Neural Attention Areas for Faster Medical Image Quality Assessment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The assessment of image quality in medical images is time-consuming and challenging for operators of medical imaging systems, particularly due to issues with patient positioning, implantable device placement, and image artifacts, leading to the need for re-takes and increased radiation exposure.
Innovation Solution
A computer-implemented method using a neural network to predict attention areas relevant to image quality in medical images, trained on a dataset of medical images with ground truth data, to assist operators in assessing image quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the Radiographer manually assesses image quality by meticulously examining numerous features, then the accuracy of image quality assessment is improved, but the time required for assessment increases significantly
Solution Approach 1:
The patent introduces an intermediary system comprising a trained machine learning model that acts as a mediator between the acquired medical image and the Radiographer's final decision. The model pre-identifies potential quality issues and highlights critical areas, allowing the Radiographer to focus their meticulous examination on specific regions of concern rather than manually examining every feature from scratch, thus maintaining assessment accuracy while reducing overall assessment time
Solution Approach 2:
The system performs preliminary action by automatically analyzing the medical image before the Radiographer's manual assessment. The trained model pre-identifies potential quality issues, generates quality scores, and highlights areas of concern, thereby preparing the groundwork for the Radiographer's detailed examination and reducing the time required for their meticulous review
2Reliability
If re-take images are acquired to ensure adequate image quality, then the reliability of diagnosis is improved, but the radiation dose to the patient increases
Solution Approach 1:
The patent implements a feedback mechanism where the trained model continuously evaluates image quality and provides immediate feedback to the Radiographer about potential issues. This real-time feedback allows for quality assurance before the image is finalized, reducing the need for re-takes and thereby minimizing additional radiation exposure to the patient while maintaining diagnostic reliability
Solution Approach 2:
The system performs preliminary quality assessment and identification of potential issues before the imaging procedure is completed. By detecting positioning errors, artifacts, or other quality problems in advance, the system enables corrective actions to be taken during the current imaging session rather than requiring additional re-take procedures, thus reducing cumulative radiation dose
3Manufacturing precision
If the Radiographer repetitively positions and adjusts the patient to achieve satisfactory image quality, then the image quality is improved, but the workflow efficiency decreases
Solution Approach 1:
The trained model performs preliminary analysis of the acquired image to identify positioning errors and quality issues before the Radiographer initiates repetitive repositioning and re-acquisition cycles. By providing immediate feedback on positioning accuracy and predicting potential quality problems, the system enables the Radiographer to make informed adjustments more efficiently, reducing the number of repetitive positioning iterations needed
Solution Approach 2:
The patent replaces the manual, trial-and-error mechanical process of repetitive patient repositioning and image re-acquisition with an automated intelligent system. The trained model automatically detects positioning errors and provides targeted guidance, substituting the inefficient mechanical cycle of repositioning-adjustment-reacquisition with a more efficient detection-guidance-correction workflow
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
A computer-implemented method of supporting the assessment of image quality in a medical image, is provided. The method comprises: receiving medical image data comprising a medical image (110) representing an anatomical region of a patient; inputting the medical image data into a neural network; predicting, using the neural network, an indication of one or more attention areas (120) in the medical image (110) relevant to the assessment of image quality, in response to the inputting; and outputting the indication of the one or more attention areas (120). The neural network is trained to predict the indication of the one or more attention areas (120) using training data comprising a plurality of training medical images representing the anatomical region, and for each training medical image, corresponding ground truth data comprising an indication of one or more attention areas for the training medical image.