Paula Cerejeiras
Monogenic signal theory in medical imaging
Medical ultrasonography imaging for nodule detection is a non-invasive diagnostic test, which combines low cost, short acquisition time, and sensitivity to the number and size of abnormal nodules. However, a chief problem is that ultrasound images have low contrast, making it hard for the experts to interpret and classify the nodules detected. In this talk we discuss techniques based on the concepts of monogenic signal which aims to enhance the edges of abnormalities. Hereby, we use a combination of Riesz transforms and monogenic curvelets in order to determine the phase and phase angle of a given image. Riesz transforms have remarkable properties: they are shift- and scale-invariant, preserve L2 inner-product, and are steerable. Based on this approach, one is able to determine size and position of abnormalities present in images.