Abstract
Early detection of cancer often requires analyzing medical images that have registered only feeble responses from the cancerous tissues. These responses are usually displayed on an intensity modulated monochrome display screen with MxN pixels where 8 to 10 bits of excitations are allowed per pixel. The limitations of the modern digital display systems and the constraints of the human visualization system make the task of detection of those feeble responses a challenging one. To compensate for these difficulties, an engineering approach has been developed to meticulously study the gradual formation of the complete grayscale image from its constituent 8 or 10 bits per pixel. Such analysis sometimes releases signatures for the cancer-infected segments that are not visible on the completely formed image. As a result, while watching the gradual formation of the complete image, a radiologist could hone on to certain suspicious areas on that image. On the other hand, since some cancerous segments are also known to add roughness on the image, the locations of those almost invisible peaks from those rough surfaces could be extracted by a computer program. On superimposing those extracted peak locations over the different layers of the image formation process, more information could be provided to the radiologist. Finally, an Artificial Neural Network (ANN) trained with fuzzy logic is used to predict the degree of belongingness of the suspicious segments of the image toward the cancerous or non-cancerous classes. This interactive and iterative approach stays under the complete command of a physician who cohesively uses the inferences from those three methods to detect cancer in its early stages and save lives.
Biography
Dr. Swapan Chakrabarti has a master’s in physics from the University of Calcutta, India; a master’s in compuational physics and a Ph.D. in electrical engineeing, both from the University of Nabraska – Lincoln, USA. He is now a faculty emeritus of the Department of Electrical Engineering and Computer Science at the University of Kansas (KU) in Lawrence-Kansas, USA. Known for student-centered teaching, Chakrabarti received departmental and university-wide teaching awards during his nearly 30 years of teaching at that department. He also worked as the Associate Director of Education for two years and the Director of Technology Transfer for three years at the National Science Foundation’s Center for Remote Sensing of Ice Sheets (CReSIS) located within the KU campus. His service experience spans over numerous university-wide committes with diversified objectives.
Some of Chakrabarti’s broad research experiences include computation of scattering cross-sections from rough surfaces, designing 16-bit displays for biomedical and scientific data visualization, combining Artificial Neural Networks with fuzzy logic for solving pattern classification problems such as radar target clasification, drug discovery, etc. A fusion of knowledge from some of these branches helps him to develop somewhat uncoventional approaches to assist detecting cancer in its early stages. He also has performed funded research over diverse fields in engineering, and published more than 50 peer reviewed journal and conference papers.
He has secured four U.S. patents, and four more are currently existing in the Patent Pending status. Over the past three years, he has been working on the development of engineering aproaches to assist physicians in detectng cancer in its earliest opossible stages and save lives
Although not all our forums are available to view online, you can find those that are on our “Forum and Seminar Recordings” page
