Marker Tracking in Fluoroscopic Images Using Pre-rendered Templates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for tracking markers in fluoroscopic images during radiation treatment face challenges with non-spherical markers, particularly rod- or wedge-shaped markers, due to their changing shapes in three-dimensional space, leading to difficulties in real-time tracking and potential erroneous detection of non-marker images.
Innovation Solution
A medical image processing device and program that learns features common to marker images from training images, using a combination of simulation and clinical images to track markers in fluoroscopic images, allowing for accurate detection of marker positions and reducing the risk of erroneous tracking by using feature extraction parameters to predict and correct marker positions in real-time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If template matching is used to detect marker position in fluoroscopic images, then marker detection can be performed automatically, but detection accuracy deteriorates when markers have non-spherical shapes that change orientation in 3D space
Solution Approach 1:
The system performs preliminary action by generating multiple pre-rendered template images of the marker from different 3D orientations and positions before the actual detection process. These templates are stored and selected based on the current fluoroscopic image characteristics, enabling accurate template matching even when the marker orientation changes during treatment.
Solution Approach 2:
The system applies dynamics by making the template selection adaptive and dynamic rather than static. The template choice changes based on the detected marker orientation and position in real-time, allowing the system to handle non-spherical markers that rotate and change appearance during the radiation treatment process.
2Measurement precision
If multiple templates are used to improve detection accuracy for non-spherical markers, then detection precision improves, but calculation time and system complexity increase
Solution Approach 1:
Multiple templates are pre-rendered and stored in advance from different 3D orientations and positions, so they are ready for immediate use during detection without requiring real-time generation. This eliminates the need for expensive real-time 3D rendering and reduces calculation time during actual marker tracking.
Solution Approach 2:
The system uses a limited set of representative templates covering the most probable marker orientations and positions rather than exhaustively searching all possible orientations. This partial action approach achieves sufficient detection accuracy while significantly reducing the computational burden compared to complete template matching.
3Measurement precision
If spherical markers are used for easy template matching, then detection accuracy improves, but marker stability deteriorates as spherical markers may move in patient's daily life
Solution Approach 1:
The system enables the use of asymmetric (non-spherical) markers such as rod-shaped or wedge-shaped markers that are more stable in the patient's body. By generating and matching 3D templates that account for different orientations, the system achieves accurate detection of these asymmetric markers without requiring them to be spherical, thus maintaining both detection accuracy and marker stability.
4Stability of the object's composition
If rod-shaped or wedge-shaped markers are used to improve stability, then marker stability improves, but detection difficulty increases due to changing shapes in fluoroscopic images
Solution Approach 1:
The system performs preliminary action by pre-generating 3D rendered templates of the specific rod-shaped or wedge-shaped marker from multiple orientations and positions. These templates are stored and used for matching against fluoroscopic images, eliminating the need for complex real-time 3D modeling and rendering during detection.
Solution Approach 2:
The system applies dynamics by adapting the template selection to match the current marker orientation and position detected in the fluoroscopic image. This dynamic template matching approach handles the changing appearance of rod-shaped or wedge-shaped markers as they rotate and move within the patient's body during treatment.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables efficient and accurate tracking of markers, improving real-time performance and reducing errors in marker detection, thereby ensuring precise radiation targeting aligned with the planned treatment position, even for markers with varying orientations.
Implementation Method 1
a fluoroscopic image PI of a patient P which is captured during treatment
Implementation Method 2
irradiating a lesion within a patient's body with radiation to destroy the lesion
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
A medical image processing device according to an embodiment includes a first image acquirer and a tracker. The first image acquirer is configured to acquire a fluoroscopic image of a patient as a first image. The tracker is configured to track an object photographed in the first image on the basis of a first feature common to object images that are a plurality of images of the object obtained by observing the object placed within a body of the patient in a plurality of directions.