Image processing is being developed around the world for the manufacturing environment, but a South African research team is pushing the boundaries and finding ways in which the technology can be used to provide better healthcare.
Professor Tshilidzi Marwala, the chair of systems engineering at the University of the Witwatersrand (Wits), says his passion for science became deep-rooted when he won the National Youth Science Olympiad in 1989. His prize took him to the United Kingdom to visit the University of Oxford and the University College of London, where he was exposed to the latest developments.
Passing matric, he attended Case Western Reserve University in the United States and, after completing his degree, he returned to South Africa to work for the CSIR.
He went on to obtain his master’s degree at the University of Pretoria and then it was back to the United Kingdom for his PhD at the University of Cambridge, which he followed with a post-doctorate fellowship at Imperial College.
Marwala’s PhD was on artificial intelligence and one of the applications for this technology is in image processing. This work found application when he joined SABMiller.
”One of the projects we worked on was using image processing to assess the returnable bottles. Thousands, if not millions, of bottles come in and they have to be checked for any cracks and other defects.
”A camera is used to look into the bottle and, based on the image of the bottle, it is predicted whether or not there is a crack,” Marwala says.
He says the accuracy of the system had to be very high because companies such as SABMiller work to the Six Sigma standards. This means there can be no more than three mistakes out of every million.
When Marwala joined Wits in 2003, he and his team decided to take image processing in a different direction.
”Instead of using an image to determine whether a bottle has high integrity or not, we started using images to diagnose diseases,” Marwala says.
One disease targeted by the team was pulmonary embolism or a blockage in the patient’s lungs.
An X-ray image of the patient’s lungs was scanned and the image processing system was used to determine if the patient was suffering from the condition.
”We carried out the project in partnership with Denel, developing the application so that military personnel can be accurately diagnosed in the field, where they do not have access to doctors and hospitals. In that situation you need technology that can perform tasks that are usually carried out by highly trained human beings.
”Funding from Denel was matched by a grant we received from the National Research Foundation’s technology and human resources for industry programme [Thrip],” Marwala says.
Having taken the leap of turning an engineering application developed for the manufacturing environment to healthcare uses, the research team is working to expand its functionality to diagnose other diseases and injuries.
Marwala says the increasing computing power of small devices, as well as the very sophisticated communications technologies available today, mean that a relatively small device can be taken into the field and x-ray images can be scanned and processed on the ground.
Taking the technology a step further, Marwala’s team took on another National Research Foundation project, using image processing to help treat patients needing radiation therapy.
”We worked on the project in partnership with iThemba Labs. When someone needs radiation therapy for a condition such as a brain tumour, the patient is scanned to determine the exact location of the tumour and then a beam of radiation is shot at the tumour to kill it.
”However, the whole process has to be carried out at high speed to avoid the slightest movement from the patient and in this case even the patient blinking can shift the target area,” Marwala says.
Another image processing application project the team has taken on is what are termed coded apertures.
One of the key factors in using images in medical applications is the clarity of the images examined by doctors or processed by image processing systems.
”We developed algorithms that take medical images and enhance the clarity. The project has resulted in us registering a provisional patent on the technology, as it has application in a broad range of areas,” Marwala says.