Machine Learning Image Classifier for Medical X-Ray Orientation Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In medical imaging, improper patient positioning during x-ray imaging can lead to false positives in diagnosing conditions like thoracic aortic aneurysms, as it is difficult to distinguish true abnormalities from false positives, especially when patients are rotated or imaged at off-center angles, resulting in unnecessary and costly computed tomography (CT) scans.
Innovation Solution
A computer system uses machine learning techniques to identify patient rotation angles and medical conditions by training a machine learning image classifier model with three-dimensional imaging data, generating artificial x-ray images, and converting between anterior-posterior and posterior-anterior views, thereby reducing false positives and improving diagnostic accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional x-ray imaging is used without accounting for patient orientation, then the imaging process is simple and quick, but diagnostic accuracy deteriorates due to false positives from improper positioning
Solution Approach 1:
The system performs preliminary detection of patient rotation angles and projection types before the main diagnostic classification. An image classifier model first analyzes the x-ray image to determine patient orientation and projection angle, then this information is used to adjust the subsequent medical condition classification, preventing false positives from improper positioning
Solution Approach 2:
The patent introduces an intermediary image classifier model that acts as a mediator between the raw x-ray image and the medical condition classifier. This intermediate step detects patient orientation and provides correction information, enabling the main classifier to accurately diagnose medical conditions regardless of patient positioning or projection type
2Reliability
If machine learning models are trained without considering patient rotation, then training data generation is simpler, but classification reliability deteriorates due to false positives from rotated images
Solution Approach 1:
During training data generation, the system performs preliminary rotation of x-ray images to create multiple versions at different angles. These rotated images are then used to train the medical condition classifier, enabling it to recognize medical conditions accurately regardless of the patient's actual rotation angle during imaging
Solution Approach 2:
The training process is designed to be universal by incorporating images at multiple rotation angles and projection types (PA and AP views). The classifier model learns to handle various imaging conditions simultaneously, making it universally applicable to different patient positions and reducing false positives from improper positioning
3Object-affected harmful factors
If false positives are not reduced, then all potential cases can be reviewed, but unnecessary CT scans increase radiation exposure and healthcare costs
Solution Approach 1:
The system uses feedback from the image classifier model that detects patient orientation and projection type. This feedback information is used to adjust the interpretation of x-ray images, allowing the system to distinguish between apparent abnormalities caused by positioning and true medical conditions, thereby reducing false positives and unnecessary follow-up scans
Data Source
AI summary
A computer system identifies a medical condition in a patient. A trained machine learning image generator is used to generate a set of training images based on three-dimensional patient imaging data, wherein each training image is labeled with a projection angle of the corresponding two-dimensional projection. Using the set of training images, a machine learning image classifier model is trained to identify patient rotation angles in x-ray images. X-ray images are processed with the machine learning image classifier model to identify patient rotation angles. A machine learning medical condition classifier model is trained to identify a medical condition using the labeled x-ray images. The machine learning medical condition classifier model determines an indication of the medical condition in a patient's x-ray image. Embodiments of the present invention further include a method and program product for identifying a medical condition in a patient in substantially the same manner described above.