Professor, Department of Electronics & communication Engineering, Sri Sairam Engineering College, India

Dr. E. Priya completed her Ph.D at MIT Campus, Anna University in the field of “Automated analysis using image processing and artificial intelligence for the diagnosis of tuberculosis images”. At present she is a Professor at the Department of Electronics and Communication Engineering, Sri Sairam Engineering College, Affiliated to Anna University, Chennai.

She has 17 years of teaching experience, 4 years of research experience and 3 years of industrial experience. She is a recipient of DST-PURSE fellow and a project participant of India-South African collaborative project titled “Development of computing tools for decision support in health assessment in rural areas”. Her areas of interest include bio-medical imaging, image processing, signal processing, application of artificial intelligence and machine learning techniques. She has published papers in refereed International Journals, Conferences and book chapters in the area of medical imaging and infectious diseases.

Speech Title: An Automated Approach for the Identification and Classification of Sputum Smear Tuberculosis Images

Abstract: Tuberculosis (TB) is a contagious airborne infectious disease for which early diagnosis is essential to control. The conventional method employed for TB identification is the microscopy-based TB screening. Manual screening using microscope is tedious and requires highly trained experts. Besides huge variability in sensitivity, manual screening for the identification of disease causing agent is a labour intensive task. Further, it is time consuming and depends on patient’s level of infection and requires large number of images to be analysed in one slide. Hence there is a need to automate the diagnostic process to improve the sensitivity and accuracy of the test.

The non-uniform illumination in microscopic digital TB images due to light source optics and camera noise degrades the visual perception of these images. The most appropriate illumination correction method is evaluated by qualitative and quantitative measures. Qualitative analysis includes intensity profiles and histogram-based statistical features serve as a quantitative measure to validate the illumination correction methods. The pre-processed sputum smear images are subjected to segmentation. More than 80% match is observed between the segmentation method and ground truth obtained from human annotated images for the overlap measures. Later the identified TB objects are segregated into bacilli and non-bacilli by extracting Fourier descriptors, Hu and Zernike moments. The most significant features are derived and are subjected to classification using extreme learning machines. Classification of sputum smear images into TB positive and negative is evaluated based on performance measures. The developed automated technique is useful for mass screening of pulmonary tuberculosis because of the higher accuracy.

Keywords: Tuberculosis, sputum smear, non-uniform illumination, histogram-based analysis, Fourier descriptors, moment features, extreme learning mchines