AI-Guided Median Sagittal Extraction From 3D Volume Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for extracting standard sections from three-dimensional volume data in medical imaging, such as the median sagittal section, are inefficient and struggle with sections having minimal feature differences, leading to prolonged processing times or inaccuracies.
Innovation Solution
A medical processing apparatus and method that utilizes a trained model to detect specific anatomical features, such as crown-rump length or anatomical structures, to define a reference line, and then extracts a first section intersecting this line, followed by determining the posture of the subject to accurately identify the median sagittal section.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If fully searching for the standard section from all azimuthal sections is used, then the standard section can be extracted automatically, but enormous time is required for the processing
Solution Approach 1:
The patent segments the search process by first identifying candidate sections based on specific anatomical features (such as the corpus callosum in brain imaging), then performing detailed analysis only on these candidates rather than searching all azimuthal sections. This divides the automated extraction process into preliminary candidate selection and final verification stages, reducing overall processing time while maintaining automation.
Solution Approach 2:
The patent performs preliminary identification of sections containing key anatomical structures before final standard section extraction. By pre-processing the volume data to locate characteristic features and narrow down candidate sections, the system prepares the data in advance for faster automated extraction, eliminating the need to search all possible azimuthal sections.
2Loss of time
If searching for the standard section using artificial intelligence is used, then time and effort of the user are reduced, but it is difficult to extract the standard section when there is little difference in features between the section to be extracted and neighboring sections
Solution Approach 1:
The patent applies local quality by enhancing the detection of specific anatomical features within the volume data. Instead of relying on general AI patterns, the system identifies and emphasizes local structural characteristics (such as the unique geometry of the corpus callosum or specific fetal landmarks) that distinguish the target section from neighboring sections, improving extraction reliability even when overall features are similar.
Solution Approach 2:
The patent changes parameters by using multiple detection criteria and thresholds for identifying standard sections. The system adjusts detection sensitivity and uses composite scoring based on multiple anatomical landmarks and structural relationships, enabling reliable differentiation of target sections from similar neighboring sections through multi-parameter evaluation rather than single-feature detection.
3Measurement precision
If user manually searches for the standard section from three-dimensional volume data, then accurate extraction can be achieved, but time and effort of the user are increased
Solution Approach 1:
The patent implements self-service by enabling the system to automatically perform section extraction using AI-assisted detection of anatomical landmarks and structural features. The volume data processing system independently identifies candidate sections, evaluates them against predefined criteria, and extracts the standard section without requiring manual user intervention, thereby maintaining accuracy while eliminating time and effort expenditure.
Data Source
AI summary
A medical processing apparatus according to the present embodiment comprises a processing circuitry configured to acquire three-dimensional volume data of a subject, extract a first section intersecting a reference line of the subject included in the three-dimensional volume data, extract a second section intersecting the first section from the three-dimensional volume data based on a posture of the subject included in the first section, and output information based on the second section.