Hang Seng Innovative Design Competition 2004

Champion: Lai Tsz Chung and Law Wai Kong

The First Semi-automatic Cirrhosis Diagnosis System in Hong Kong

Mr. Lai Tsz Chung, fresh graduate of BEng in Computer Science, and Mr. Law Wai Kong, fresh graduate of BEng in Computer Engineering, have been awarded champion of the Hang Seng Innovative Design Competition 2004, which were held on June 19, 2004 (Sat) at The Hong Kong University of Science and Technology. The competition is amongst the best Final Year Projects in the School of Engineering at HKUST.

The award-winning project is "Application of image processing techniques to tissue texture analysis and image compression", under supervision of Prof. Albert Chung of the Department. This project is considered to be the first semi-automatic software for cirrhosis check in Hong Kong, and the accurracy is up to 80%. It is planned that the technology will have a clinical test at Whales Hospital.

"It is so exciting that we could win the champion of this competition. I think the project itself is meaningful, which could contribute to the clinical aspect. I wish to thank our advisor, Prof. Albert Chung, and our clinical consultant, Dr.Simon Yu (Radiologist from the Prince of Wales Hospital, Hong Kong), for their support and advise in this project.", the winners said.

According to World Health Organization (WHO), cirrhosis causes hundreds of thousands of deaths worldwide each year. Nowadays, cirrhosis detection mainly relies on manual analysis of ultrasonic liver images and histological analysis of patients' liver samples. Detection results are
highly dependent on the experience and ability of sonographers to observe certain confusing textural characteristics. Histological analysis is more accurate, but it is invasive and involves complicated processes which require much more time and resources.

"Severity information plays an essential role in the cirrhosis medical treatment. In this project, we aim to provide a semi-automatic three severity levels cirrhosis classification system. Ultrasound scanning machine calibration and machine dependent parameters are not required by our system. Without machine dependent measurement, this classification system is able to deliver classification result quickly after basic training procedure. With the aid of this system, patients will be able to obtain preliminary diagnostic results within several minutes once ultrasonic liver scan is available", Mr. Law said.

Ten-fold Stratified Cross-validation is performed for accuracy testing. It is proved that the system can provide up to 81% accuracy for classification of "with-or-without cirrhosis", and 71% accuracy for classification of "no, early or advanced cirrhosis".

The project has been reported in