Deep Learning Vertigo Diagnosis via Eye Head Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for diagnosing vertigo, such as Frenzel glasses and videonystagmography, are prone to variability in diagnostic accuracy due to human observation and require expensive equipment, limiting their widespread use.
Innovation Solution
A device and method utilizing a deep learning model to automatically generate information on vertigo by tracking and quantifying eye and head movements from video recordings, removing noise such as eye blinking, and calculating a gain to assess vestibular function without the need for complex equipment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Frenzel glasses or video Frenzel glasses are used to directly observe eye movement, then diagnostic information can be obtained, but diagnostic accuracy varies depending on doctor's experience and observation skill
Solution Approach 1:
The patent replaces the manual observation and assessment method (mechanical/physical observation through Frenzel glasses) with an automated image processing system using deep learning models. The system automatically detects eye position, tracks eye movement, and generates diagnostic information, eliminating human variability in observation and assessment.
Solution Approach 2:
The system enables self-assessment capability where the automated algorithm independently performs eye movement detection, tracking, and diagnostic information generation without requiring expert human intervention. The deep learning model self-corrects and standardizes the diagnostic process.
2Measurement precision
If videonystagmography equipment with gyro sensor and eye tracker is used to objectify test results, then diagnostic accuracy increases, but equipment cost becomes expensive
Solution Approach 1:
The patent extracts only the essential function needed for eye movement tracking from complex videonystagmography equipment. Instead of using gyro sensors and dedicated eye trackers, the system uses standard video recording devices combined with deep learning-based image processing to achieve the same objective measurement function.
Solution Approach 2:
The system creates a virtual copy of the eye tracking function through software algorithms rather than physical hardware. The deep learning model processes video frames to generate eye movement data, effectively copying the functionality of expensive eye tracking equipment using computational methods.
3Loss of information
If current videonystagmography equipment is used to observe eye movement, then eye movement can be recorded, but noise from eye blinking distorts test results
Solution Approach 1:
The system performs preliminary classification of video frames into blinking states and non-blinking states before generating final eye movement measurements. By identifying and separating blinking frames in advance, the system prevents noise contamination of the eye movement data.
Solution Approach 2:
The patent converts the harmful effect of eye blinking (noise in eye movement data) into a useful classification feature. The deep learning model learns to recognize blinking patterns and uses this information to selectively exclude or correct affected frames, transforming a source of error into a quality control mechanism.
Data Source
AI summary
Provided are a device and method for generating information on vertigo by tracking changes in eyes and head position, a computer-readable recording medium, and a computer program. More specifically, according to the device and method, movement of a patient's eyes and head is tracked in a video of the patient on the basis of a deep learning model to generate information on vertigo. The computer program for implementing the method is stored in the computer-readable recording medium.