Automatic Coronary Sinus Segmentation with Multi-Plane Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current medical image segmentation methods, particularly for the coronary sinus, are time-consuming and prone to human error, lacking accuracy and reproducibility due to significant patient-to-patient variation and the complexity of the structure, with existing algorithms failing to effectively segment this anatomical structure.
Innovation Solution
An automatic segmentation method using neural networks processes 3D CT images to generate a 3D mask of the coronary sinus by decomposing the image into 2D planes and employing multiple neural networks for probability mapping, followed by a weighted ensemble and thresholding to create a precise 3D model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual segmentation is performed by experienced cardiologists, then segmentation accuracy is improved, but segmentation time increases significantly (approximately 1 hour per case)
Solution Approach 1:
The patent replaces the manual mechanical segmentation process performed by cardiologists with an automated computational system based on neural networks. The system processes 3D CT images through multiple neural networks that perform probability mapping and weighted ensemble operations to automatically generate segmented coronary sinus masks, eliminating the need for time-consuming manual tracing while maintaining high accuracy through algorithmic precision.
Solution Approach 2:
The patent creates a digital 3D model copy of the coronary sinus from the patient's CT scan data. By generating a virtual representation that can be processed computationally, the system allows for rapid automated segmentation without requiring physical manual measurement or tracing, thus reducing time while preserving the anatomical details needed for accurate segmentation.
2Productivity
If known algorithms are used for segmentation, then processing speed is improved, but segmentation accuracy deteriorates due to significant patient-to-patient variation in coronary sinus structure
Solution Approach 1:
The patent employs neural networks that can dynamically adapt to varying anatomical parameters across different patients. The system processes 3D images by decomposing them into 2D planes and using multiple neural networks to capture the variability in coronary sinus structure. This allows the algorithm to maintain high accuracy across diverse patient populations while operating automatically at high speed.
Solution Approach 2:
The patent transforms the 3D coronary sinus segmentation problem into a series of 2D probability mapping problems that can be solved more efficiently by neural networks. By decomposing the 3D volume into 2D slices and processing them through multiple neural networks with weighted ensemble, the system achieves both high speed and high accuracy in handling patient variability.
3Adaptability or versatility
If manual segmentation is performed, then adaptability to patient variation is improved, but reproducibility deteriorates due to human factor involvement
Solution Approach 1:
The patent replaces human cardiologist segmentation with an automated neural network system that eliminates human variability. The computational algorithm consistently applies the same processing steps to all patients, ensuring reproducible results while maintaining adaptability to individual anatomical variations through the flexibility of neural network-based probability mapping and ensemble methods.
Data Source
AI summary
Method, executed by a computer, for identifying a coronary sinus of a patient, comprising: receiving a 3D image of a body region of the patient; extracting 2D axial images of the 3D image taken along respective axial planes, 2D sagittal images of the 3D image taken along respective sagittal planes, and 2D coronal images of the 3D image taken along respective coronal planes; applying an axial neural network to each 2D axial image to generate a respective 2D axial probability map, a sagittal neural network to each 2D sagittal image to generate a respective 2D sagittal probability map, and a coronal neural network to each 2D coronal image to generate a respective 2D coronal probability map; generating, based on the 2D probability maps, a 3D mask of the coronary sinus of the patient.