Research Interests and Experiences
I have been working in the field of imaging and image processing since final year of my B.Sc program. I am mostly interested in combining the physics of the imaged phenomenon with mathematical image processing techniques to better understand and analyze the image.
My current research field is factor analysis of dynamic contrast enhanced images. Factor analysis tries to extract functional information about the images by breaking down the image into a few segments (Factor images) and extracting their corresponding behavior curves (Factor curves). I have developed a factor analysis algorithm that incorporates the a priori information about the physics of contrast enhancement (Pharmacokinetic modeling of the organ of interest) with a purely data driven factor analysis technique (Independent Component Analysis) to generate physiologically meaningful images and curves which as the same time carry maximum amount of information about the image. Applying this new constrained ICA algorithm shows better results compared to either purely data driven or purely model based analyses. For more information please refer to my talk at the Sunnybrook Health Sciences Centre.
I worked on Elastography during my master's program. Elastography is imaging mechanical properties of soft tissues in order to detect any abnormality in the body and also to diagnose type of the abnormality. I worked on modeling the tissue using non-linear (Hyperelastic) models, which provide more realistic models of the imaged soft tissue compared to the classical Elastography techniques that assume a linear model for the tissue[PDF]. I demonstrated the feasibility of using non-linear models of soft tissues for abnormality detection and diagnosis[PDF]. I also developed an optical flow based large motion estimation technique for soft tissue displacement measurement via taking advantage of the mechanical model of the tissue[PDF]. For more information you can refer to my master's thesis or check my thesis defense talk at University of Western Ontario.
My undergraduate thesis was on Iris Recognition which is one of the most accurate and non-duplicable personal identification techniques. I developed a feature extraction algorithm based on orientations of Gabor filter and Daubechies2 wavelet transform[PDF]. This feature is used in a matching algorithm to recognize identity of the person based on his/her eye scan. I developed a Graph Cut based segmentation technique for segmenting pupil area from the rest of the eye[PDF]. I also developed a robust segmentation technique for segmenting Iris region from the rest of the eye[PDF]. For more information please refer to my publication or if you can read Persian, you may check my B.Sc thesis (University of Tehran).