Date of Submission


Date of Award


Institute Name (Publisher)

Indian Statistical Institute

Document Type

Doctoral Thesis

Degree Name

Doctor of Philosophy

Subject Name

Computer Science


Computer Vision and Pattern Recognition Unit (CVPR-Kolkata)


Chaudhuri, Bidyut Baran (CVPR-Kolkata; ISI)

Abstract (Summary of the Work)

Automation of image analysis in the bio-medical ficld is one of the important achievements of applied image processing research. The rapid development in the electronic instrumentation during 1960s and 70s made it possible to automate the routine process of diagnosis and prognosis of many discases. Development of high resolution imaging instruments such as X-ray CT, MRI, etc., for macro imaging and electron microscope, confocal microscope, etc., for micro imaging has given a tremendous boost to the advancement of medical field. Advancement in the field of computing has made it possible to reconstruct the pictures of internal organs of the body in a non-invasive method.Based on the imaging instrument, we can broadly classify the medical images into two types. They are macro-images obtained by macro-imagìng instruments such as MRI, CT- scan, Ultra-sound, etc., and micro-images obtained by micro-imaging instruments such as light microscope, electron microscope, confocal microscope, ctc.. This thesis mainly concerns with the analysis of micro-imaging data sets obtained using confocal microscope. More specifically, we consider 3-D histo-pathological images where the aim is to automatically segment the cells and measure the quantitative features of the tissue and the cells of a histo- pathological specimens.The histo-pathological images are relatively complex in the sense that the cells are arranged in different patterns often touching or overlapping on each other. Moreover, during different pathological disorders, the change in the features of tissue and cells belonging to different organs is different. Even the cells and tissue of the same organ may show different characteristics during different sub-classes of the same disease, Such problems, along with the inconsistency in defining the features by different pathologists as well as lack of standard procedures to define the pathological features make automatic processing of these images more difficult. As a result very few automated systems are found to be successful.The necessity of automation of the analysis of histo-pathological images stems from the fact that the quantitative study of the tissue specimen by visual approach is very difficult. The visual inspection of the tissue specimen under the microscope depends upon the understanding of the physiological processes and the ability to diagnose by comparing each sample to cases that the expert has seen before. This qualitative or subjective evaluation is appropriate to identify different diseases. The same can not be said while differentiating the different sub-classes of the same disease (Firestone et al., 1996). Visual inspection is less effective in quantification of tissue characteristics. The number of specimens to be inspected, the time factor, consistency and the reproducibility make the automatic quantitative study of the biological specimens, a significant and important part of tissue specimen analysis. The automation of the quantitative study of histo-pathological images exists in several levels. A completely automatic system is the one where the human interaction may be 1 needed only for preparing and loading the specimens. In the semi-automatic systems, the human interaction is necessary to feed a priori information for processing the image data. There is a task-specific automation in which only a particular task or the region in the image is subject to automatic image analysis.


ProQuest Collection ID:

Control Number


Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.


Included in

Mathematics Commons