Development of Intelligent Microscopic Image Detecting Platform for Microorganisms

Project: National Science and Technology CouncilNational Science and Technology Council Academic Grants

Project Details


Optical microscopes are the most useful and cost acceptable pieces of equipment widely used in clinical examinations of microorganisms such as parasite eggs, wastewater recycling and health checks of foreign laborers. Although a well-trained technician can identify a single target in minutes, however, microscopic examination and quantitation of mixed helminth infection eggs is still a horrible even for experienced medical technologists. In the last two decades, there are many computational methodologies published for microscopic examinations. However, most laboratories still use manual eye-inspections. The human validating step is too complicated to achieve, and costs of the system are too high to deploy. Currently, AI cannot replace human experts, instead, it can do a great job on pre-scanning. The present three years proposal will focus on bridging both human experts and machine predictions by developing a high accuracy microscopic image detection architecture and establishing an efficient human-validating platform. In the first year, we propose to digitalize microscopic images of microorganisms and helminth eggs under high- and low magnification ratios. Morphological signatures that have been proof informative for microbial classification are extracted and integrated into the middle layer of the object detection deep-learning architecture. In the second year, the AI models generated for the high magnification ratio will be used to develop a new recognition model for low magnification ratio. Then a distributed computing system for object detection of microscopic images will be developed. In the third year, an intelligent microscopic image analysis system for microorganisms based on the benefits of both low-/high- magnification ratios will be established to convert the existing microscopic image capturing system into an AI-based target recognition system without further costs. A virtual microscope that is highly friendly to manual validation will be developed; this system can serve as an ideal education platform for training of medical technicians. Besides, a platform for microscopic image detection community will be initialized. Results for the present study can be used as a basics to develop a highly efficient and cost acceptable microscopic examination system based on deep-learning strategies. A smoothing workflow of AI-assistant microbial identification abilities will be developed to improve the clinical microscopic examinations by lowering the human workloads and increasing the efficiencies.

Project IDs

Project ID:PB10907-4374
External Project ID:MOST109-2221-E182-049
Effective start/end date01/08/2031/07/21


Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.