Extracting 3D Structure of Blood Cells Using Shape From Focus Strategies

Automatic cell counting, detection, classification and segmentation performed on images containing blood cells are among the areas where artificial intelligence and image processing techniques are most used. The images on which these operations are performed are generally 2-dimensional (2D). However, due to the loss of size and depth of focus, when imaging the three-dimensional (3D) structures of microscopic samples in the 2D plane, images cannot fully represent the true anatomical structure of the cells. This situation negatively affects the success of processes such as automatic counting, detection, classification and segmentation of the cell. To address this deficiency, researchers propose to use not only 2D images but also 3D structural data, which provides a more detailed and accurate representation of blood cells. In this context, the subject of this study is based on two main functions. The first of these is the preparation of a comprehensive data set consisting of microscope image sequences focused on different regions to extract the 3D structure of blood cells. The second is the development of shape from focus (SFF) strategies for extracting the 3D structure of blood cells in microscopic systems. In the study, reference-free performance analysis procedures are used to determine which focusing function is most suitable for extracting the 3D structure of blood cells. The performance analysis procedures used in the study are Surface Smoothness (SS), Surface Normal Variability (SNV) and Internal Consistency (CA - Cronbach's Alpha). Both the results of the performance analysis procedures and the visual findings obtained in the study show that the Variance focus function is most suitable for extracting the 3D structure.