Hidayetoğlu tackles complex imaging as CSE Fellow

6/8/2017 Julia Sullivan, ECE ILLINOIS

Submitted with faculty co-advisors ECE Professors Chew and Hwu, his proposal aims to decrease computing cost for complex medical imaging while preserving clarity and improving speed.

Written by Julia Sullivan, ECE ILLINOIS

ECE ILLINOIS PhD candidate Mert Hidayetoglu has been selected as one of eight 2017-18 Computational Science and Engineering (CSE) Fellows

Mert Hidayetoglu
Mert Hidayetoglu
His proposal, “Large Inverse-Scattering Solutions on Supercomputers and its Application to Imaging,” has real-world implications for decreasing the cost and computing time of complex medical imagining. 

Diffraction tomography is used to find the shape of an object, often employed by medical imaging for diagnostics. Many commercial imaging technologies, like ultrasound, are built around the simple linear relationship between the object and the scattered field of reflected waves collected by receivers. This single scattering methodology works well when the object has a low contrast to the background.

However, some imaging produces a nonlinear multiple-scattering field, resulting in a distorted image. A rigorous solution of the nonlinear problem has dramatically high computational cost and requires more than a human lifetime with traditional methods. 

Submitted with faculty co-advisors ECE Professors Weng Cho Chew and Wen-mei Hwu, Hidayetoğlu’s proposal aims to decrease computing cost for these more complex images while preserving clarity and improving speed. His first approach proposes decreasing the computational complexity with the multilevel fast multipole algorithm (MLFMA). The second employs computational power from Blue Waters. Working at the supercomputing facility at the National Center for Supercomputing Applications (NSCA), the team will spread the computational cost across a large number of parallel computing nodes.

An illustration of the proposed two-dimensional MLFMA parallelization scheme.
An illustration of the proposed two-dimensional MLFMA parallelization scheme.
“The proposed work requires knowledge from different disciplines, including waves and fields, integral equations, numerical methods, fast algorithms, parallel and high-performance computing,” Hidayetoğlu said. “For imaging with real data, we will use benchmark measurements available in electromagnetics and acoustics societies, since the proposed work can be applied to both.”

Hidayetoğlu is well-positioned to attack this research problem. Mert has been working with the electromagnetics group at ECE ILLINOIS for the last two years under supervision of Prof. Chew, co-advised by Prof. Hwu at the Coordinated Science Laboratory. Both professors are also CSE affiliates. He is the first author of more than a dozen conference papers and received the ECE ILLINOIS 2016-17 Professor Kung Chie Yeh Fellowship. Next spring, he will present his research results alongside this year’s other CSE Fellows at the annual CSE Symposium. 


Share this story

This story was published June 8, 2017.