Handheld Fundus Imaging With Real-Time Retina Alignment Guidance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fundus imaging devices, such as the Optomed Aurora ®< IQ, require skilled operation to correctly align the camera for capturing still images or videos of the fundus, posing challenges for less experienced users.
Innovation Solution
A handheld fundus imaging apparatus equipped with a machine vision algorithm using AdaBoost for retina detection, combined with a user interface that provides real-time alignment aids, facilitates accurate alignment and capture of fundus images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If traditional fundus imaging devices are used, then imaging capability is achieved, but operation difficulty increases for less experienced users
Solution Approach 1:
The system performs automatic retina detection and alignment using machine vision algorithms with AdaBoost classification. The device autonomously identifies the retina location and guides alignment without requiring manual adjustment by the operator, enabling the system to serve itself in the alignment process.
Solution Approach 2:
The patent replaces manual mechanical alignment operations with automated optical and computational methods. Machine vision algorithms analyze captured images to detect retina features and calculate alignment parameters, substituting the need for skilled manual positioning with automated image processing and classification systems.
2Measurement precision
If manual alignment methods are used, then device simplicity is maintained, but detection precision decreases
Solution Approach 1:
The system implements a feedback loop where captured images are analyzed by machine vision algorithms to detect retina features, and the results are used to generate alignment guidance displayed to the operator. This closed-loop feedback enables precise retina detection and alignment by continuously monitoring and adjusting based on image analysis results.
Solution Approach 2:
The patent introduces an intermediary machine vision system that acts as a mediator between the camera and the operator. The AdaBoost-based classification algorithm processes raw images to extract retina features and generates alignment information, serving as an intelligent intermediary that enhances detection precision without requiring direct manual intervention.
3Ease of operation
If automated retina detection is implemented, then ease of use improves, but processing time increases
Solution Approach 1:
The system performs preliminary action by pre-training AdaBoost classification algorithms with labeled retina image data before actual use. This preliminary training phase enables the system to rapidly classify and detect retina features during operation, reducing processing time during actual imaging while maintaining high detection accuracy through pre-learned patterns.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6B
AI summary
Apparatus and method for imaging fundus of eye. The apparatus (100) includes an optical system (116), an image sensor (114) to capture a still image or a video through the optical system (116), a user interface (102) including a display (104) to display data to a user of the apparatus, and one or more processors (106). The one or more processors (106) cause performance of at least the following: setting the image sensor (114) to capture an aiming video; detecting a retina of the eye in the aiming video; setting the display (104) to display the aiming video with a highlight of the detected retina; and setting the image sensor (114) to capture one or more final still images of the fundus or a final video of the fundus.