Image processing techniques and segmentation evaluation
MetadataShow full item record
This thesis presents contributions in the field of microscopic image analysis, in particularthe automatic segmentation of fluorescent images of cell nuclei and colon crypts. Theevaluation methodology of the segmentation results is detailed and a new evaluation criterionis presented.The proposed discrepancy method is based on the comparison: machine segmentation vs.ground-truth segmentation. This error measure eliminates the inconveniences that appear inthe case of concave objects and allows easy control of the method sensibility regarding theobjects shape similarity according to the field in which it is used.An analysis of the most used image processing methods in microscopic imagesegmentation is presented by considered both the pathological fields: cytology and histology.Segmentation methods are also proposed for both fields: segmentation of the nuclei (used incytometry) and crypts segmentation (used in hystometry).Since the critical problem in microscopic images from tissues with colon carcinoma isthe touching nuclei, three techniques are proposed to find the boundaries oftouching/clustered nuclei. Since all methods need accurate background delineation, twoapproaches are proposed for this matter.The segmentation problem of specific chained configurations is solved using the pointswith high concavity and a set of templates and rules to validate and to pair these points. Theclustered/touching cell nuclei within complex structures are separated using the shape of thesection profile or a cross-correlation with a specific template of the separation areas.Regarding the histological structures, two automatic segmentation techniques robustlyidentify the epithelial layer/crypts. Both proposed methods use hierarchical approaches likemorphological hierarchy or anisotropic diffusion pyramid. A useful study of the samplingstep and a comparison between the hierarchy (without sampling) and the pyramid (withsampling) is presented. The significant implication of these techniques consists of the coarseto-fine approach. First the high level information is preferred against the local one to allow aneasy detection of the positions for the interest objects. Next, a more detailed analysis of thehierarchical representations is performed in order to obtain an accurate segmentation.The evaluation has been done by comparison against ground-truth segmentations or byvisual inspecting by a human expert. The results confirmed that the proposed methods couldefficiently solve the segmentation problems of microscopic images.