Automated OCT Image Quality Control with Deep Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current retinal optical coherence tomography (OCT) image quality assessment is subjective, time-consuming, and prone to inter-rater agreement issues due to manual evaluation, which affects the accuracy of segmentation and detection of subtle changes in neurological disorders.
Innovation Solution
A fully automated quality analysis method (AQUA-OCT) using deep convolutional neural networks (DCNNs) to evaluate key image quality aspects based on the OSCAR-IB criteria, providing real-time quality control and adaptable to various OCT devices and scan types.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual evaluation by expert graders is used to assess OCT image quality, then domain knowledge and subjective judgment are applied, but the process is time-consuming and has mediocre inter-rater agreement
Solution Approach 1:
The patent replaces the manual mechanical evaluation process with an automated deep learning system. Multiple DCNNs process OCT images automatically, substituting human graders with algorithmic analysis that delivers consistent, rapid quality assessment without subjective variability or time constraints.
Solution Approach 2:
The patent creates a digital replica of expert grading capability through trained neural networks. The DCNNs are trained on labeled data to replicate and standardize the quality assessment function, producing consistent results that mirror expert judgment while eliminating inter-rater variability and processing time.
2Reliability
If manual quality control is implemented, then quality standards can be enforced, but considerable domain knowledge is required and results remain subjective
Solution Approach 1:
The patent divides the quality control task into multiple specialized DCNNs, each targeting specific quality criteria (OSCAR-IB components). This segmentation allows complex quality assessment to be broken into manageable algorithmic modules that process different aspects independently and combine results, ensuring consistent application of standards without requiring human domain knowledge.
Solution Approach 2:
The patent transforms subjective quality judgment into objective parameter-based assessment. The system evaluates quantifiable image parameters (signal strength, centering, completeness) through DCNNs, converting subjective expert opinion into measurable, consistent metrics that enforce quality standards reliably across all assessments.
3Productivity
If automated methods are used for quality analysis, then speed and consistency improve, but accuracy may be reduced without domain knowledge
Solution Approach 1:
The patent performs preliminary training of DCNNs on extensively labeled OCT data with known quality outcomes. This pre-training phase allows the system to learn from expert-annotated examples before deployment, ensuring that automated assessment achieves high accuracy by incorporating domain knowledge during training rather than requiring it during operation.
Solution Approach 2:
The patent combines multiple DCNN outputs into a comprehensive quality assessment. By merging the results of several specialized networks that evaluate different quality aspects, the system achieves both speed and accuracy, leveraging the strengths of each individual network while maintaining high overall assessment precision.
Data Source
AI summary
Retinal optical coherence tomography (OCT) with intraretinal layer segmentation is increasingly used not only in ophthalmology but also for neurological diseases such as multiple sclerosis (MS). Signal quality influences segmentation results, and high-quality OCT images are needed for accurate segmentation and quantification of subtle intraretinal layer changes. Among others, OCT image quality depends on the ability to focus, patient compliance and operator skills. Current criteria, for OCT quality define acceptable image quality, but depend on manual rating by experienced graders and are time consuming and subjective. In this paper, we propose and validate a standardized, grader-independent, real-time feedback system for automatic quality assessment of retinal OCT images. We defined image quality criteria for scan centering, signal quality and image completeness based on published quality criteria and typical artifacts identified by experienced graders when inspecting OCT images. We then trained modular neural networks on OCT data with manual quality grading to analyze image quality features. Quality analysis by a combination of these trained networks generates a comprehensive quality report containing quantitative results. We validated the approach against quality assessment according to the OSCAR-IB criteria, by an experienced grader. Here, 100 OCT files with volume, circular and radial scans, centered on optic nerve head and macula, were analyzed and classified. A specificity of 0.96, a sensitivity of 0.97 and an accuracy of 0.97 as well as a Matthews correlation coefficient of 0.93 indicate a high rate of correct classification. Our method shows promising results in comparison to manual OCT grading and may be useful for realtime image quality analysis or analysis of large data sets, supporting standardized application of image quality criteria.