Development of Bright Field Microscope Image Resource and Identification of Robust Focus Measure Functions for Automated Capturing of Images

Abstract

An efficient healthcare system always takes advantage of observation and interpretation of medical data that include narrative, textual data, numerical measurement, recorded signals, radiographs and pathological images. Medical imaging is one of the most important tools for disease diagnosis. Digitization and analysis of imaging data have been attracting attention in recent years due to their societal impact in the domains of diagnosis, observation and the training of doctors. From simple chest X-rays (CXR) to pathological microscopic images (sputum smear image) require the highest levels of quality for acquisition, storage and processing. Infectious disease like tuberculosis which is caused by Mycobacterium tuberculosis requires radiological (CXR) and pathological (sputum smear microscopy) tests for the effective diagnosis. The former is not a confirmatory test due to its non-specific and redundant patterns. Therefore, Ziehl-Neelsen stained conventional bright field microscopic (CM) test is the most widely used confirmatory method in low and middle income countries. However, the manual screening of tuberculosis bacilli using sputum smear CM microscopy may misdiagnose 33 to 50% of active cases due to patient load at the hospital. The majority of current issues on tuberculosis diagnosis can be addressed by incorporating automated methods. Autofocusing, auto-stitching, and image segmentation and classification are the three sequential steps in automated microscopy system for the tuberculosis screening. However, lack of unified datasets impedes the development of robust algorithms on these three domains. Keeping in view of these limitations, the proposed thesis work is based on four objectives which are described in the four different chapters (Chapter 2, 3, 4 and 5). In the 1st objective, Ziehl-Neelsen Sputum smear Microscopy image Database (ZNSM-iDB) has been developed to facilitate the development of algorithms and methods related to automated microscopy system, and it is freely available at http://14.139.240.55/zn