CTP-based VOF detection method and system

Abstract

The present invention relates to a computed tomography perfusion (CTP)-based venous output function (VOF) detection method and system, and the CTP-based VOF detection method performed by the CTP-based VOF detection system, according to the present invention, comprises: removing bias noise from a contrast agent concentration graph in a brain region of a CTP image; excluding a graph that has a peak earlier than the peak time of an arterial input function (AIF) and aligning, to the same position as the peak of the AIF, peak values of remaining graphs; calculating a Pearson correlation coefficient with the AIF so as to select a graph having a Pearson correlation coefficient greater than or equal to a predefined threshold; generating a mask on the basis of a peak value range of a VOF candidate group selected from the selected graphs, so as to extract a final VOF graph; generating a maximum intensity projection (MIP) from the CTP image and extracting a blood vessel image; and checking a veinous positions through an operation between the mask and the VOF candidate group so as to select, as a final VOF, the VOF candidate closest to a center relative to the midline.

Description

CTP-based VOF detection method and system

[0001] The present invention relates to a method and system for detecting Venous Output Function (VOF) based on Computed Tomography Perfusion (CTP).

[0002] Computed Tomography Perfusion (CTP) imaging technology for evaluating cerebral hemodynamics tracks the series of processes in which a contrast agent enters the cerebral vascular system through arteries and is discharged through veins over time, thereby calculating various perfusion parameters such as cerebral blood flow (CBF), cerebral blood flow velocity (CBV), and mean transit time (MTT). To accurately calculate these perfusion parameters, a reference curve is required that indicates how the contrast agent changes within the vascular system.

[0003] One of these reference curves, the Venous Output Function (VOF), is a signal that reflects the process of contrast agent being expelled through the veins after passing through brain tissue, and plays a key role in the normalization process of perfusion analysis. When the VOF is calculated accurately, the scale and dynamic characteristics of changes in contrast agent concentration are stably aligned during the calculation of perfusion indices, which determines the reliability of the overall CTP analysis.

[0004] Generally, CTP analysis utilizes the VOF in conjunction with the Arterial Input Function (AIF), which represents the characteristics of contrast agent arrival in arteries. While the AIF reflects the initial dynamic changes as the contrast agent enters the brain via the arteries, the VOF represents the dynamic characteristics of the later phase as the contrast agent is expelled through the veins. Using both AIF and VOF together allows for the correction of time delay and dispersion phenomena occurring between arteries and veins. Such correction is essential for obtaining accurate hemodynamic parameters, enabling more accurate early diagnosis of diseases such as cerebral infarction and assessment of lesion extent.

[0005] However, it is difficult to automatically and accurately detect VOF due to various factors, such as the anatomical diversity of venous structures, image noise, patient movement, and individual differences in contrast agent delivery patterns. In particular, existing algorithms often target a single vein or rely on simple pixel-based criteria, leading to a problem where they fail to achieve a level of accuracy suitable for stable clinical use. Since inaccurate VOF detection can distort the results of the overall perfusion analysis, an accurate VOF detection algorithm is essential for the reliability and accuracy of CTP analysis.

[0006] Accordingly, there is a need to develop a new VOF detection technology to reliably extract contrast agent efflux signals from the cerebral venous system and improve the accuracy of AIF-VOF-based perfusion parameter calculations.

[0007] The present invention was devised to solve the aforementioned problems, and aims to establish a more sophisticated and reliable candidate screening system by combining multiple criteria comparisons (peak value and time to reach peak, etc.) with an Arterial Input Function (AIF) and Full Width at Half Maximum (FWHM)-based filtering, and to secure a detection mechanism that can operate flexibly under various patient conditions by applying dynamic Pearson correlation coefficient-based filtering that reflects temporal change patterns.

[0008] In addition, the present invention ensures biological validity by ensuring that the selected candidate anatomically matches the actual venous pathway through spatial location-based correction and vascular structure verification, implements an algorithm that operates stably even when there are variations in image quality or patient movement by adding an alternative evaluation method using Faux CTA and midline-centered correction, and applies adaptive window settings to reflect the diversity of vascular structures for each patient and increase the accuracy of venous structure extraction.

[0009] Through this, the present invention aims to effectively solve problems such as instability, low accuracy, and lack of clinical validity of existing VOF detection algorithms through this multi-stage refinement process, and to provide a high-precision VOF detection technology that can be stably applied in various clinical environments.

[0010] A CTP-based VOF detection method performed by a CTP-based VOF detection system according to an embodiment of the present invention for solving the aforementioned problem comprises: a first step of removing bias noise on a contrast agent concentration graph in the brain region of a CTP image; a second step of excluding graphs having a peak faster than the peak time of an AIF (Arterial Input Function) from among the graphs from which the bias noise has been removed; a third step of aligning the peak values ​​of the excluded graphs with the same position as the peak of the AIF and calculating the Pearson correlation coefficient with the AIF; a fourth step of selecting graphs in which the Pearson correlation coefficient is greater than or equal to a predefined threshold; a fifth step of selecting a group of VOF candidates from the selected graphs; a sixth step of generating a mask based on the peak value range of the selected VOF candidates; and a seventh step of extracting a final VOF graph using the mask and the average value of the farthest cluster. The method comprises: a step 8 of generating a Maximum Intensity Projection (MIP) from the CTP image and extracting a blood vessel image by setting a window to clearly reveal the blood vessel structure; a step 9 of identifying possible vein locations by performing an AND operation between the generated mask and the VOF candidate group; and a step 10 of selecting the VOF candidate closest to the center based on the Midline from the possible vein locations and selecting it as the final VOF.

[0011] According to another embodiment of the present invention, the third step performs a shift operation to align the Peak value of the selected graph with the Peak of the AIF to the same position, and calculates the Pearson correlation coefficient with the AIF.

[0012] According to another embodiment of the present invention, the fifth step calculates the Full Width at Half Maximum (FWHM) of each selected graph and excludes graphs smaller than the FWHM of the AIF to select VOF candidates.

[0013] According to another embodiment of the present invention, the seventh step removes VOF candidates in the center and upper regions of the image using the mask, and extracts a final VOF graph using the average value of the farthest cluster.

[0014] A Computed Tomography Perfusion (CTP)-based Venous Output Function (VOF) detection system according to an embodiment of the present invention comprises: a noise removal unit for removing bias noise on a contrast agent concentration graph in the brain region of a Computed Tomography Perfusion (CTP) image; a graph filtering unit for excluding graphs having a peak faster than the peak time of an Arterial Input Function (AIF) among the graphs from which the bias noise has been removed; a correlation coefficient calculation unit for aligning the peak values ​​of the excluded graphs with the same position as the peak of the AIF and calculating the Pearson correlation coefficient with the AIF; a graph selection unit for selecting graphs in which the Pearson correlation coefficient is greater than or equal to a predefined threshold; a VOF candidate selection unit for selecting a group of Venous Output Function (VOF) candidates from the selected graphs; a mask generation unit for generating a mask based on the range of peak values ​​of the selected VOF candidates; and a VOF graph extraction unit for extracting a final VOF graph using the mask and the average value of the farthest cluster. The system comprises: a blood vessel image extraction unit that generates a Maximum Intensity Projection (MIP) from the CTP image and extracts a blood vessel image by setting a window to clearly reveal the blood vessel structure; a vein location verification unit that identifies possible vein locations by performing an AND operation between the generated mask and a group of VOF candidates; and a VOF selection unit that selects the VOF candidate closest to the center based on the Midline from the possible vein locations and selects it as the final VOF.

[0015] According to another embodiment of the present invention, the correlation coefficient calculation unit performs a shift operation to align the Peak value of the selected graph with the Peak of the AIF to the same position, and calculates the Pearson correlation coefficient with the AIF.

[0016] According to another embodiment of the present invention, the VOF candidate selection unit calculates the Full Width at Half Maximum (FWHM) of each selected graph and excludes graphs smaller than the FWHM of the AIF to select VOF candidates.

[0017] According to another embodiment of the present invention, the VOF graph extraction unit removes VOF candidates in the center and upper regions of the image using the mask, and extracts a final VOF graph using the average value of the farthest cluster.

[0018] According to the present invention, distortion caused by image noise and baseline drift can be effectively suppressed by precisely adjusting the baseline of the perfusion signal through bias noise removal. This stabilizes the initial reference value of the time-concentration curve and fundamentally improves the accuracy of the feature analysis and candidate selection processes in subsequent steps.

[0019] In addition, by performing multiple criteria comparisons with AIF and simultaneously considering peak values ​​and time-to-peak, the limitations of existing methods that relied on a single criterion can be overcome, and only more reliable vein candidate signals can be selected. Furthermore, by combining this with FWHM (Full Width at Half Maximum)-based filtering, the candidate pool can be selected more precisely by reflecting the gradual concentration change patterns exhibited by actual vein signals.

[0020] Furthermore, the present invention evaluates the similarity of temporal change patterns by applying dynamic Pearson correlation, thereby ensuring flexibility to adapt to various patient conditions or changes in contrast agent injection patterns. This overcomes the limitations of a simple static characteristic value matching method and enables more stable candidate selection that reflects actual hemodynamic signal characteristics.

[0021] In addition, by combining spatial location-based correction and vascular structure verification processes, the possibility of false detection is significantly reduced by confirming whether the selected candidate location corresponds to an anatomically valid venous structure. This ensures that the VOF location selected by the algorithm matches the actual biological venous pathway, thereby increasing the reliability of the result interpretation.

[0022] The present invention also includes alternative analysis utilizing Faux CTA, thereby providing a stable auxiliary judgment criterion even when the quality of the original image is degraded or venous signals are not sufficiently secured under specific conditions. By adding midline-based position correction, alignment errors caused by patient movement, image tilting, etc., are minimized, further expanding the scope of application of the algorithm.

[0023] In addition, by dynamically adjusting the area likely to contain vein structures according to the situation through the adaptive window setting, the precision of the blood vessel extraction process can be improved and false detections caused by unnecessary area analysis can be reduced.

[0024] This multi-stage refinement process ensures that individual elements work complementarily to provide high accuracy and reproducibility, as well as secure the validity and reliability required in a clinical environment. As a result, the VOF detection algorithm according to the present invention operates stably under various image quality and patient conditions and has the effect of significantly improving the precision of CTP-based perfusion analysis.

[0025] FIG. 1 is a flowchart illustrating a method for detecting Computed Tomography Perfusion (CTP) Venous Output Function (VOF) according to an embodiment of the present invention.

[0026] FIG. 2 is a drawing for explaining a mask generated according to an embodiment of the present invention.

[0027] FIG. 3 is a drawing showing the VOF closest to the center based on the Midline generated according to an embodiment of the present invention.

[0028] FIG. 4 is a configuration diagram of a CTP-based VOF detection system according to an embodiment of the present invention.

[0029] The present invention is capable of various modifications and may have various embodiments, and specific embodiments are illustrated in the drawings and described in detail in the description of the invention. However, this is not intended to limit the present invention to specific embodiments, and it should be understood that it includes all modifications, equivalents, and substitutions that fall within the spirit and scope of the invention.

[0030] However, in describing the embodiments, if it is determined that a detailed description of related known functions or configurations could unnecessarily obscure the essence of the invention, such detailed description is omitted. Furthermore, the sizes of each component in the drawings may be exaggerated for illustrative purposes and do not represent the actual sizes applied.

[0031] Furthermore, throughout the specification, when a component is referred to as being "connected" or "joined" with another component, it should be understood that the component may be directly connected or joined to the other component, but unless specifically stated otherwise, it may also be connected or joined through an intermediate component. Additionally, throughout the specification, when a part is described as "including" a component, unless specifically stated otherwise, this means that it may include additional components rather than excluding other components.

[0032] FIG. 1 is a flowchart illustrating a method for detecting a Computed Tomography Perfusion (CTP) Venous Output Function (VOF) according to an embodiment of the present invention. FIG. 2 is a diagram illustrating a mask generated according to an embodiment of the present invention, where FIG. 2(a) is a graph showing 100 selected VOF candidates for plotting a graph, and FIG. 2(b) is a diagram showing a mask generated based on the peak value range of the VOF candidates. FIG. 3 is a diagram showing the VOF closest to the center based on the midline generated according to an embodiment of the present invention.

[0033] A CTP-based VOF detection method according to one embodiment of the present invention includes a series of processes for finally detecting VOF by analyzing the time-concentration graph of all pixels extracted from a CTP image.

[0034] In addition, the CTP-based VOF detection method according to one embodiment of the present invention is performed by a CTP-based VOF detection system, wherein the CTP-based VOF detection system according to one embodiment of the present invention may be composed of a computer terminal, a server, or a dedicated device, or the components providing each function may each be composed of a computer terminal, a server, or a dedicated device. Alternatively, the CTP-based VOF detection system according to one embodiment of the present invention may have each component providing each function composed of hardware or software, and more specifically, each component may be composed of modules classified by operation functions performed by a processor.

[0035] From now on, a method for detecting CTP (Computed Tomography Perfusion) VOF (Venous Output Function) according to an embodiment of the present invention will be described with reference to FIGS. 1 to 3.

[0036] First, for each graph acquired in the brain region of the CTP (Computed Tomography Perfusion) image, bias noise is removed to resolve the problem where the baseline prior to contrast agent injection is distorted by noise (S110). By analyzing the initial section of the graph to estimate the bias value and correcting it across the entire graph, all signals start from the same baseline, making quantitative comparison possible in subsequent analysis steps.

[0037] In addition, among the graphs from which bias noise has been removed, signals that appear at a time earlier than the peak of the AIF (Arterial Input Function) are excluded because they cannot be seen as veins (S120). By utilizing the physiological characteristic that vein signals reach their peak later than arterial signals, a significant portion of non-venous signals are removed at this stage.

[0038] Afterward, the graphs remaining after exclusion undergo a shifting process to align their peak positions for comparison with AIF, and the Pearson correlation coefficient with AIF is calculated (S130). At this time, a shift operation is performed to align the peak values ​​of the selected graphs with the peaks of AIF, and the Pearson correlation coefficient between each graph and AIF is calculated in the aligned state. Only graphs that satisfy a set threshold are selected as targets for analysis in the next step (S140). This is a measure to maintain only signals whose temporal change patterns are similar to AIF at a certain level or higher.

[0039] Afterwards, a group of candidates for the Venous Output Function (VOF) is selected from the selected graphs (S150). At this time, for the graphs that satisfy the correlation coefficient criteria, the Full Width at Half Maximum (FWHM) is calculated once again to reflect the unique signal width of the vein. Graphs with an FWHM that is excessively narrower than the AIF are removed because they do not match the characteristics of the vein signal, and through this process, a group of morphologically consistent VOF candidates is secured.

[0040] Subsequently, as shown in FIG. 2, a mask is generated using the Peak value distribution of the remaining VOF candidates (S160). The mask generated based on the Peak value range indicates an area in the image space where vein signals are likely to exist, and provides a reference area for subsequent structure-based analysis.

[0041] When a mask is created, the spatial distribution formed by the candidate group is clustered, and the final VOF graph is extracted using the average value of the cluster formed at the farthest location among them (S170). At this time, the VOF candidates in the center and upper regions of the image are removed using the mask, and the final VOF graph is extracted using the average value of the farthest cluster.

[0042] Subsequently, a Maximum Intensity Projection (MIP) is generated from the above CTP image, and a window is set to clearly reveal the vascular structure, thereby extracting a vascular image (S180). At this time, an appropriate window is set to emphasize structures with high contrast agent concentration, and through this process, an image is obtained in which vascular structures, including arteries and veins, are clearly revealed.

[0043] Afterwards, an AND operation is performed between the generated mask and the VOF candidate group to identify possible vein locations (S190).

[0044] Accordingly, as shown in FIG. 3, among several candidates identified as veins, the candidate closest to the center based on the midline of the brain is selected as the final VOF (S200). Since the midline is a reference that reflects the left-right symmetry of the brain structure, the closer the location is to the center, the higher the probability that it corresponds to the main vein (superior sagittal sinus, etc.). Through this, positional errors caused by image tilting or patient movement can be corrected, and the final VOF corresponding to the actual vein structure can be reliably derived.

[0045] Through such procedures, the present invention implements a multi-stage algorithm that includes signal purification, arterial-vein comparison, shape-based filtering, spatial verification, and combined vascular structure analysis, thereby enabling VOF detection that is much more accurate and stable than existing methods.

[0046] FIG. 4 is a configuration diagram of a CTP-based VOF detection system according to an embodiment of the present invention.

[0047] From now on, the configuration of a CTP-based VOF detection system according to an embodiment of the present invention will be described with reference to FIG. 4.

[0048] A CTP-based VOF detection system according to one embodiment of the present invention may be composed of a computer terminal, a server, or a dedicated device, or each component providing each function may be composed of a computer terminal, a server, or a dedicated device. Alternatively, each component providing each function of the CTP-based VOF detection system according to one embodiment of the present invention may be composed of hardware or software.

[0049] More specifically, a CTP-based VOF detection system (100) according to one embodiment of the present invention comprises a noise removal unit (110), a graph filtering unit (120), a correlation coefficient calculation unit (130), a graph selection unit (140), a VOF candidate selection unit (150), a mask generation unit (160), a VOF graph extraction unit (170), a blood vessel image extraction unit (180), a vein location verification unit (190), and a VOF selection unit (195).

[0050] The noise removal unit (110) removes bias noise included in the contrast agent concentration graph extracted from the Brain region of the Computed Tomography Perfusion (CTP) image. The noise removal unit (110) can receive the CTP image from the database (200).

[0051] The graph filtering unit (120) excludes graphs among the remaining graphs after noise removal that have a peak appearing faster than the peak time of the AIF (Arterial Input Function), leaving only valid graphs.

[0052] Next, the correlation coefficient calculation unit (130) shifts the peak positions of the filtered graphs to match the peak positions of the AIF, and then calculates the Pearson correlation coefficient between each graph and the AIF.

[0053] The graph selection unit (140) selects only graphs in which the calculated correlation coefficient is greater than or equal to a predefined threshold.

[0054] The VOF candidate selection unit (150) calculates the FWHM (Full Width at Half Maximum) of each selected graph and forms a VOF (Venous Output Function) candidate group by excluding graphs whose value is smaller than the FWHM of the AIF.

[0055] The mask generation unit (160) generates a mask based on the Peak value range of the selected VOF candidate group.

[0056] The VOF graph extraction unit (170) derives a final VOF graph using the average value of the cluster formed at the furthest position from the generated mask. Specifically, first, a mask is applied to remove VOF candidates located in the center and upper regions of the image, and then the average graph of the furthest cluster is extracted as the final VOF.

[0057] The blood vessel image extraction unit (180) generates a Maximum Intensity Projection (MIP) from the CTP image and obtains a blood vessel image with the blood vessel structure emphasized through appropriate window settings.

[0058] The vein location verification unit (190) performs an AND operation on the generated mask and the VOF candidate group to identify an area where the vein is likely to be located.

[0059] Finally, the VOF selection unit (195) selects the candidate closest to the center based on the Midline among the possible vein locations and confirms it as the final VOF.

[0060] In the detailed description of the present invention as described above, specific embodiments have been described. However, various modifications are possible within the scope of the present invention. The technical concept of the present invention should not be limited to the aforementioned embodiments, but should be defined by the claims as well as equivalents thereof.

Claims

1. A first step of removing bias noise from the contrast agent concentration graph in the brain region of a CTP (Computed Tomography Perfusion) image; A second step of excluding graphs having a peak faster than the peak time of the AIF (Arterial Input Function) from among the graphs from which the above bias noise has been removed; Step 3, aligning the peak values ​​of the excluded remaining graphs with the peaks of the AIF and calculating the Pearson correlation coefficient with the AIF; Step 4, which selects graphs where the Pearson correlation coefficient is above a predefined threshold; Step 5, selecting a group of VOF (Venous Output Function) candidates from the selected graphs; Step 6: Generating a mask based on the peak value range of selected VOF candidates; Step 7, extracting the final VOF graph using the above mask and the average value of the farthest cluster; Step 8, generating a Maximum Intensity Projection (MIP) from the above CTP image and extracting a blood vessel image by setting a window to clearly reveal the blood vessel structure; Step 9, identifying possible vein locations by performing an AND operation between the generated mask and the VOF candidate group; and Step 10: From the above possible venous locations, selecting the VOF candidate closest to the center based on the Midline and selecting it as the final VOF; A CTP-based VOF detection method characterized by including 2. In Claim 1, The above third step is, A CTP-based VOF detection method characterized by performing a shift operation to align the peak value of a selected graph with the peak of the AIF to the same position, and calculating the Pearson correlation coefficient with the AIF.

3. In Claim 1, The above fifth step is, A CTP-based VOF detection method characterized by calculating the Full Width at Half Maximum (FWHM) of each selected graph and excluding graphs smaller than the FWHM of the AIF to select VOF candidates.

4. In Claim 1, The above seventh step is, A CTP-based VOF detection method characterized by removing VOF candidates in the center and upper regions of an image using the above mask, and extracting a final VOF graph using the average value of the farthest cluster.

5. A noise removal unit that removes bias noise on the contrast agent concentration graph in the brain region of a CTP (Computed Tomography Perfusion) image; A graph filtering unit that excludes graphs having a peak faster than the peak time of the AIF (Arterial Input Function) among the graphs from which the above bias noise has been removed; A correlation coefficient calculation unit that aligns the peak values ​​of the remaining graphs with the same position as the AIF peak and calculates the Pearson correlation coefficient with the AIF; A graph selection unit that selects graphs in which the Pearson correlation coefficient is greater than or equal to a predefined threshold; A VOF candidate selection unit that selects a VOF (Venous Output Function) candidate group from selected graphs; A mask generation unit that generates a mask based on the peak value range of selected VOF candidates; A VOF graph extraction unit that extracts a final VOF graph using the above mask and the average value of the farthest cluster; A blood vessel image extraction unit that generates a Maximum Intensity Projection (MIP) from the above CTP image and extracts a blood vessel image by setting a window to clearly reveal the blood vessel structure; A vein location verification unit that performs an AND operation between the generated mask and the VOF candidate group to verify possible vein locations; and A VOF selection unit that selects the VOF candidate closest to the center based on the Midline at the above possible venous location and selects it as the final VOF; A CTP-based VOF detection system characterized by including 6. In Claim 5, The above correlation coefficient calculation unit is, A CTP-based VOF detection system characterized by performing a shift operation to align the peak value of a selected graph with the peak of the AIF to the same position, and calculating the Pearson correlation coefficient with the AIF.

7. In Claim 5, The above VOF candidate selection unit is, A CTP-based VOF detection system characterized by selecting VOF candidates by calculating the Full Width at Half Maximum (FWHM) of each selected graph and excluding graphs smaller than the FWHM of the AIF.

8. In Claim 5, The above VOF graph extraction unit is, A CTP-based VOF detection system characterized by removing VOF candidates in the center and upper regions of an image using the above mask, and extracting a final VOF graph using the average value of the farthest cluster.

Images