Biennial Report 2005-2007 - Saha Institute of Nuclear Physics

96 Biennial Report 2005-07progress.Abhijit SanyalMED3.2.1.11 Feasibility of Temperature imagingThe property of ultrasound attenuation of agar hydrogel that changes with temperature linearlyfrom around 20 ◦ C to 42 ◦ C is explained on the basis of visco-elastic fluid model. It is also shown thatthe excess water content does not play any significant role in contributing ultrasound attenuationof the gel. From DSC measurement we have shown that phase transition takes place at around20 ◦ C and 42 ◦ C which is in good agreement with those as observed in ultrasound attenuation of thegel too. It is also conjectured that this linearity region could be used to estimate temperature ofhydrogel like polymers in noninvasive manner.To estimate the spatial temperature profile within liquid phantoms under insonification an instrumenthas been designed and fabricated in our workshop.Madhusudan Roy, Supratic ChakrabortyMED3.2.1.12 Development of ultrasonic probeSimulation of electrical properties of silicon oxide membrane: Silicon-based non-contact air-coupledultrasonic transducers using Micro-Electro-Mechanical-System (MEMS) technology for ultrasoundscan and non-destructive evaluation from the perspective of medical imaging has been emerging asa viable alternative to existing 1-D piezoelectric transducer array due to its difficulties for real-timeimaging of dynamic structure such as heart. We have procured PZFEX software and installed it.We have started working with it. Keeping in view of the next five year plan period project we havecompleted preliminary works in connection with clean room, customizations of instruments etc.Supratic Chakraborty, Madhusudan RoyMED3.2.1.13 Texture analysis of ultrasound images of human liverTexture analysis based on spatial gray-level dependence (SGLD) matrix computation is carriedout over the input ultrasound images of livers to extract features, namely, maximum probability,uniformity, entropy, element difference moment of order 2, inverse element difference moment oforder 2, homogeneity and correlation. In this work seventy-six ultrasound scan images of humannormal livers and twenty-four ultrasound images of fatty livers as identified by a radiologist onthe basis of echogenecity, have been collected from hospital. From each ultrasound image of theliver five sub-images, termed as region of interest (ROI), have been suitably cropped and takenas an input image for analysis. An unsupervised neural network learning technique, namely, SelfOrganising Map (SOM) has been employed to generate profile plots. Using Student’s t like statisticfor each feature as a measure of distinction between normal and fatty livers, two most appropriatefeatures, namely, maximum probability (Maxp) and uniformity (Uni) are selected from this profileplots. These two features are found to form clusters with little overlap for normal and fatty livers.