High Throughput Computer-Aided Fluorescence Diagnosis System
20140324 - 20150923
Dr Xiaohua Wu
We plan to achieve the following deliverables within the fifteen-month project period: 1. Develop a first-ever high throughput fluorescence microscopic scanning platform with real-time auto focusing and adaptive depth of field extension capabilities, with the throughput up to ten times higher than current commercially available slide scanners 2. Develop a generic facilitating software system for fluorescence microscopy based diagnosis including image processing and object classification functionalities, with the speed and accuracy optimizable for different application requirements 3. Develop a first-ever automated Tuberculosis whole smear screening system by customizing the developed high throughput scanning platform and diagnosis facilitating software, achieving sensitivity significantly higher than current manual microscopic diagnosis 4. Conduct a large scale evaluation test for the developed smear screening system with a collaborating party to quantitatively measure its performance in speed, sensitivity and specificity, and prepare the test report for further system development
Mr Tony Leung Dr Sammy Wong Dr Ivy Law Mr Harley Li Mr Hanbin Jian
Department of Microbiology, Chinese University of Hong Kong Henan Center for Disease Control and Prevention TB Reference Laboratory Kindstar Global MetaSystems Asia Co., Ltd. [Sponsor] Speed Fair Co., Ltd (Technology Licensing) [Sponsor] Speed Fair Co., Ltd. [Sponsor] Stanley Ho Centre for Emerging Infectious Diseases, Chinese University of Hong Kong
Pathology diagnostics is currently facing both opportunities and challenges. On the one hand, more advanced microscopy techniques such as fluorescence and confocal microscopy are migrating from research laboratories to clinical benches, offering pathologists powerful tools to tackle complex tasks like cancer diagnosis, while such imaging equipments often require manual operation with low throughput. On the other hand, more affordable computation technologies including image processing and machine learning are emerging to provide pathologist strong assistance for their diagnostic decision making, while such software often fails to meet the strict clinical requirements on speed and accuracy. The project aims to address these technology challenges by developing a high throughput fluorescence imaging system and accompanying diagnosis facilitating software with real-time image processing capability. Such a system can be utilized for various fluorescence microscopy based diagnostic applications such as immunofluorescence, fluorescence in situ hybridization, high content assay screening and so on. In particular, as the first application example, we will focus on Tuberculosis microbiological diagnosis and develop the first ever automated whole sputum smear screening system, which will provide significantly higher sensitivity than current manual microscopic diagnosis.