Two-Stage Medical Image Classification System
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing volume of medical images burdens radiologists, as conventional methods and machine learning algorithms are inefficient in filtering normal images from abnormal ones, leading to wasted time on reviewing non-actionable data and missing subtle findings.
Innovation Solution
A two-stage classification system that uses trained classifiers to differentiate between normal and abnormal medical images, with the first stage confirming normalcy and the second stage detecting abnormalities, reducing the workload for radiologists by accurately filtering out non-actionable data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If radiologists review all medical images manually, then diagnostic accuracy is maintained, but workload and time consumption increase significantly
Solution Approach 1:
The patent introduces an automated classification system as an intermediary between image acquisition and radiologist review. This system uses machine learning algorithms to pre-screen images and generate classification results, serving as a mediator that filters normal images before they reach radiologists, thereby reducing workload while maintaining diagnostic accuracy through automated preliminary analysis
Solution Approach 2:
The patent segments the image review process into two distinct stages: automated classification stage and radiologist review stage. The automated system handles the initial screening and classification of normal vs. abnormal images, while radiologists focus only on reviewing images flagged as abnormal or uncertain, thus dividing the workload efficiently between machine and human expertise
2Productivity
If machine learning algorithms are used to filter normal images, then radiologist workload is reduced, but false negatives may occur
Solution Approach 1:
The automated classification system performs preliminary screening and filtering of normal images before radiologist review. By conducting this preliminary action, the system prepares the workflow by identifying and flagging only potentially abnormal images for radiologist attention, reducing workload while maintaining safety through subsequent human verification of flagged cases
Solution Approach 2:
The system implements feedback mechanisms where radiologist reviews of classified images provide ground truth labels that are used to continuously train and improve the classification algorithms. This feedback loop ensures that false negatives are identified and corrected over time, progressively reducing the false negative rate while maintaining workload reduction benefits
3Device complexity
If single-stage classification is used, then system complexity is reduced, but diagnostic accuracy decreases
Solution Approach 1:
The classification system is segmented into multiple sequential stages: initial automated classification, radiologist review of flagged images, and feedback-based reclassification. This multi-stage segmentation allows each stage to specialize in specific tasks, improving overall classification accuracy while managing complexity through modular design where each stage builds upon the previous one
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Provided are computer-implemented methods and systems for classifying a medical image data set. In particular, a method is provided comprising the steps of receiving the medical image dataset of a patient, of providing a first classification stage configured to classify the medical image dataset as normal or not-normal, of providing a second classification stage different than the second classification stage and configured to classify the medical image dataset as normal or not-normal, and of subjecting the medical image dataset to the first classification stage so as to classify the medical image dataset as normal or not-normal. Further, the method comprises subjecting the medical image dataset to the second classification stage so as to classify the medical image dataset as normal or not-normal, if the medical image dataset is classified as normal in the first classification stage.