Medical Image Segmentation Using Anatomical Landmark Positioning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current medical image processing methods for image segmentation, such as those used in nuclear magnetic resonance imaging and ultrasound-guided procedures, face challenges with low precision and the risk of over-segmentation due to close pixel values between the prostate and bladder wall, and require extensive training data and time for machine learning models, leading to instability in detection.
Innovation Solution
A medical image processing apparatus and method that detects feature points and generates feature target regions based on their positional relationships, using a combination of machine learning models and deep learning techniques to specify a target region accurately, thereby performing precise image segmentation while preventing over-segmentation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning models are used for image segmentation, then segmentation accuracy can be improved, but training time and data requirements increase
Solution Approach 1:
The patent extracts and stores anatomical landmark positions and spatial relationships from pre-labeled medical images as training data. This preliminary extraction of feature points and their geometric relationships enables the machine learning model to be trained on pre-processed anatomical information, reducing the need for time-consuming manual annotation and accelerating the training process while maintaining segmentation accuracy.
2Measurement precision
If machine learning models are used for image segmentation, then segmentation accuracy can be improved, but detection stability deteriorates
Solution Approach 1:
The patent incorporates a feedback mechanism where the system iteratively refines segmentation results by comparing predicted anatomical landmark positions with ground truth labels. The model adjusts its parameters based on the discrepancy between predicted and actual landmark positions, enabling continuous improvement of detection stability and accuracy through iterative training and validation cycles.
Solution Approach 2:
The patent performs preliminary extraction and verification of anatomical landmark positions from pre-labeled images, creating a curated dataset of verified feature points and their spatial relationships. This preliminary preparation ensures that the training data is of high quality and consistent, which significantly improves the stability and reliability of subsequent detection results.
3Measurement precision
If manual drawing is performed for image segmentation, then segmentation precision can be controlled, but work efficiency decreases
Solution Approach 1:
The patent implements a self-service approach where the machine learning model automatically performs image segmentation by predicting anatomical landmark positions and generating segmentation masks based on learned spatial relationships. The system autonomously processes medical images without requiring manual intervention for each segmentation task, dramatically improving work efficiency while maintaining precision through the model's trained accuracy on anatomical feature recognition.
4Measurement precision
If feature points are used for target region specification, then segmentation precision is improved, but system complexity increases
Solution Approach 1:
The patent segments the complex image segmentation task into distinct modular components: (1) extraction of anatomical landmark positions, (2) calculation of spatial relationships between landmarks, (3) prediction of target region boundaries based on learned patterns, and (4) generation of segmentation masks. This segmentation of the overall process into independent, manageable modules reduces system complexity while maintaining high segmentation precision through specialized processing at each stage.
Data Source
AI summary
A medical image processing apparatus according to an embodiment includes processing circuitry. The processing circuitry is configured to generate, from a medical image, a plurality of feature target regions for specifying a target of image segmentation. The processing circuitry is configured to specify a target region indicating the region in which the target is present, on a basis of the plurality of feature target regions. The processing circuitry is configured to perform, in the target region, the image segmentation on the target.