MTANN Rib Suppression for Lung Nodule Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computer-aided diagnostic (CAD) schemes for lung nodule detection in chest radiographs face challenges in detecting nodules overlapping with ribs and clavicles, leading to high false positive rates and reduced sensitivity and specificity due to the difficulty in suppressing the contrast of these anatomical structures.
Innovation Solution
A multi-resolution massive training artificial neural network (MTANN) is employed to modify the appearance of ribs and clavicles in chest radiographs, using dual-energy subtraction techniques and multi-resolution decomposition/composition methods to effectively suppress their contrast, thereby enhancing the detection of lung nodules.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional CAD schemes are used for lung nodule detection, then the detection process is straightforward, but the false positive rate increases and detection accuracy decreases due to rib and clavicle interference
Solution Approach 1:
The patent extracts and removes the interfering bone structures (ribs and clavicles) from the chest radiograph image by training a neural network to generate a bone structure mask, which is then subtracted from the original image. This extraction approach eliminates the harmful interference from bone structures while preserving lung nodules for accurate detection
Solution Approach 2:
The patent applies different processing qualities to different regions of the image: bone structures are suppressed with high contrast reduction, soft tissues are maintained with moderate contrast, and lung nodules are preserved with enhanced visibility. This local quality differentiation allows selective modification of image regions based on their diagnostic importance
2Reliability
If the contrast of ribs and clavicles is suppressed to improve nodule visibility, then detection sensitivity improves, but the visibility of soft tissues may be compromised
Solution Approach 1:
The neural network is trained to apply different contrast modification levels to different tissue types: aggressive contrast suppression for bone structures (ribs and clavicles), moderate preservation for soft tissues, and enhanced visibility for lung nodules. This localized quality control ensures that suppressing bone contrast does not compromise soft tissue visibility
Solution Approach 2:
The patent dynamically adjusts the contrast parameter for different anatomical structures based on their type and diagnostic importance. Bone structures undergo significant contrast reduction, soft tissues maintain their original contrast or experience slight enhancement, and lung nodules are enhanced for better detection. This parameter differentiation resolves the contradiction between bone suppression and soft tissue preservation
3Measurement precision
If a trained image processing device is used to modify anatomical structure appearance, then detection accuracy improves, but the device complexity and training requirements increase
Solution Approach 1:
The patent introduces a trained neural network as an intermediary device between the original chest radiograph and the final detection process. This intermediary processes the image to suppress bone structures while preserving lung nodules, thereby improving detection accuracy. The complexity is justified by the significant performance improvement in detecting nodules that overlap with bone structures
Data Source
AI summary
A method, system, and computer program product for modifying an appearance of an anatomical structure in a medical image, e.g., rib suppression in a chest radiograph. The method includes: acquiring, using a first imaging modality, a first medical image that includes the anatomical structure; applying the first medical image to a trained image processing device to obtain a second medical image, corresponding to the first medical image, in which the appearance of the anatomical structure is modified; and outputting the second medical image. Further, the image processing device is trained using plural teacher images obtained from a second imaging modality that is different from the first imaging modality. In one embodiment, the method also includes processing the first medical image to obtain plural processed images, wherein each of the plural processed images has a corresponding image resolution; applying the plural processed images to respective multi-training artificial neural networks (MTANNs) to obtain plural output images, wherein each MTANN is trained to detect the anatomical structure at one of the corresponding image resolutions; and combining the plural output images to obtain a second medical image in which the appearance of the anatomical structure is enhanced.