Automated Change Detection of Serial MR Images Using Compressed Sensing

We aim to develop a new algorithm to detect changes in serial MR examinations. Computer‐based change detection system is an important tool to automatically process abundant information produced by imaging systems and assist physicians to identify clinically important changes in the images. Many existing methods are computationally costly. The emerging mathematical theory of compressed sensing (CS) allows perfect recovery of a sparse signal from fewer samples. In serial MR images, the stationary content and its variations due to noise and acquisition‐related changes are compressible and changes over time are spatially sparse. We thus utilize the CS techniques to develop novel change detection algorithms that reduce computational complexity and extract information robust to the presence of noise. Our algorithm is expected to aid radiologists to make early diagnosis and be used as a training tool for young practitioners as well as to advance the development of change detection technology at the A.U.G. Signal for geological applications.

Faculty Supervisor:

Dr. Hongmei Zhu

Student:

Varvara Nika

Partner:

Hospital for Sick Children

Discipline:

Mathematics

Sector:

Medical devices

University:

York University