A New On-line Learning Method for Diagnostic Medical Sonography

Supervisors: Professor Sandy Cochran, Dr Shufan Yang

The aim of this project is to investigate new on-line learning algorithms, architectures and models which are inspired by the image processing capability of the visual system in the human brain, and subsequently apply these models to Diagnostic Medical Sonography applications.

The development of Sonography provides a large number of assistive diagnosis and measurement of neuronal activities in the brain; however the process of discriminating between the images of normal brain activity with those of patients exhibiting characteristics of early stage disease is a very difficult and challenging problem, normally taking a very long time for finial diagnosis. The process is time-consuming and prone to error due to human fallibilities and fatigue, although early diagnosis could also be vital for recovery. Thus, a new image processing technology is required to record very long ultrasound sweeps with arbitrary trajectories.

The project is built on the supervisors’ team previous work experience in the reconfigurable computing and biological-inspired spiking neural modelling to dedicated reconfigurable hardware based on Field Programmable Gate Arrays (FPGA). Emulating biological signal processing on an FPGA is an economical option for complex systems modelling, prior to proceeding to fully integrated circuit design and fabrication. This work is an important step towards the eventual goal of incorporating advanced system-on-chip module into a realistic ultrasound-guided diagnoses system. The results of this research can be applied to solve problems not only in the medical imaging domain but also in other industrial/commercial image processing applications.