GB2463450A - Region of Interest Tuning for Dynamic Imaging - Google Patents
Region of Interest Tuning for Dynamic Imaging Download PDFInfo
- Publication number
- GB2463450A GB2463450A GB0816192A GB0816192A GB2463450A GB 2463450 A GB2463450 A GB 2463450A GB 0816192 A GB0816192 A GB 0816192A GB 0816192 A GB0816192 A GB 0816192A GB 2463450 A GB2463450 A GB 2463450A
- Authority
- GB
- United Kingdom
- Prior art keywords
- interest
- region
- initial region
- model
- rol
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
- 238000003384 imaging method Methods 0.000 title description 12
- 238000000034 method Methods 0.000 claims abstract description 13
- 230000001419 dependent effect Effects 0.000 claims 1
- 238000011156 evaluation Methods 0.000 claims 1
- 101000657326 Homo sapiens Protein TANC2 Proteins 0.000 description 6
- 102100034784 Protein TANC2 Human genes 0.000 description 6
- 239000000700 radioactive tracer Substances 0.000 description 4
- 238000002591 computed tomography Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000002599 functional magnetic resonance imaging Methods 0.000 description 3
- 238000002595 magnetic resonance imaging Methods 0.000 description 3
- WQZGKKKJIJFFOK-GASJEMHNSA-N Glucose Natural products OC[C@H]1OC(O)[C@H](O)[C@@H](O)[C@@H]1O WQZGKKKJIJFFOK-GASJEMHNSA-N 0.000 description 2
- 206010028980 Neoplasm Diseases 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 239000008103 glucose Substances 0.000 description 2
- 238000012831 peritoneal equilibrium test Methods 0.000 description 2
- 238000012636 positron electron tomography Methods 0.000 description 2
- 238000012877 positron emission topography Methods 0.000 description 2
- 238000002603 single-photon emission computed tomography Methods 0.000 description 2
- 230000031018 biological processes and functions Effects 0.000 description 1
- 210000000988 bone and bone Anatomy 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 230000010339 dilation Effects 0.000 description 1
- 230000003628 erosive effect Effects 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 238000011850 initial investigation Methods 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 230000035790 physiological processes and functions Effects 0.000 description 1
- 238000002600 positron emission tomography Methods 0.000 description 1
- 102000005962 receptors Human genes 0.000 description 1
- 108020003175 receptors Proteins 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 208000024891 symptom Diseases 0.000 description 1
- 210000001519 tissue Anatomy 0.000 description 1
- 238000011179 visual inspection Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G06T7/0081—
-
- G06T7/2006—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/215—Motion-based segmentation
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Medical Informatics (AREA)
- Quality & Reliability (AREA)
- Radiology & Medical Imaging (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
A method of selecting a region of interest in image data from an image sequence of a subject is described. An initial region of interest in the image data is identified. A model for an expected progression in the image sequence is determined, and the actual progression in the image sequence for the initial region of interest is measured. The measured and expected progressions are compared, and the initial region of interest is evaluated based on the result of the comparison. The initial region may be modified according to the result. In particular, where an error between the measured and expected progressions is determined according to the model, a region of interest for which the error is lower than that for the initial region of interest is selected.
Description
Region of Interest Tuning for Dynamic Imaging 1.1 Definitions, Acronyms, and Abbreviations CT Computed Tomography fMRI Functional MRI tRW Inveon Research Workplace MRI Magnetic Resonance Imaging PET Positron Emission Tomography ROt Region of Interest SPECT Single Photo Emission Computed Tomography TAC Time Activity Curve 1.2 References c'J (\j 2 WHAT IS THE TECHNICAL PROBLEM SOLVED BY THIS INVENTION? 0 Two distinct branches of medical imaging are structural imaging (e.g., a typical CT or MRI scan) and functional imaging (e.g., a PET, SPECT or fMRI scan). Whereas structural imaging can show the anatomical boundaries and structure of different tissues in the subject (e.g., position of bones or presence of unexpected material such as a tumour), functional imaging can provide information on the physiological function of the subject's body, for example showing areas of the body that are using unexpectedly large amounts of glucose (often a symptom of a tumour).
When using functional imaging, tracers are often introduced into the body that will highlight certain biological processes (different tracers can be used to highlight different processes, such as use of glucose or receptor binding in the brain). Imaging can either be performed to generate a single image, showing average tracer distribution over a fixed period, called static imaging, or used to generate a number of images forming a series of snapshots showing the tracer distribution Over time, called dynamic imaging.
Dynamic imaging can provide a fuller picture of the way in which the tracer is taken up, and in particular it is possible to plot the average uptake within a particular region of the image against time, to create a time-activity curve (TAC). The shape of this curve then contains information about how the tracer was taken up in that region during the course of the scan.
Apart from merely looking at the shape of the TAC, it is also possible to fit a model to TAC to summarise the uptake pattern in a handful of biologically meaningful parameters. Such parameters can then be compared between scans allowing sophisticated comparison of scans (e.g., of the same patient over time, or between patients), as well as allowing a better understanding of the underlying physical processes involved. Such analysis is known as kinetic modelling.
Kinetic-model analysis requires the delineation of one or more regions of interest (ROls) within a dynamic imaging study. The time activity curves (TACs) recorded within these ROls are used to fit the kinetic model from which the parameters of interest (e.g., Ki, k2 k3 etc) are derived. The goodness of this fit, appropriateness of the model, and suitability of the ROls drawn can be determined by, for example, assessment of the residuals of the fit (chi-squared) and the standard errors corresponding to each kinetic parameter.
The problem addressed by this invention is the delineation of the ROls. Currently ROls are manually defined based on a visual inspection of the image, perhaps making use of techniques such as thresholding of image voxels. However, the user must rely on experience to create their ROl delineation, and if they are unhappy with the accuracy of the kinetic modelUng results must decide manually how to adjust the regions of interest to improve the results. Our invention details a method for automatically O tuning (or optimising) this delineation to enable the best-possible fit of the model by minimising the standard errors of the kinetic parameters, or some other metric of goodness of fit.
ci 3 How HAS THIS PROBLEM BEEN ADDRESSED UNTIL Now? O Currently, the process of delineating the 3D ROls necessary for compartment modelling requires a significant amount of manual input. For example, the options available in the Inveon Research Workplace (IRW) allow a user to adjust the position and size of a series of 3D geometric shapes (e.g., sphere, cube etc.) within the image volume, with the voxels contained within this shape defining the ROl. This set of voxeis may then be refined in a semi-automated manner by thresholding, i.e., removing all voxels that lie outside manually-specified intensity limits, or seed-point thresholding, which is thresholding combined with the restriction that the resulting ROl must consist of only a single component connected to the seedpoint.
In addition to requiring significant manual input, the existing methods for ROl definition allow a user to refine their ROl based only on how well it fits' the underlying intensity image. The impact of any adjustments on the quality of the derived kinetic model is not assessed.
4 How DOES THE PROPOSED INVENTION SOLVE THE PROBLEM? The proposed invention addresses the problem of ROl delineation by defining an algorithm that enables fully-automated tuning of an ROl based on an initial seed-point. The algorithm achieves this by selecting the intensity threshold used to define the ROI which minimises the residual error in the fitted compartment model. A more
detailed description is given below.
4.1 PJgorithm description
1. Seed point definition: a. A single voxel is selected by the user to act as a seedpoint for ROl optimisation.
2. Initialisation: a. The intensity of the seedpoint voxel is selected as the initial threshold, with all connected voxels having an intensity equal-to or greater-than this threshold delineating the ROl.
b. The kinetic model is then fitted to the data based on the defined ROl, and standard errors of the resultant kinetic parameters calcUlated.
3. Optimisation: a. The threshold (as defined in 2a) is reduced by an amount sufficient to increase the number of voxels defining the ROl by a small number (e.g., 1).
b. The kinetic model is refit to this new ROL and the errors recalculated.
c. Steps 3a and 3b are repeated until an appropriate stopping criteria is met (e.g., significant increase in error).
O 4.2 Prototype investigation The algorithm described in Section 4.1 has been implemented and initial results are given below. Figure 4.1 illustrates the relationship between threshold and the error in model fit for Case 1. In these initial investigations, error was calculated as the sum of standard errors as percentages of the derived kinetic parameter value (e.g., Ki, k2 C\J and k3); however, alternative methods for computing a combined error may be more O appropriate. The minimum error corresponds to a threshold of 1.14x104. The ROl corresponding to this threshold is shown in Figure 4.2. The corresponding figures for case 2 are shown in Figure 4.3 and Figure 4.4 respectively (with a threshold of 3.56x104 producing the minimum error).
Figure 4.1. Graph illustrating relationship between the threshold used to define the ROl.
and the sum of the standard errors (as percentages of derived kinetic parameter value) for case 1. The minimum error corresponds to a threshold of 1.14 x 1O.
Figure 4.2. Screenshot displaying ROl corresponding to optimal threshold for case 1.
Figure 4.3. Graph illustrating relationship between the threshold used to define the ROl and the sum of the standard errors (as percentages of derived kinetic parameter value) for case 2. The minimum error corresponds to a threshold of 3.56 x 1O.
Figure 4.4. Screenshot displaying ROI corresponding to optimal threshold for case 2.
WHAT ARE THE ALTERNATIVE WAYS IN WHICH THE INVENTION MAY BE
REALISED? 1. ROI adjustment: As opposed to defining and adjusting a ROl based solely on the intensity threshold, the algorithm could be extended to include other measures such as a smoothness constraint, or any other function of the image I ROl (for example standard morphological operations such as erosion and dilation).
2. Optimisation: Any optimisation algorithm may be used to search the parameter space in order to identify the optimal solution (e.g., best threshold in implementation described in Section 4). This could range from an exhaustive search, or in situations where an exhaustive search is impractical due to the number of parameters being optimised, a more intelligent algorithm (e.g., genetic algorithm).
3. Model: The method could be extended from traditional compartment models, to any kinetic modelling approach in which some error of the fit is calculated (e.g. Patlak).
4. Combination of errors from different parameters: although our examples have used the s-n f errors (expressed as a percentage of the derived parameter), Q other techniques may be more appropriate depending on the relative importance of the different parameters. In such cases, a weighted average of the errors could be used, for example.
C'J 5. Metric of goodness of fit: although our example has used the standard error of O the kinetic parameters, any appropriate alternative metric may be used, e.g., chisquared (residuals of fit). 0.
Figure 4.5: Fig 4.2 redrawn with ROt in out line Figure 4.6: Figure 4.4 redrawn with ROl(s) in out line
Claims (3)
- SCLAIMS1. A method of selecting a region of interest in image data from an image sequence of a subject, comprising: identifying an initial region of interest in the image data; determining an appropriate model for an expected progression in the image sequence; measuring a progression in the image sequence for the initial region of interest; comparing the measured progression for the initial region of interest with the expected progression according to the model; and evaluating the initial region of interest according to a result of the comparison.
- 2. A method according to Claim 1, further comprising modifying the initial region of interest according to a result of the evaluation.
- 3. A method according to Claim 1 or Claim 2, further comprising: setting the region of interest as being dependent on a particular variable; determining, for values of the variable, an error between the measured progression in the image sequence for the given region of interest and the expected progression according to the model; and selecting a region of interest for which the error is lower than that for the initial region of interest.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
GB0816192A GB2463450A (en) | 2008-09-05 | 2008-09-05 | Region of Interest Tuning for Dynamic Imaging |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
GB0816192A GB2463450A (en) | 2008-09-05 | 2008-09-05 | Region of Interest Tuning for Dynamic Imaging |
Publications (2)
Publication Number | Publication Date |
---|---|
GB0816192D0 GB0816192D0 (en) | 2008-10-15 |
GB2463450A true GB2463450A (en) | 2010-03-17 |
Family
ID=39888825
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
GB0816192A Withdrawn GB2463450A (en) | 2008-09-05 | 2008-09-05 | Region of Interest Tuning for Dynamic Imaging |
Country Status (1)
Country | Link |
---|---|
GB (1) | GB2463450A (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO1996014795A1 (en) * | 1994-11-11 | 1996-05-23 | Board Of Regents Of The University Of Washington | Cortical tissue imaging methods and device |
US20040167395A1 (en) * | 2003-01-15 | 2004-08-26 | Mirada Solutions Limited, British Body Corporate | Dynamic medical imaging |
US20050096543A1 (en) * | 2003-11-03 | 2005-05-05 | Jackson John I. | Motion tracking for medical imaging |
US20070016016A1 (en) * | 2005-05-31 | 2007-01-18 | Gabriel Haras | Interactive user assistant for imaging processes |
-
2008
- 2008-09-05 GB GB0816192A patent/GB2463450A/en not_active Withdrawn
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO1996014795A1 (en) * | 1994-11-11 | 1996-05-23 | Board Of Regents Of The University Of Washington | Cortical tissue imaging methods and device |
US20040167395A1 (en) * | 2003-01-15 | 2004-08-26 | Mirada Solutions Limited, British Body Corporate | Dynamic medical imaging |
US20050096543A1 (en) * | 2003-11-03 | 2005-05-05 | Jackson John I. | Motion tracking for medical imaging |
US20070016016A1 (en) * | 2005-05-31 | 2007-01-18 | Gabriel Haras | Interactive user assistant for imaging processes |
Also Published As
Publication number | Publication date |
---|---|
GB0816192D0 (en) | 2008-10-15 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Chen et al. | Automatic segmentation of individual tooth in dental CBCT images from tooth surface map by a multi-task FCN | |
RU2571523C2 (en) | Probabilistic refinement of model-based segmentation | |
Pedrosa et al. | LNDb: a lung nodule database on computed tomography | |
EP3035287B1 (en) | Image processing apparatus, and image processing method | |
Egger et al. | Pituitary adenoma volumetry with 3D Slicer | |
Zaidi et al. | Comparative methods for PET image segmentation in pharyngolaryngeal squamous cell carcinoma | |
US9646229B2 (en) | Method and system for bone segmentation and landmark detection for joint replacement surgery | |
WO2020028352A1 (en) | Methods and systems for segmenting organs in images using a cnn-based correction network | |
US9547894B2 (en) | Apparatus for, and method of, processing volumetric medical image data | |
AU2012312482B2 (en) | Methods of spatial normalization of positron emission tomography images | |
US20140301624A1 (en) | Method for interactive threshold segmentation of medical images | |
CN105912874A (en) | Liver three-dimensional database system constructed on the basis of DICOM (Digital Imaging and Communications in Medicine) medical image | |
Fajar et al. | Reconstructing and resizing 3D images from DICOM files | |
CN107004305A (en) | Medical image editor | |
Buyssens et al. | Eikonal based region growing for superpixels generation: Application to semi-supervised real time organ segmentation in CT images | |
CN106462974A (en) | Optimization of parameters for segmenting an image | |
CN113196340A (en) | Artificial Intelligence (AI) -based Standardized Uptake Value (SUV) correction and variance assessment for Positron Emission Tomography (PET) | |
Bakas et al. | Segmentation of gliomas in multimodal magnetic resonance imaging volumes based on a hybrid generative-discriminative framework | |
CN104146766B (en) | Scanning means, scan method and medical image equipment | |
Wang et al. | Automatic segmentation of coronary arteries in CT imaging in the presence of kissing vessel artifacts | |
Cavalcanti et al. | Unmixing dynamic PET images with variable specific binding kinetics | |
CN107920793A (en) | Image reconstruction process method, image reconstruction process program and the laminagraph device for being equipped with the image reconstruction process program | |
Mohammed et al. | Liver segmentation: A survey of the state-of-the-art | |
CN104915989A (en) | CT image-based blood vessel three-dimensional segmentation method | |
Contijoch et al. | Shape analysis of simulated breast anatomical structures |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
WAP | Application withdrawn, taken to be withdrawn or refused ** after publication under section 16(1) |