Method for characterising collateral circulation

Abstract

The invention relates to a method for generating a desynchronisation map that contains both morphological and physiological quantitative information by implementing the following two steps: detecting the transition time between the parenchymal phase and the cerebral venous phase in the cerebral circulation region; and generating the desynchronisation map. Upon detection of the cerebral venous phase, the image corresponding to the start of the cerebral venous phase is detected, for example using a bidirectional dynamic maximum enhancement algorithm, which measures changes in the entropy of the crossing point over the entire data acquisition cycle. This image is used to determine the desynchronisation map. The desynchronisation map provides the following information: complexity and density of collateral circulation, presence of a collateral venous phase, and quantification of the temporal desynchronisation between cerebral circulation and collateral circulation, measured in seconds.

Description

Description Title of the invention: Method for characterizing collateral circulation technical field

[0001] The present invention relates to a method for processing angiographic images obtained during an acquisition phase in which a contrast agent is injected into the vascular network of an organ experiencing arterial occlusion. The organ may be a human or animal brain undergoing a stroke. The invention can be applied to any type of organ that is the source of cerebral, cardiac, and peripheral vascular diseases. Prior art

[0002] In general, ischemic strokes (IS) are considered life-threatening emergencies. IS are caused by the blockage of a cerebral artery by clots (or fragments of atherosclerotic plaques), resulting in ischemia of the brain tissue.

[0003] The introduction of mechanical thrombectomy (MT) has enabled the effective treatment of ischemic strokes. Mechanical thrombectomy is a procedure performed by interventional neuroradiologists in the arteriography suites of authorized centers, which allows for the mechanical recanalization of the occluded artery through endovascular navigation of dedicated devices (stents, aspiration catheters) and the retrieval of the clot / fragment obstructing the artery.

[0004] In contrast, the clinical efficacy of mechanical thrombectomy decreases over time, as cerebral ischemia is a dynamic event that progresses over time. Nevertheless, in some cases, even very late recanalization can achieve a good clinical outcome.

[0005] During the occlusion of a cerebral artery, the cerebral vascular system can establish collateral circulation through anastomoses between the occluded artery and neighboring arteries, which retrogradely supply the cerebral territory of the occluded artery. This collateral circulation is highly variable, almost individual, and its hemodynamics are poorly understood. This circulation represents a kind of "hemodynamic reserve" for the brain, can modulate the progression of cerebral ischemia, and is considered a factor capable of determining the clinical course and prognosis of patients with ischemic stroke. Collateral circulation follows well-defined circulatory phases: an activation phase, an arterial phase, a capillary or parenchymal filling phase, and a venous phase.

[0006] The permeability of collateral circulation in the various phases is linked to the efficiency of the collateral circulation itself. Indeed, it is possible to observe extensive but inefficient collateral circulations.

[0007] Arteriographic analysis of collateral circulation remains very limited. An approach to analyzing cerebral arteriographic images has been proposed by the MR-CLEAN team for evaluating recanalization of the occluded artery after the TM procedure. However, this type of approach, which used the Minimum Intensity Projection (MIPP) method, only allowed for the isolation of the arterial circulatory phase, which is necessary to assess recanalization. This approach is presented in the following documents:

[0008] - Prasetya H, Ramos LA, Epema T, Treurniet KM, Emmer BJ, van den Vineyard IR, Zhang G, Kappelhof M, Berkhemer OA, Yoo AJ, Roos YB, van Oostenbrugge RJ, Dippel DW, van Zwam WH, van der Lugt A, de Mol BA, Majoie Baring, CB EV HA; MR CLEAN Registry Investigators ; « qTICI: Quantitative assessment of brain tissue reperfusion on digital subtraction angiograms of acute ischemic stroke patients .” ; Int J Stroke 2021 Feb;16(2):207-216 doi: 10.1177 / 1747493020909632 Epub 2020 Feb 25. PMCID: 32098584;

[0009] - Su R, Cornelissen SAP, van der Sluijs M, van Es ACGM, van Zwam WH, Dippel DWJ, Lycklama G, van Doormaal PJ, Niessen WJ, van der Lugt A, van Walsum T. Trans Med Imaging. 2021 Sep;40(9):2380-2391. doi: 10.1109 / TMI.2021.3077113. Epub 2021 Aug 31. PMID: 33939611.)

[0010] Several teams have focused on analyzing the impact of different circulatory phases, particularly the venous phase and collateral circulation, finding significant correlations between venous drainage time and the collateral circulation profile. However, this type of analysis has primarily focused on perfusion CT and CT angiography methods. Such analyses are presented in the following documents:

[0011] - Faizy TD, Kabiri R, Christensen S, Mlynash M, Kuraitis GM, Broocks G, Flottmann F, Marks MP, Lansberg MG, Albers GW, Fiehler J, Wintermark M, Heit JJ. “Favorable Venous Outflow Profiles Correlate With Favorable Tissue-Level Collaterals and Clinical Outcome.” Stroke. 2021 May;52(5): 1761-1767. 33682452 .;

[0012] - Singh N, Bala F, Kim BJ, Najm M, Ahn SH, Fainardi E, Rubiera M, Khaw AV, Zini A, Goyal M, Menon BK, Almekhlafi M. “Time-resolved assessment of cortical venous drainage on multiphase CT angiography in patients with acute ischemic stroke.”,' Neuroradiology. 2022 May;64(5): 897-903. doi:10.1007 / s00234-021-02837-l. Epub 2021 Oct 26. PMID: 34704112.)

[0013] The only tool based on cerebral arteriography data is the ASITN / SIR scale (American Society of Interventional and Therapeutic Neuroradiology / Society of Interventional Radiology). This scale has shown significant limitations regarding the intra- and inter-observer reproducibility of assessments, as described in the medical literature. Such a tool is notably presented in the following document:

[0014] - Ben Hassen W, Malley C, Boulouis G, et al. “Inter- and intraobserver reliability for angiography leptomeningeal collateral flow assessment' by the ASITN / SIR scale. Journal of Neurolnterventional Surgery 2019;11:338-341.

[0015] Such a tool does not allow for an accurate measurement of the quality of collateral circulation.

[0016] We know the following document: David Arthur and Sergei Vassilvitskii, “k- means++: the advantages of careful seeding.”, SODA “07: Proceedings of the eighteenth annual ACM-SIAM symposium on Discrete algorithms”, pages 1027-1035, Philadelphia, PA, USA, 2007. Society for Industrial and Applied Mathematics. ISBN 978-0-898716-24-5. This document describes a k-means classification technique.

[0017] We also know the document F. Petitjean, A. Ketterlin, and P. Gan parski, “A global averaging method for dynamic time warping, with applications to clustering.”, Pattern Recognition, 44(3):678-693, 2011. This document describes a technique, different from the K-means algorithm, for calculating an average of a set of sequences.

[0018] We are familiar with document US20130077839, which describes a method for visualizing changes in blood flow in a sequence of digital subtraction angiography (DSA) images. This method uses temporal contrast curves to generate dynamic images of blood flow.

[0019] We know of document US9375157 describing an angiographic examination method for the representation of blood flow properties.

[0020] We also know of document US8731262, which describes a method for detecting arterial, parenchymal, and venous phases using light intensity. Document CN110942826 allows for determining the onset of a venous phase. And document EP3005943 describes a method for imaging collateral circulation to better characterize it.

[0021] The document “Angiography collateral venous phase: a novel landmark for leptomeningeal collaterals evaluation in acute ischemic stroke”; Consoli A, Pizzuto S, Sgreccia A, Di Maria F, Coskun O, Rodesch G, Lapergue B, Felblinger J, Chen B, Bracard S. J Neurointerv Surg. 2022 Dec 20:jnis-2022-019653. doi: 10.1136 / jnis-2022-019653, highlights the need for a more precise assessment of collateral circulation based on quantitative and qualitative analysis. Currently available scales do not allow for this type of evaluation.

[0022] However, in the previously described systems, the currently proposed collateral circulation characteristics are neither precise nor easy to use, and they do not sufficiently take into account the pathophysiology of collateral circulation, which should be considered separately from cerebral circulation from a hemodynamic perspective. The presence of a substantial network of collateral vessels does not necessarily indicate good collateral circulation. Indeed, these vessels may be present but non-functional. The existence of a venous phase in the collateral circulation region, visualized by cerebral arteriography, appears to be a sign associated with collateral circulation functionality.

[0023] Furthermore, the hemodynamic efficiency of collateral circulation is poorly understood. Indeed, there is no system or algorithm that can automatically calculate the efficiency of collateral vessels.

[0024] The present invention aims at a new method for accurately assessing the extent of collateral circulation.

[0025] Another objective of the invention is to be able to evaluate the effectiveness of collateral circulation. Description of the invention

[0026] At least one of the objectives is achieved with a process for processing angiographic images obtained during an acquisition phase in which a contrast agent is injected into a vascular network of an organ in a state of arterial occlusion; this process includes the following steps: - obtaining a projection map displaying arteries and veins from the angiographic images, - delineation on the projection map of a region of cerebral circulation and a region of collateral circulation, the contrast agent being able to propagate in the region of cerebral circulation during a parenchymal phase and then during a venous phase, - identification of a region of interest (ROI) in the cerebral circulation region, - Determination, in the region of interest, of the instant of the switchover from the parenchymal phase to the venous phase; - Development of an individualized vascular network map distinguishing arteries and veins in the regions of cerebral and collateral circulation from angiographic images. - development of a time-to-peak (TTP) map illustrating the time taken by the contrast agent to reach a maximum contrast level for each pixel of the organ, based on angiographic images, and - generation of a desynchronization map by applying the vascular network map to the time-to-peak (TTP) map with identification of cerebral circulation and collateral circulation regions, and identification of the tipping point.

[0027] Desynchronization refers to a time lag between the arterial / venous phases of the collateral circulation region and those of the cerebral circulation region. The cerebral circulation is considered the reference vascular territory for calculating desynchronization.

[0028] Peak time mapping is a map illustrating the time it takes for a pixel or voxel to reach its maximum contrast. It is a temporal parameter derived from the dynamic analysis of digital subtraction angiography (DSA) image sequences. It represents the time required for a given pixel in the imaging dataset to reach maximum intensity after the administration of a contrast agent. In the context of parametric mapping, peak time (TTP) is calculated for each spatial unit (pixel) based on temporal intensity curves obtained from a DSA image sequence. These intensity curves track the contrast enhancement over time. The point at which the intensity reaches its highest value corresponds to the "peak" time, and this value is recorded and mapped for each spatial unit.The resulting parametric map is a visual representation where the value assigned to each pixel encodes its respective TTP. Such maps are particularly useful for evaluating hemodynamic parameters, as they allow the identification of regions exhibiting delayed perfusion, altered blood flow, or abnormal vascular dynamics.

[0029] Implementation details: 1. Data acquisition: Sequential DSA images are acquired over time after administration of the contrast agent, capturing the dynamic vascular improvement. 2. Signal analysis: for each pixel, a time-intensity curve is generated, illustrating the change in intensity over time. 3. Peak identification: the maximum intensity value on the time-intensity curve is identified, and the corresponding time point is recorded as the TTP. 4. Generating a map: the TTP values ​​for all pixels are compiled into a two-dimensional or three-dimensional parametric map, visually representing the temporal dynamics of contrast enhancement in the anatomical region of interest.

[0030] By encoding TTP values ​​in a parametric map, the invention facilitates improved diagnostic capabilities, enabling clinicians or researchers to quantitatively and visually assess vascular flow dynamics and efficiently identify pathophysiological changes.

[0031] The invention includes identifying the region of interest in the cerebral circulation, enabling the determination of the moment of transition between the parenchymal and venous phases within that region. This region of interest can be determined manually, but it is preferably determined automatically.

[0032] The organ can be a human or animal brain or any other organ containing arteries and veins.

[0033] With the method according to the invention, it is now possible to assess the effectiveness of collateral circulation, symbolizing the presence of a venous phase in the collateral region. The images used are preferably those obtained by digital subtraction angiography (DSA), a medical imaging technique used to clearly visualize blood vessels in tissues, and also therapeutically for performing mechanical thrombectomy procedures. Other techniques for acquiring images of blood vessels can be used, such as multiphase CT angiography.

[0034] This collateral circulation can also be quantified based on the pathophysiological concept of desynchronization. Indeed, although the collateral circulation and the cerebral circulation are anatomically continuous, they may exhibit some hemodynamic differences due to potential microthrombotic phenomena in the microcirculation supplied by collateral vessels. Consequently, these two circulations may not have the same hemodynamic profile and may thus be desynchronized. The smaller the difference between these two circulations (desynchronization), the greater the functionality of the collateral circulation. The present invention uses this principle of desynchronization to assess collateral circulation.

[0035] The process according to the invention advantageously leads to the generation of a desynchronization map carrying quantitative information that is both morphological and physiological.

[0036] The method according to the invention can be used directly in the arteriography suite by neuroradiologists to assess the extent and effectiveness of collateral circulation in patients with large vessel occlusion (M1 segment of the middle cerebral artery, M1-MCA). This would allow for better management of these patients in a hyperacute setting (assessment of patient prognosis, blood pressure management, risk of hemorrhagic transformation after mechanical thrombectomy). Furthermore, the method according to the invention, by its real-time application, can be performed at the very beginning of a procedure, for example, just before thrombectomy, whereas MRI and CT scans are generally performed 30 to 90 minutes earlier.Finally, the method according to the invention can have an important impact on research studies, as collateral assessment has become an important element of research protocols focused on acute ischemic stroke.

[0037] For this reason, measuring desynchronization directly using digital subtraction angiography (an imaging technique used to perform mechanical thrombectomies) should be a more appropriate method.

[0038] It has been observed that the functionality of collateral vessels can be analyzed based on knowledge of the dynamics of the venous phase of collateral circulation. This allows, in particular, the differentiation between an extensive but inefficient collateral vessel network and a potentially less extensive but efficient one. Therefore, identifying the venous phase of collateral circulation is a useful parameter in understanding the hemodynamic characteristics of collateral circulation.

[0039] According to an advantageous feature of the invention, the projection map can be a maximum enhancement projection image (MEP) obtained by calculating a dynamic maximum enhancement projection (dMEP) on a total number N of images from a series of angiographic images according to the formula: MEP = dMEP(t N dMEP(t) being a function defined by: dMEP(td) = min (DSAÇt ...DSA(ty) (1) with DSA the series of angiographic images by digital subtraction, DSA t) being the f eme image in the DSApour image series < t < t N , N being the total number of images in the DSA image series.

[0040] The function dMEP(t) is calculated as the minimum of intensity projections among images at t.

[0041] Other techniques can be used to determine projection mapping, such as MinlP, or Minimum Intensity Projection. Minimum intensity projection is a data visualization method that allows the detection of low-density structures within a given volume. This method uses all the data from a volume of interest to generate a single two-dimensional image. It projects the pixel with the lowest attenuation value from each view across the entire volume onto a 2D image.

[0042] Maximum intensity projection (MIP) can also be used. Both maximum intensity projection and minimum intensity projection are volume rendering techniques in which appropriate editing methods are used to define the volume of interest.

[0043] According to one embodiment of the invention, the ROI used to determine the tipping point can be defined as the superior longitudinal venous sinus along a midline when the organ is the brain. The superior longitudinal venous sinus can be automatically detected from angiographic images.

[0044] Advantageously, the moment of switching from a parenchymal phase to a venous phase in the region of interest can be the moment of acquisition of an angiographic image for which said region of interest reaches a predetermined level of contrast or an acquisition moment determined manually by the operator.

[0045] According to an advantageous feature of the invention, determining the tipping point can include the following steps: - definition of a dynamic maximum enhancement projection function dMEP by dMEPÇt) with DSA the series of angiographic images by digital subtraction, DSA t) being the f eme image in the DSA image series for , N being the total number of images in the DSA image series, - definition of a maximum dynamic rear enhancement projection function (dMEP) calculated in reverse time order according to the formula: dMEP(t) = min (DSA(t) ...DSA(t) N ) (2) - application of the dMEP function progressively on all images of a series of angiographic images, application of the dMEP function regressively on all images of the series of angiographic images, and for each measurement of dMEP and dMEP, calculation of an entropy on the region of interest; - the switching instant being the first instant of an angiographic image for which the entropy of the dMEP function in the region of interest is greater than the entropy of the dMEP function in the same region of interest.

[0046] According to an advantageous feature of the invention, entropy can be calculated according to the following mathematical formula for Shannon entropy:

[0047] E = -i =1 ...MPi log2(Pi) (3) with pt the probability of finding a pixel of intensity belonging to the i eme grey level category in the image (i=l ...M, 1 corresponding to the lowest grey level and M the highest).

[0048] According to an advantageous feature of the invention, the step of delimiting a region of cerebral circulation and a region of collateral circulation may include a segmentation step by applying a deep learning model, for example nnUNet, refined on projection mapping.

[0049] Preferably, the projection map is a maximum enhancement MEP projection image.

[0050] It is also possible to plan to subdivide the collateral circulation region into two regions by segmentation according to a distribution of the cerebral vascular territory, namely: an extended collateral circulation region and a pure collateral circulation region.

[0051] According to an advantageous feature of the invention, vascular network mapping is calculated using time-density evolution curves of each pixel of the organ as multi-variant input parameters of a K-means classification based on the application of the dynamic time-deformation (DTW) center-of-center algorithm.

[0052] According to an advantageous feature of the invention, prior to the generation of the desynchronization map, the time-at-peak (TTP) map can be normalized by dividing each pixel value by the maximum value in the time-at-peak (TTP / max(TTP)) map; and the switchover time can also be normalized in the same way, by dividing by the maximum value in the time-at-peak (TTP) map.

[0053] A TTP is defined as the time elapsed until maximum opacity, represented by the peak of the curve. A DSA TTP map is obtained by calculating the TTP for each pixel.

[0054] According to an advantageous feature of the invention, the method may further include a step of generating desynchronization maps displaying values ​​lower or higher than the normalized tipping time (similarly to the previous section, nPV=PV / max(TTP)), so as to detect the presence or absence of a collateral venous phase.

[0055] This procedure allows for verification of the presence or absence of collateral circulation. It enables an evaluation of the venous phase of the collateral circulation and thus allows for the deduction of the efficiency of the collateral vessels. This, in turn, makes it possible to identify a functional collateral network in contrast to those that are not.

[0056] According to an advantageous feature of the invention, the method may further include a step of determining a desynchronization distance as being a ratio between an average value of peak times (TTP) in the cerebral circulation region and an average value of peak times (TTP) in the collateral circulation region.

[0057] This desynchronization value is preferably calculated for the entire duration of both the arterial and venous phases.

[0058] The desynchronization value provides information on the functionality of the collateral circulation region.

[0059] According to an advantageous feature of the invention, the method may further include a step of determining a vascular density calculated from the mapping of the vascular network. This calculation may, for example, be a normalization, and in particular the division of the overall surface area of ​​the vascular network by the surface area of ​​the collateral region.

[0060] According to an advantageous feature of the invention, the method may further include a step of determining a vascular complexity calculated by fractal analysis of the collateral circulation region in the projection mapping (MEP).

[0061] Vascular density and vascular complexity of the collateral circulation region allow for the evaluation of collateral circulation morphology.

[0062] According to another aspect of the invention, a computer program product is provided that can be downloaded from a communication network and / or stored on a computer-readable medium and / or executable by a microprocessor, comprising program code instructions for implementing the angiographic image processing method according to the method of the invention.

[0063] According to another aspect of the invention, a computer-readable recording medium is provided comprising instructions which, when executed by a computer, cause the computer to implement the method according to the invention. Description of the figures and methods of realization

[0064] Other advantages and features of the invention will become apparent upon reading the detailed description of implementations and embodiments, which are by no means limiting, and the following attached drawings:

[0065] Figure 1 is a schematic view of a system for implementing the method according to the invention,

[0066] Figure 2 is a schematic view of a flowchart illustrating different stages of the process according to the invention,

[0067] Figure 3 is a schematic view of a projection map showing the boundaries of the cerebral circulation region and the collateral circulation region.

[0068] Figure 4 is a schematic view illustrating the location of the crossing point as a region of interest in a superior longitudinal venous sinus,

[0069] Figure 5 is a schematic view illustrating the detection of the moment of transition from a parenchymal phase to a venous phase by bidirectional application, forward and backward, of the dynamic function dMEP,

[0070] Figure 6 is a view of a customized vascular atlas superimposed on a vascular network mapping image.

[0071] Figure 7 is a view of a desynchronization map of a patient with good and functional collateral circulation, and

[0072] Figure 8 is a view of a desynchronization map of a patient with inefficient collateral circulation and absence of collateral venous phase in the collateral circulation region.

[0073] The embodiments described below are not exhaustive; in particular, variants of the invention may be implemented comprising only a selection of features described hereafter, isolated from the other features described, if this selection of features is sufficient to confer a technical advantage or to differentiate the invention from the prior art. This selection includes at least one preferably functional feature without structural details, or with only a portion of the structural details if this portion alone is sufficient to confer a technical advantage or to differentiate the invention from the prior art.

[0074] In a digital subtraction angiography system, opacification of cerebral vascular structures begins a few seconds after injection of the contrast agent and ceases when the contrast leaves the hemisphere drained by the cerebral venous system. In stroke cases, from a functional perspective, two circulatory systems can be identified: the cerebral circulation and the activated collateral circulation. The circulatory phases "artery > parenchyma / capillary > venous phase" must be considered for both circulations, and therefore temporal markers represent the transit time between two consecutive phases. It has been observed that the onset of the venous phase of the collateral circulation provides important information regarding the functional efficiency of the collateral circulation.Thus, the desynchronization between cerebral circulation and collateral circulation is amplified even more, mainly in the venous phase.

[0075] Therefore, the present invention is based on the identification of the venous phase for cerebral and collateral circulation. Cerebral circulation can be used as a reference circulation. The invention also makes it possible to identify the point at which the parenchymal phase switches to the venous phase. Arteries and veins are labeled and then overlaid with a peak timing map to detect any desynchronization. This allows for the identification of the presence or absence of the collateral venous phase, which is a key indicator for assessing collateral efficiency. Furthermore, the invention enables the morphological characterization (extent) of collateral circulation, using parameters such as the complexity and density of the vascular network.

[0076] Figure 1 is a simplified schematic view of a medical imaging system capable of implementing the method according to the invention. The system comprises an arteriography machine 10 for acquiring arteriographic images, DSA2D sequences, when a contrast agent is injected into the blood vessels of a patient lying on the bed of the arteriography machine 10. The image data signals can be in DICOM format (Digital Imaging and Communications in Medicine). Other formats can also be used.

[0077] A controller 11 is provided to control the operation of the arteriography machine 10. A processing unit 12 is connected both to the arteriography machine 10 for image retrieval and to the controller 11 to initiate image acquisition sequences. This processing unit 12 includes hardware and software means for data acquisition, data storage, data processing, and wired or wireless input and output communication.

[0078] The processing unit 12 includes, in particular, a microprocessor or microcontroller configured to implement the steps of the process according to the invention. The images used for the invention are acquired live but may also be from a prior acquisition and stored in memory. The processing unit 12 manages image processing applications such as, for example, visualization applications, computer-aided diagnostic (CAD) applications, medical image rendering applications, anatomical segmentation applications, image recording applications, or any other type of medical image processing application.

[0079] A display screen 13 is connected to the processing unit 12 for viewing images and maps.

[0080] We will now describe the implementation of the method according to the invention for a patient who is to undergo a mechanical thrombectomy for an occlusion of the M1 segment of the middle cerebral artery.

[0081] The method according to the invention will allow the morphology (extent) of collateral circulation to be evaluated, in particular by means of vascular complexity and density. It will also allow the efficiency of collateral circulation to be evaluated, which has been defined as the existence of a venous phase in the collateral region and quantified based on the pathophysiological concept of desynchronization.

[0082] In Figure 2, the input data are angiographic images 20 obtained using the digital subtraction angiography technique, DSA, which consists of a series of projected digital X-ray images with 2D contrast subtracted from a pre-contrast image. These angiographic images 20 are acquired in anteroposterior projection.

[0083] Preferably, the acquisition rate is at least six frames per second, but the rate can go down to two frames per second.

[0084] The process according to the invention makes it possible to generate two maps: a projection map 21 and a peak time map, TTP, 22 from the angiographic images 20.

[0085] From the projection mapping, in step 23 we delineate the cerebral circulation region and the collateral circulation region, the latter of which can also be subdivided.

[0086] In step 24, the point at which the parenchymal phase switches to the venous phase is identified using an entropy calculation algorithm. With the detection of this switching point, a personalized hemispheric vascular atlas can be reconstructed using a k-means classification algorithm based on the dynamic time-deformation (DTW) centroid. This results in a vascular network map in step 25 in which arteries and veins are separated.

[0087] During step 26 of generating a desynchronization map, the vascular atlas or vascular network 25 is applied to the time map at peak 22. This generates: - Desynchronization mapping 261 allows for the evaluation of the effectiveness of collateral circulation, - a desynchronization distance 262 which is a quantitative data, allowing the ratio of desynchronization between the two circulations to be indicated, - a vascular network density 263 which is also a quantitative data allowing the morphology of the vascularization to be evaluated, - a complexity of the vascular network 264 which is also a quantitative data allowing to evaluate the morphology of the vascularization.

[0088] Steps 21-23:

[0089] We will now describe the method of generating the enhancement projection maps and peak times (TTP), steps 21 and 22 of Figure 2, as well as the distinction between cerebral and collateral regions, step 23 of Figure 2.

[0090] The projection map is a maximum enhancement MEP projection image obtained by calculating a dynamic maximum enhancement (dMEP) projection over a total number t Nof angiographic images acquired during a sequence, according to the formula:

[0091] MEP = dMEP(t N )

[0092] dMEP(t) being a function defined by:

[0093] dMEPÇt) = min (DSAÇt ...DSA t)) (1)

[0094] with DSA the series of angiographic images, DSA(t being the f eme image in the DSA image series for < t < t N , t N being the total number of images in the DSA image series.

[0095] The function dMEP(t) is calculated as the minimum of intensity projections among images tl at t.

[0096] The peak time mapping indicates, for each pixel, the moment when the contrast agent concentration reaches its maximum; it is a TTP map of the "pixel-wise time to peak" type in English.

[0097] Next, in step 23, the segmentation is performed by applying a deep learning model, for example nnUNet, refined on the projection map. Three regions are defined for the segmentation: the cerebral region and two collateral circulation regions according to the distribution of the cerebral vascular territory, namely: the extended collateral circulation, and the pure collateral circulation region.

[0098] Figure 3 is an example of a projection map according to the MEP principle. We distinguish the cerebral circulation region 30 (which corresponds to the entire vascular region of the cerebral hemisphere where the contrast agent is injected), the collateral circulation region 31 including the extended collateral circulation 32, which corresponds to the region of pure collateral circulation including the part of the cerebral circulation in continuity with the collateral circulation 31 and the region of pure collateral circulation 33, which corresponds to the region of collateral circulation without considering the continuity with the region of cerebral circulation.

[0099] Step 24:

[0100] We will now describe the method for identifying the tipping point, step 24 of figure 2.

[0101] The switchover time is defined as the moment of the first image in which cerebral venous circulation can be visualized. In Figure 4, region 40 is the ROI of the superior longitudinal venous sinus (SLVS). slvs on the midline, the line separating the two cerebral hemispheres. This ROI 40 is determined to easily calculate the moment of transition from the cerebral parenchymal phase to the venous phase. The ROI 40 is automatically identified on the projection map as follows: it corresponds to the most supero-medial region, which is, for example, located between, on the one hand, the most medial part of the extended collateral circulation region, and on the other hand, the midline separating the two cerebral hemispheres.

[0102] To determine the switching instant, the maximum dynamic enhancement projection function is applied in a forward direction (denoted dMEP) and in a backward direction (denoted dMEP). dMEP / dMEP is calculated from both directions for the entire duration of a DS A image acquisition cycle. The forward direction begins at the start of the acquisition, and the backward direction begins at the end of the acquisition. An entropy (E) R0[sLVS dMEP(t) is calculated in the ROI slvs in the forward direction. An entropy (E R0 [ SLVS dMEP(ty) is calculated in the ROI slvs in the reverse direction. Figure 5 illustrates the two-way calculation.

[0103] Entropy measures the information content of ROI sh , s of the current image. Its mathematical form is defined by:

[0104] E = -i =1 ...MPi log2(Pi) (3)

[0105] where pt is the probability of finding a pixel of intensity belonging to the i emeGray level category in the image (i=l ...M, 1 corresponding to the lowest gray level and M the highest). The higher the entropy, the more information it contains. Thus, in the forward dynamic MEP direction, the information inside ROI sh , s increase until the contrast agent penetrates and remain relatively stable until the end. In the rear dynamic MEP direction, the information remains almost stable at the beginning because the contrast agent's maximum is reached at the start of the venous phase and remains so until the end of the acquisition cycle. The first crossing of the two curves corresponds to the moment when the information inside the crossover point reaches a maximum for the first time. Therefore, the start of the venous phase is identified by detecting the first image in which (E ROIsLVS dMEP(t)) has exceeded (E ROIsLVS dMËP(t)).

[0106] Step 25:

[0107] We will now describe the method for generating the vascular network map, step 25 of figure 2.

[0108] To extract morphological and physiological information, the arterial / venous vascular network must first be labeled. A custom atlas of cerebral and collateral circulation is created using k-means classification based on the dynamic time warp (DTW) center of mass algorithm, using the dtai distance library implemented in Python v3.10. The DTW-based k-means method is a well-known time series classification algorithm, typically used for one-dimensional signals, n-dimensional signals, or satellite images. However, it has never been used for DSA image series from mechanical thrombectomy procedures.The evolution of the temporal density of pixels at different types of vessels (arteries or veins) and circulatory systems (cerebral or collateral) differs in terms of the arrival time of the contrast agent, the direction of wavefront propagation, and the background amplitude. The temporal density evolution curves of each pixel can be considered as multi-variable input parameters for K-means classification based on the dynamic time warp (DTW) center of mass algorithm. Arteriovenous labeling is applied in two steps:

[0109] 1) on DSA images before the tipping point, using 4 classes and a window size (called the Sakoe-Chiba bandwidth) of 60; the cerebral e / collateral artery region is identified by selecting the two best matching classes

[0110] 2) On the entire DSA series, using 5 classes and a window size of 60; the cerebral / collateral venous region is identified by selecting the best corresponding class. The output clustering maps lead to the separation of the arterial and venous regions in Figure 6. The grouping of the arterial and venous regions yields a mask, representing the entire vascular network, which is used in the next step. [YES] Step 26:

[0112] We will now describe the method for generating the desynchronization map, step 26 of figure 2.

[0113] The desynchronization map is defined as a normalized peak time map multiplied by the mask created in step 25. The peak time map (TTP) is normalized by dividing each value of this peak time map by the maximum TTP value. Thanks to this normalization, all of a patient's data are mapped on the same scale (between 0 and 1).

[0114] Figure 7 is an example of a desynchronization map of a patient with functional and healthy collateral circulation. As mentioned above, the arterial and venous atlas is used as a mask and superimposed on the normalized peak time map.

[0115] The two small images shown on the right correspond to before (top) and after (bottom) the passage point which displays with a normalized nPV value of 0.53. This example shows good spatial coverage of the collateral circulation region in each phase (arterial, parenchymal and venous), which corresponds to low desynchronization.

[0116] Figure 8 is an example of a desynchronization map of a patient with inefficient collateral circulation and an absence of a venous phase in the collateral circulation region. Compared to the situation in Figure 7, although the collateral region is well-improved in the arterial and parenchymal phases, flow can barely reach the collateral region in the venous phase, with the normalized nPV value of the point of passage being 0.46.

[0117] The collateral region and its subdivisions are delineated and shown in Figures 7 and 8. The normalized value of the transition point serves as the reference value for the extent of desynchronization. The two small images in each of Figures 7 and 8 represent the desynchronization map for values ​​above the reference value, and for values ​​below the reference value. This allows visualization of the presence or absence of a venous collateral phase. The detection of the venous collateral phase is therefore based on the desynchronization map.

[0118] Morphological indices are also generated in the form of density and complexity: collateral circulation density is calculated based on the customized collateral circulation atlas, using the overall surface area of ​​the vascular network divided by the surface area of ​​the collateral region. Its complexity is calculated using the fractal dimension extracted from the collateral circulation region of the projection map.

[0119] Thus, the invention enables the generation of a desynchronization map containing quantitative morphological and physiological information by implementing two steps: cerebral venous phase detection and desynchronization map generation. During cerebral venous phase detection, the onset image of the cerebral venous phase is detected, for example, using a dual-directional dynamic maximum enhancement algorithm, which measures the entropy changes of the ROISVLS over the entire data acquisition cycle. This phase and the personalized cerebral vascularization map are used to determine the desynchronization map.Desynchronization mapping provides the following information: complexity and density of collateral circulation, presence of a collateral venous phase, and quantification of the temporal desynchronization between cerebral circulation and collateral circulation measured in seconds.

[0120] Of course, the invention is not limited to the examples just described. Many modifications can be made to these examples without departing from the scope of the present invention as described.

Claims

Demands 1. A method for processing angiographic images obtained during an acquisition phase in which a contrast agent is injected into a vascular network of an organ in a state of arterial occlusion; this method comprising the following steps: - obtaining a projection map displaying arteries and veins from angiographic images, - delineation on the projection map of a region of cerebral circulation and a region of collateral circulation, the contrast agent being able to propagate in the region of cerebral circulation during a parenchymal phase and then during a venous phase, - identification of a region of interest (ROI) within the cerebral circulation, - determination, within the region of interest, of the instant of the switch from the parenchymal phase to the venous phase. - development of an individualized vascular network map distinguishing arteries and veins in cerebral and collateral circulation regions from angiographic images, - development of a peak time map illustrating the time taken by the contrast agent to reach a maximum contrast level for each pixel of the organ from angiographic images, and - generation of a desynchronization map by applying the vascular network map to the time-to-peak (TTP) map with identification of cerebral circulation and collateral circulation regions, and identification of the tipping point.

2. A method according to claim 1, characterized in that the projection map is a maximum enhancement projection image (MEP) obtained by calculating a dynamic maximum enhancement projection (dMEP) on a total number N of images from a series of angiographic images according to the formula: MEP = dMEP(t w )dMEP(t) being a function defined by: dMEPÇt) = min (DSAÇt ... DSA t)) (1) with DSA the series of angiographic images by digital subtraction, DSA(t being the f eme image in the DSApour image series < t < t N , N being the total number of images in the DSA image series.

3. A method according to claim 1, characterized in that the region of interest is the superior longitudinal venous sinus ROI sivs) on a midline when the organ is a brain.

4. Method according to claim 3, characterized in that the instant of switching from a parenchymal phase to a venous phase in the region of interest ROI slvs ) is the moment of acquisition of an angiographic image for which said region of interest reaches a predetermined level of contrast.

5. A method according to any one of the preceding claims, characterized in that the determination of the tipping moment comprises the following steps: - definition of a dynamic maximum enhancement projection function dMEP by: dMEPÇt) = min (DSAÇt ... DSA t)) (1) with DSA the series of angiographic images by numerical subtraction, DSA(t being the f eme image in the DSA image series for , N being the total number of images in the DSA image series, - definition of a maximum dynamic rear enhancement projection function (dMEPj calculated in reverse time order according to the formula: dMEPÇt) = min (DSA(t) ...DSA(t N y) (2) - application of the dMEP function progressively on all images of a series of angiographic images, application of the dMEP function regressively on all images of the series of angiographic images, and for each measurement of dMEP and dMEP, calculation of an entropy on the region of interest; - the switching instant being the first instant of an angiographic image for which the entropy of the dMEP function in the region of interest is greater than the entropy of the dMEP function in the same region of interest.

6. A method according to claim 5, characterized in that the entropy is calculated according to the following mathematical formula for Shannon entropy: with pt the probability of finding a pixel of intensity belonging to the i eme grey level category in the image.

7. A method according to any one of the preceding claims, characterized in that the step of delimiting a region of cerebral circulation and a region of collateral circulation includes a segmentation step by applying a refined deep learning model to the projection map.

8. A method according to any one of the preceding claims, characterized in that the vascular network mapping is calculated using time-density evolution curves of each pixel of the organ as multi-variant input parameters of a K-means classification based on the application of the dynamic time-strain (DTW) center of gravity algorithm.

9. A method according to any one of the preceding claims, characterized in that prior to the generation of the desynchronization map, the time-at-peak (TTP) map is normalized by dividing each pixel value by the maximum value in the time-at-peak (TTP) map; and the switchover time is also normalized in the same way, by dividing by the maximum value in the time-at-peak (TTP) map.

10. Method according to claim 9, characterized in that it further comprises a step of generating desynchronization maps displaying values ​​lower or higher than the normalized tipping time, so as to detect the presence or absence of a collateral venous phase.

11. A method according to any one of the preceding claims, characterized in that it further comprises a step of determining a desynchronization distance as being a ratio between an average value of the peak times in the cerebral circulation region and an average value of the peak times in the collateral circulation region.

12. A method according to any one of the preceding claims, characterized in that it further comprises a step of determining a vascular density calculated from the vascular network mapping.

13. A method according to any one of the preceding claims, characterized in that it further comprises a step of determining a vascular complexity calculated by fractal analysis of the collateral circulation region in the projection map.

14. Product computer program downloadable from a communication network and / or stored on a computer-readable medium and / or executable by a microprocessor, characterized in that it includes program code instructions for the implementation of the angiographic image processing method according to any one of the preceding claims.

15. Computer-readable recording medium comprising instructions which, when executed by a computer, cause the computer to carry out the method according to any one of claims 1 to 13.

Images