Technical Name Automatic microscopy system for Mycobacterium tuberculosis identification by using artificial intelligence deep learning
Project Operator National Cheng Kung University
Project Host 孫永年
Acid-fast bacilli sputum microscopy is one of the most effective detection methods for tuberculosis in clinical practice. The system is comprised of high-speed automatic focusing image acquisition and artificial intelligence deep learning tuberculosis identification technologies; the accuracy is over 94%, which can specifically meet the urgent needs in clinical microscopy examination.
Scientific Breakthrough
The system acquires more than 600 fields of view per smear, which is twice as many as manual microscopy. Generally, the accuracy of manual sputum microscopy is about 30%~70%, and the system can be as high as 90%~97%. The system takes about five minutes to process a smear, which saves 75% of time when compared with manual checking. After applying this system, specialists can work more efficiently.
Industrial Applicability
The system is compatible with the microscopes currently used in medical institutions and yields economic benefits. In high tuberculosis burden countries, the system provides a higher quality and efficient inspection platform to improve the screening accuracy of tuberculosis. This gives relatively early diagnosis and appropriate treatment for the patients, and prevents the disease from spreading.
Keyword Tuberculosis sputum microscopy automatic microscopy image acquisition automatic M. tuberculosis identification artificial intelligence deep learning Mycobacterium tuberculosis CNN microscope identification
  • Contact
other people also saw