolás-Saenz, Bioengineering and Aerospace Engineering Department, Univ. Carlos III Madrid (Spain). co-authors: Sara Guerrero-Aspizua1,2,3, Javier Pascau1,5, Arrate Muñoz-Barrutia1,5 affiliations: 1- Departamento de Bioingenieria e Ingenieria Aeroespacial, Universidad Carlos III de Madrid, 28911 Leganes, Spain 2- Centre for Biomedical Network Research on Rare Diseases (CIBERER), U714, 28029 Madrid, Spain 3- Hospital Fundación Jiménez Díaz e Instituto de Investigación FJD, 28040 Madrid, Spain 4- Epithelial Biomedicine Division, CIEMAT, 28040 Madrid, Spain Instituto de Investigación Sanitaria Gregorio Marañon, 28007 Madrid, Spain. Abstract: We present a novel method to assess the variations in protein expression and spatial heterogeneity of tumor biopsies with application in computational pathology. This was done using different antigen stains for each tissue section and proceeding with a complex image registration followed by a final step of color deconvolution to detect the exact location of the proteins of interest. For proper assessment, the registration needs to be highly accurate for the careful study of the antigen patterns. However, accurate registration of histopathological images comes with three main problems: the high amount of artifacts due to the complex biopsy preparation, the size of the images, and the complexity of the local morphology. Our method manages to achieve an accurate registration of the tissue cuts and segmentation of the positive antigen areas. Method: We developed a transformation T that consists of a robust pre-alignment and a global and a local transformation. The combined transformation T can be expressed as T=TAlign+TGlobal+TLocal. The alignment and global motion are used to describe the overall motion of the biopsy and appropriately register the biopsy global shape. The local motion involves the alignment of the internal structure of the biopsy. The color deconvolution algorithm is based on the bisection of the red-blue color joint histogram of the images through their diagonal. Results: We used the relative Target Registration Error (rTRE) to evaluate the registration performance and found that the median error was reduced in each step (Final Median rTRE of 0.007). We compared our algorithm to several registration methods and were able to prove that our algorithm can compete with state-of-the-art methods, outperforming most of them. The segmentation performance was evaluated comparing the masks generated by our algorithm with a ground truth generated with WEKA. The metrics used were Dice similarity coefficient (0.97 ± 0.0352) and Hausdorff distance (10.96?m ± 3.97?m). The final Heterogeneity Map is created using the registered tissue cuts. Each transformed section is subjected to our color deconvolution algorithm, which gives as output a binary mask of the areas positive for the antigen. These masks are stacked over the fixed image of the patient. Conclusion The advantages of our proposed pipeline are its robustness and, most importantly, its completely automatic and non-supervised character. Compared to other segmentation and registration algorithms, our proposal yields similar registration results and improved segmentation results without the need of manual annotations nor training.