Masoud Nosrati ('15) developed an algorithm to improve the speed and accuracy of medical image segmentation.

Alumnus wins prestigious dissertation award for novel medical image analysis technique

April 10, 2016
Print

Masoud S. Nosrati, a recent PhD graduate from SFU’s School of Computing Science, has received a prestigious doctoral dissertation award from the Canadian Image Processing and Pattern Recognition Society.

Nosrati, who graduated in October 2015, developed a novel technique to help computers identify key features in medical images. This groundbreaking advance could improve diagnosis and treatment for patients, and reduce healthcare costs and waiting times.

An algorithm that helps computers "see" inside the human body

The healthcare industry generates a huge amount of data every day from imaging techniques such as MRI, CT, and ultrasound scans.

While this data allows clinicians and scientists to peer inside the human body and extract valuable information, the process of visually analyzing each image can be time-consuming and error-prone.

“Different doctors, or even the same one, may give different answers at different times,” says Nosrati’s PhD supervisor professor Ghassan Hamarneh.

“There is a strong need to develop automatic, accurate, and robust computer vision systems to handle this flood of big image data.”

With this in mind, Nosrati’s thesis focuses on medical image segmentation, a process that plays a key role in automating medical image analysis by partitioning an image into meaningful parts.  

This technique can help doctors localize key structures, such as the boundary of a tumour, which can help inform diagnosis or treatment – for example, choosing between radiotherapy and surgery for cancer patients.  

Specifically, Nosrati developed algorithms to improve the speed and accuracy of segmentation in microscopy image analysis and image-guided robotic surgery. The latter superimposes a computer-generated image on the surgeon's view during an operation, enhancing the view of internal organs and tissue such as tumours.  This helps ensure that the surgeon doesn’t make unnecessary cuts and completely removes the cancerous tissue.

Above: Image-guided robotic surgery. On the right, a pre-operative 3D kidney and tumour model. With Nosrati’s framework, a surgeon can see an overlay of the locations and boundaries of these anatomical features, guiding the procedure and minimizing trauma.

About Masoud S. Nosrati

Nosrati has produced papers in publications including IEEE Transactions on Medical Imaging (IEEE TMI), International Journal of Computer Assisted Radiology and Surgery (IJCARS ), IEEE International Conference on Computer Vision (IEEE ICCV) and International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI).

He is currently a vision and machine-learning researcher at SPORTLOGiQ in Montreal, working with computer vision systems for hockey player analysis software.

Of his paper win, he says: “This is a great honour that my thesis has been acknowledged as the best PhD thesis in Canada in the area of computer vision. I will do my best to justify CIPPRS trust and contribute to the computer vision field by introducing advanced algorithms to solve real world computer vision problems.”

The Canadian Image Processing and Pattern Recognition Society (CIPPRS) Doctoral Dissertation Award is given annually to the top Ph.D. thesis in the areas covered by the Conference on Computer and Robot Vision (CRV).