3D Ultrasound Ocular Segmentation via Deep Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for acquiring, processing, and visualizing ultrasound imaging of the eye, particularly for conditions like glaucoma, are underutilized due to a lack of clinical expertise and the need for specialized ultrasonographers, limiting effective diagnosis and treatment.
Innovation Solution
An apparatus and method utilizing deep learning models for segmenting ocular structures, noise reduction, and 3D volume rendering of the eye, enabling improved alignment, noise reduction, and visualization of ultrasound data, including the use of graphical user interfaces to facilitate easier clinical workflows.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning models and automated processing methods are implemented, then diagnostic accuracy and efficiency are improved, but device complexity and implementation difficulty increase
Solution Approach 1:
The patent introduces deep learning models as intermediary components between the ultrasound imaging system and clinical decision-making. These models automatically perform segmentation, noise reduction, and feature extraction, acting as intelligent mediators that translate raw imaging data into diagnostically relevant information without requiring manual expert intervention at each processing stage.
Solution Approach 2:
The automated processing pipeline enables the system to perform diagnostic tasks independently through self-service mechanisms. The deep learning models automatically segment ocular structures, reduce noise, and generate diagnostic metrics without requiring continuous human oversight, allowing the system to serve itself in performing complex analytical functions that previously required specialized ultrasonographers.
2Measurement precision
If specialized ultrasonographers are required for accurate imaging and analysis, then measurement precision is improved, but ease of operation and accessibility deteriorate
Solution Approach 1:
The system empowers non-specialized operators to achieve expert-level diagnostic accuracy through automated deep learning processing. The intelligent algorithms automatically perform tasks that previously required specialized training, including image acquisition optimization, noise reduction, structural segmentation, and diagnostic metric calculation, enabling the system to serve itself in performing expert-level analysis without human specialists.
Solution Approach 2:
The patent replaces the mechanical dependency on human expert skills and manual processing techniques with automated computational systems. Deep learning models substitute for the manual expertise of specialized ultrasonographers, automatically performing image enhancement, segmentation, and diagnostic analysis through algorithmic processing rather than human manual intervention.
3Reliability
If multiple surgical interventions and diagnostic exams are conducted for complex conditions like childhood glaucoma, then treatment effectiveness is improved, but loss of time and treatment burden increase
Solution Approach 1:
The patent performs preliminary diagnostic actions by automatically analyzing ultrasound images to provide comprehensive diagnostic information in a single examination. The deep learning models pre-process and interpret imaging data to deliver accurate diagnoses and treatment planning information before surgical intervention, reducing the need for multiple follow-up diagnostic exams and helping to minimize the total treatment time required for complex conditions.
Data Source
AI summary
A first set of embodiments relates to an apparatus comprising one or more processors configured to: access three-dimensional (3D) ultrasound imaging of an eye; generate at least one segmented ocular structure by segmenting at least one ocular structure represented in the 3D ultrasound imaging using at least one deep learning ocular structure segmentation model configured to generate a predicted segmentation volume of the at least one ocular structure based on at least one portion of the 3D ultrasound imaging; compute at least one clinical metric associated with the at least one segmented ocular structure based on the at least one segmented ocular structure; and display at least one of: the at least one segmented ocular structure, the at least one clinical metric, the 3D ultrasound imaging, or at least one portion of the 3D ultrasound imaging.