Engineers at Deakin University in
Lung disease, including cancer, is usually detected with the aid of CT (computed tomography) and MRI (magnetic resonance imaging) scans.
However, interpreting the results of these tests can be challenging and may lead to false detections.
Dr Kouzani said: ‘Currently, expert radiologists need to view many images per patient to try and identify nodules that may be cancerous.
‘This large amount of data increases the complexity of inspection and interpretation.
‘Recent studies show that radiologists can differ in their interpretation of nodules in one patient.
‘Automated approaches can therefore help improve the precision of lung nodule detection and serve as a preliminary interpreter to assist radiologists.
‘We have developed a system that can automatically identify lung nodules of varying sizes and shapes in CT images as a tool to improve the accuracy of cancer detection.’
While other automated methods have been developed over the years, the Deakin system has proved to be more accurate.
Dr Kouzani added: ‘We used many sample nodule, non-nodule and false detection patterns to effectively train the system.
‘The experimental results demonstrate that the system performs well.
‘The nodule detection rate is higher than that of the existing systems and at the same time, the false detection rate is lower.’
The researchers have been working on the project for a year and a half and expect the system to be available for hospital trials by early next year.