System and method for determining a segment for ablation

Abstract

Disclosed herein are systems and methods for determining one or more arrhythmia targets for ablation. The method may include receiving one or more maps, identifying abnormalities in the one or more maps, combining the one or more maps, and defining one or more arrhythmia targets based on overlap of the identified abnormalities in the combined one or more maps.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims the benefit of U.S. Provisional Patent Application No. 62 / 598,162 filed on December 13, 2017, U.S. Provisional Patent Application No. 62 / 607,084 filed on December 18, 2017, and PCT Patent Application No. PCT / US2018065278 filed on December 12, 2018, which are incorporated herein by reference in their entirety. Technical Field

[0003] The present technology relates to determining arrhythmia targets for ablation, and more particularly, to multimodal image mapping and risk conditions for determining atrial or ventricular segments to be ablated. Background Art

[0004] Sudden cardiac arrest (SCA) is the largest cause of death in developed countries. The majority of SCAs are attributed to arrhythmias, either ventricular tachycardia (VT) or atrial fibrillation (AF). Current invasive visualization techniques are used for applications targeting arrhythmia treatment, such as cardiac catheterization. Invasive procedures can pose significant risks to patients. Current treatments for arrhythmias include invasive catheter ablation. Non-invasive ablation using stereotactic body radiation therapy has recently been developed for the treatment of arrhythmias. For invasive and non-invasive ablations, imaging and clinical data are manually converted by clinicians into targets for treatment, resulting in treatment variability among users. In addition, the type of ablation used and the target selection (e.g., size, location) are determined ad hoc without formal decision support.

[0005] Therefore, there is a need for improved identification of arrhythmia target volumes and treatment planning for non-invasive and invasive arrhythmia ablation, and improved decision support for selection of treatment for arrhythmia ablation.A non-invasive procedure for mapping and treating arrhythmias in a patient is achieved. Summary of the invention

[0006] Provided herein is a method for determining one or more arrhythmia targets for ablation. The method may include receiving one or more maps selected from electrical maps, anatomical maps, functional maps, and combinations thereof; identifying abnormalities in one or more maps; combining one or more maps; and defining one or more arrhythmia targets based on the overlap of the identified abnormalities in the combined one or more maps. In one aspect, the electrical map may be an electrocardiogram image, the anatomical map may be at least one of a computed tomography image or a magnetic resonance image, and the functional map may be at least one of an optical emission computed tomography image, a positron emission tomography image, or an echocardiogram image. In some aspects, the abnormality may be identified as one or more segments on one or more maps, one or more contours on one or more maps, or a combination thereof. The abnormality may be identified manually, automatically, or a combination thereof. In one aspect, the method may further include adjusting one or more arrhythmia targets based on additional data. The additional data may be selected from the weight of the received mapping, confidence score, risk, clinical data, demonstration data, prior knowledge of a specific patient, prior knowledge of previous patients, toxicity, efficacy, data quality, data importance, data reproducibility, physiology of the heart, size of the scar, number of ventricular tachycardias, type of cardiomyopathy, transmurality, location of the abnormality, age, gender, size of the heart, ejection fraction, thickness of the heart, medications, comorbidities, or a combination thereof. In some aspects, the method may also include ablating the one or more arrhythmia targets. Ablating the at least one arrhythmia target may be non-invasive, such as using SBRT. In some aspects, the method may also include weighting the received mapping. In other aspects, the method may also include determining a confidence score.

[0007] In some aspects, the method may further include generating a decision support module. The decision support module may also include output of one or more arrhythmia targets and one or more of the following: a description of scar pattern, a description of scar burden size and location, a suggested amount of ablation to achieve, a confidence score for the combined mapping, a list of affected structures, general recommendations, an expected success rate of non-invasive therapy, or an expected success rate of using other treatment modalities.

[0008] Further provided herein is at least one non-transitory computer-readable medium storing instructions that, when executed by at least one processor, cause the at least one processor to: receive one or more maps selected from an electrical map, an anatomical map, a functional map, and combinations thereof; combine the one or more maps; and determine one or more arrhythmia targets for ablation based on the combination of the one or more maps. The at least one processor may also adjust the one or more arrhythmia targets for ablation. The at least one processor may also weight the received maps. In one aspect, the at least one processor may determine a confidence score. The electrical map may be an electrocardiogram image, the anatomical map may be at least one of a computed tomography image or a magnetic resonance image, and the functional map may be at least one of an optical emission computed tomography image, a positron emission tomography image, or an echocardiogram image. In another aspect, the at least one processor may: receive one or more clinical maps, including at least one of demographics, surgical history, information about ventricular tachycardia, and a previous electroanatomical map; supplement the combined one or more maps with the one or more clinical maps; and determine one or more updated ablations for one or more arrhythmia targets for ablation based on the supplemented combined maps.

[0009] In some aspects, the at least one processor can also generate a decision support module. The decision support module can also include output of one or more arrhythmia targets and one or more of the following: a description of scar pattern, a description of scar burden size and location, a recommended amount of ablation to achieve, a confidence score for the combined mapping, a list of structures at risk, general recommendations, expected success rate of non-invasive therapy, or expected success rate of using other treatment modalities.

[0010] Additional aspects and features are partially set forth in the following description and will become apparent to those skilled in the art upon reading the specification or may be learned by practicing the disclosed subject matter. A further understanding of the nature and advantages of the present disclosure may be achieved by reference to the remainder of the specification and drawings, which are a part of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] This disclosure contains at least one drawing executed in color. Copies of this disclosure with color drawing(s) will be provided by the Office upon request and payment of the necessary fee.

[0012] In order to describe the manner in which the above and other advantages and features of the present disclosure can be obtained, the principles briefly described above will be described in more detail by reference to specific embodiments thereof illustrated in the accompanying drawings. It should be understood that these drawings depict only exemplary embodiments of the present disclosure and are therefore not to be considered limiting of its scope, and the principles herein are described and explained with additional specificity and detail through the use of the accompanying drawings:

[0013] Figure 1A A flow chart of an example method of 100 is illustrated.

[0014] Figure 1B A flow chart of an example method of 101 is illustrated.

[0015] Figure 2 A flow chart of an example method of 200 is illustrated.

[0016] Figure 3A , 3B and 3C illustrate examples of cross-sections of ventricular segments.

[0017] Figure 4 An embodiment of an example system is illustrated.

[0018] Figure 5 Illustration of an example machine learning environment.

[0019] Fig. 6A illustrates example decision support tool output.

[0020] Figure 6B illustrates example decision support tool output.

[0021] Figure 6C illustrates example decision support tool output.

[0022] Figure 7 Illustration of an example workflow for electrophysiology-guided, noninvasive cardiac radiation ablation.

[0023] Fig. 8A Shown are the total number of ventricular tachycardia (VT) episodes that occurred in each of the five study patients for 3 consecutive months before treatment and for 12 months after treatment, including appropriate shocks induced by an implantable cardioverter-defibrillator (ICD), appropriate ICD antitachycardia pacing, and persistent untreated VT. In patients 4 and 5, the number of VT episodes was significantly greater than in patients 1, 2, and 3; therefore, the numbers shown for patients 4 and 5 were divided by 30 to allow for a comparison of the same size.

[0024] Figure 8B Shown in Fig. 8A Total number of ICD shocks for all five patients during the same time frame.

[0025] Figure 8C Shown in Fig. 8A The number of episodes of antitachycardia pacing occurred in all five patients within the same time frame.

[0026] Fig.9A Shown are serial assessments of left ventricular ejection fraction after treatment for each patient, as assessed by echocardiography. The mean increase was 6 percentage points (range, .2 to 22).

[0027] Fig. 9B Serial chest CT scans after treatment are shown for patient 1. The treated area is shown in blue. At 3 months, there were contiguous focal inflammatory changes within the lung parenchyma, which had almost disappeared at 12 months.

[0028] Fig. 10A Shows marked dilation of small vessels at the interface of dense fibrosis (upper right) and viable myocardium (lower left) in a postmortem heart specimen obtained from patient 5, who suffered a fatal stroke 3 weeks after treatment. There was no acute myocardial inflammation or acute cellular necrosis.

[0029] Fig. 10B The rectangular "boxcar" shaped nucleus and surrounding dilated arterioles and venules are shown, consistent with radiation exposure. The endothelial cells are normal in appearance (black arrows) and show long, thin, nonreactive nuclei. Hypertrophic cardiomyocytes are also seen (white arrows).

[0030] Fig.11 Treatment efficacy estimates are shown. Eighteen patients survived to 6 months. Patients with incessant VT or sustained slow VT below the ICD detection rate are marked with diamonds (n = 5); these episodes are not included in the total. Patients with cardiomyopathy with PVC involvement are marked with a plus sign (n = 2) and are expressed as the PVC burden (percentage) captured on the 24-hour Holter monitor. Each row represents a patient; the blue line represents before ablation and the red line represents after ablation. The upper limit is artificially truncated at 200 episodes. Patients are arranged by recurrence during follow-up, ranging from the largest (bottom) to the smallest (top).

[0031] Fig.12 Figure 24-hour PVC burden and change in left ventricular ejection fraction. Two patients with PVC-related cardiomyopathy were enrolled and treated. The PVC burden measured by 24-hour Holter monitor was 24% and 26% at baseline, respectively. Longitudinal PVC burden (red line) is shown. Left ventricular ejection fraction was measured by echocardiography. Longitudinal LV ejection fraction (green line) is shown. LV ejection fraction improved in both patients as PVC burden decreased.

[0032] Fig.13AKaplan-Meier curves showing overall survival for all patients. The actuarial overall survival was 89% at 6 months and 72% at 12 months.

[0033] Fig. 13B Stacked bar chart showing the use of antiarrhythmic drugs by patient at baseline and after 6 months of treatment. The y-axis represents the total number of antiarrhythmic drugs used, and the size of each color is proportional to the number of drugs used in that class of antiarrhythmic drugs. Amiodarone use was divided into high-dose (≥300 mg / day) and low-dose (<300 mg / day). Class I drugs consisted of mexiletine and flecainide. Class III drugs consisted of sotalol.

[0034] Fig. 13C Graphical representation of the mean scores reported at baseline, 6 weeks, and 6 months of treatment in 3 selected domains of the Short Form-36 questionnaire for the 18 patients alive at 6 months - blue for social functioning, green for change in health status, and purple for general health status. Asterisks indicate significant changes in mean scores over time (p<0.05).

[0035] Fig.14 The ECG of the patient in Example 3 is displayed.

[0036] Fig.15A and 15B MRI mapping of the patient in Example 3 is shown.

[0037] Fig.16A and 16B The CT map of the patient in Example 3 is shown.

[0038] Fig.17 The PET map of the patient in Example 3 is shown.

[0039] Fig.18A , 18B and 18C show the ECGI map of the patient in Example 3.

[0040] Fig.19 The outline of the image of the patient in Example 3 is shown.

[0041] Fig. 20A , 20B , 20C, 20D, 20E and 20F show segmental abnormalities identified for the input mapping of Example 3. Figure 20G The output of the method is displayed using color-coded probabilities to define one or more arrhythmia targets.

[0042] Fig.21A and 21B The ECG of the patient in Example 4 is displayed.

[0043] Fig. 22CT mapping of the patient in Example 4 is shown.

[0044] 23A and 23B show PET maps of the patient in Example 4.

[0045] Fig.24A , 24B , 24C and 24D show ECGI mapping of VT1 and VT2 for the patient in Example 4. Fig.24E and 24F Additional ECGI maps for the patient in Example 4 are shown.

[0046] Fig.25A and 25B The outline of the image of the patient in Example 4 is shown.

[0047] Fig.26A , 26B , 26C, 26D, 26E, 26F, 26G and 26H show the segmental abnormalities identified from each input map of the patient in Example 4. Fig.26I Displays the output of the method defining one or more arrhythmia targets using color-coded probabilities. DETAILED DESCRIPTION

[0048] Various embodiments of the present disclosure are discussed in detail below. Although specific implementations are discussed, it should be understood that this is done for illustrative purposes only. Those skilled in the relevant art will recognize that other components and configurations may be used without departing from the spirit and scope of the present disclosure. Therefore, the following description and drawings are illustrative and should not be interpreted as limiting. Many specific details are described to provide a thorough understanding of the present disclosure. However, in some cases, in order to avoid confusing the description, well-known or conventional details are not described. Reference to an embodiment or embodiments in the present disclosure may be a reference to the same embodiment or any embodiment; and, such reference means at least one of the embodiments.

[0049] Reference to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present disclosure. The phrase "in one embodiment" appearing in various places in the specification is not necessarily all referring to the same embodiment, nor is a separate or alternative embodiment mutually exclusive of other embodiments. In addition, various features are described that may be exhibited by some embodiments but not others.

[0050] As used herein, "magnetic resonance imaging" (MRI) refers to the use of magnetic fields and radio waves to form images of the body. Typically, when used in cardiac situations, cardiovascular magnetic resonance imaging (CMR) involves ECG gating, which counteracts artifacts created by the beating heart.

[0051] As used herein, "computed tomography" (CT) refers to X-ray images taken from a patient at different angles to produce tomographic (cross-sectional) images.

[0052] References to "electrocardiographic imaging" (ECGI) refer to techniques that non-invasively reconstruct epicardial potentials, electrocardiograms, and activation sequences (isochrones) from electrocardiographic body surface potentials. In short, the patient undergoes a CT or MRI scan while wearing an electrode vest that records electrical activity. The primary electrical activity signal comes from cardiac electrical activity. The electrical information from the body surface can then be registered to a patient-specific heart model derived from the CT or MRI images to display the characteristics of the cardiac electrical activity mapped onto the patient's anatomy. Useful information includes: where the heart starts to beat, the order in which cardiac tissue depolarizes, and which parts of the heart have abnormal depolarization behavior. Typically, the patient is donned with the electrode vest, the electrical data acquired, and the CT scan performed in less than 30 minutes.

[0053] Reference to "stereotactic body radiation therapy" (SBRT), also known as stereotactic ablative radiation therapy (SABR) or stereotactic radiosurgery (SRS), means the precise delivery of high-dose radiation in a small number (usually less than 5) components to targets within the body with minimal exposure to normal adjacent tissue.

[0054] In the context of the present disclosure and in the specific context in which each term is used, the terms used in this specification generally have their ordinary meaning in the art. Alternative language and synonyms may be used for any one or more of the terms discussed herein, and there is no special significance for whether the term is elaborated or discussed in this article. In some cases, synonyms for certain terms are provided. The enumeration of one or more synonyms does not exclude the use of other synonyms. The examples used anywhere in this specification, including examples of any term discussed herein, are illustrative only and are not intended to further limit the scope and meaning of the present disclosure or any example term. Similarly, the present disclosure is not limited to the various embodiments given in this specification.

[0055] Without intending to limit the scope of the present disclosure, examples of instruments, devices, methods and related results thereof according to embodiments of the present disclosure are given below. Note that for the convenience of the reader, titles or subtitles may be used in the examples, which in no way limit the scope of the present disclosure. Unless otherwise defined, the technical and scientific terms used herein have the meanings commonly understood by those of ordinary skill in the art to which the present disclosure belongs. In the event of a conflict, this document and its definitions shall prevail.

[0056] Additional features and advantages of the present disclosure will be set forth in the following description and will be apparent in part from the description, or may be learned by practicing the principles disclosed herein. The features and advantages of the present disclosure may be realized and obtained by the tools and combinations particularly pointed out in the appended claims. These and other features of the present disclosure will become more apparent from the following description and the appended claims, or may be learned by practicing the principles set forth herein.

[0057] The methods described herein may be used to implement systems and methods for identifying arrhythmia targets using multimodal imaging techniques.

[0058] Underlying cardiomyopathy and scarring most commonly cause ventricular arrhythmias. Scarring caused by cardiomyopathy (e.g., a previous heart attack) forms the substrate for abnormal electrical circuits within the heart, which can lead to VT. Depending on the symptoms, severity, and cause of the arrhythmia, treatment options include, but are not limited to, antiarrhythmic medications, pacemaker / defibrillator placement, surgical ablation, catheter-based ablation (endocardial, epicardial) using radiofrequency energy to cause thermal injury, SBRT non-invasive ablation, and / or a combination thereof.

[0059] In the absence of ventricular scarring (idiopathic ventricular tachycardia), invasive catheter ablation is largely effective. However, catheter ablation of cardiomyopathy ventricular tachycardia is associated with a recurrence rate of up to 50% at 6 months.

[0060] In electrophysiology (EP) guided non-invasive cardiac radioablation (ENCORE), patients are targeted pre-treatment by using available cardiac imaging and electrophysiological data and visually registering with information provided by non-invasive ECGI during induced VT. For example, in the ENCORE-VT workflow, multimodal imaging can be combined with scar imaging and offline electrical mapping to define targets for ablation with SBRT. The plan is then developed in the radiation therapy planning system (TPS). On the day of treatment, the patient is immobilized to prevent patient motion, the treatment unit is aligned with the patient, and then the treatment is delivered using a linear accelerator.

[0061] Currently, electrophysiologists and radiation oncologists manually convert data into targets on the SBRT planning CT through time-consuming and imprecise iterative review of offline targeting data from TPS.

[0062] Currently, identification of the arrhythmic component of the ventricular scar in ENCORE is imperfect and requires manual creation of target volumes for ablation using offline visual estimation between various imaging modalities. Such visual estimation is highly user-dependent and requires significant consistency and skill from individual physicians. Therefore, results are likely to vary simply due to practitioner inconsistency and experience.

[0063] The disclosed technology satisfies the need in the art for automatic identification of one or more arrhythmias in a patient and user-independent objective treatment planning. Systems, methods, and computer-readable storage media for implementing multimodal techniques for identifying one or more arrhythmic components of ventricular tachycardia and generating a treatment plan are disclosed.

[0064] The methods disclosed herein define a radiotherapy target (treatment region) from multimodality imaging, where some or all of the modalities are used to define a single region of interest, which can then be combined in a weighted manner to generate a final target volume.

[0065] Any method for imaging arrhythmias can be used in combination with one or more other methods for imaging arrhythmias because the acquired data can be complementary and non-repeatable. In some embodiments, imaging can include electrophysiological features and / or anatomical features. The imaging method can be non-invasive. Non-invasive imaging methods can include but are not limited to CT, MRI, PET, SPECT, ECGI, and 12-lead EKG. However, multimodal anatomical (CT, MRI), functional (PET, SPECT), and electrophysiological (ECGI) images exist in a variety of file formats, and the presentation method often makes it difficult to integrate all information in one environment both technically and visually. Therefore, images or mappings from various sources can be converted to compatible formats for combination or comparison. In one example, two or more groups of images or mappings can be aligned with each other. Therefore, a non-invasive system for imaging can be used in combination to inform the non-invasive treatment of arrhythmias in subjects.

[0066] Anomalies or regions of interest on each image map can be defined manually or automatically. In at least one example, expert-defined targets can be used to train a model (e.g., a neural network) whose input is information from the multimodal images or the images themselves. The model can then be used on a new set of multimodal images to identify anomalies or infer target volumes (where the target volume is unknown).

[0067] 17 The segmentation model can be used to define the anatomical subregions of the ventricle. The segmentation model simplifies the problem from labeling thousands of individual voxels to labeling only 17 regions. In addition, the segmentation model overcomes the challenge of registering the individual multimodal images to each other.

[0068] The combined non-invasive mapping or segmentation model can then provide consistent, objective identification of targets for treatment.Consistency in initial target delineation is critical so that future identification of the "correct" or "best" target for patient prognosis assessment is not confounded by heterogeneity in targeting performed by different physicians.

[0069] In one example, a method for defining at least one arrhythmia target may include scoring each mapping separately for VT presence. In another example, a method for defining at least one arrhythmia target may include directly inferring a target probability.

[0070] Semi-automatic or automatic therapy selection based on patient and target attributes can provide unified treatment selection and allow for more reliable evaluation of results. In various aspects, procedure recommendations can include but are not limited to non-invasive ablation, invasive ablation, or a combination thereof. In one aspect, non-invasive ablation methods can include but are not limited to stereotactic body radiation therapy, stereotactic ablative radiation therapy, stereotactic radiosurgery, fractionated radiation therapy, hyperfractionated radiation therapy, high frequency / focused ultrasound, or laser.

[0071] Radiation may be delivered using x-rays / photons (typically using a linear accelerator), gamma rays (e.g., using a Co-60 unit), or charged particles (e.g., protons, carbon, helium, etc.). A variety of delivery systems exist, all of which have various methods of delivery. In one example, a non-invasive treatment may include SBRT ablation. Radiation therapy delivery may require optimization of the amount of immobilization, assessment, and motion in treatment planning and delivery, and the ability to create and deliver compact, precise dose distributions to maximize the dose to the target while minimizing the dose to healthy tissue, and image guidance may be required. Radiation therapy delivery may include the use of independent measurements of breathing and cardiac motion to define a patient-specific motion trajectory of the heart to precisely deliver treatment and reduce the size of the target volume, thereby mitigating toxicity associated with off-target radiation to normal tissue.

[0072] I. Target Recognition

[0073] like Figure 1A As shown, first a method for defining arrhythmia targets for ablation is described herein. Figure 1B When describing various embodiments, these changes will be described herein. Now, the present disclosure turns to Figure 1A .

[0074] Provided by way of example Figure 1A In addition, although the example method is shown in a specific order of blocks, it will be understood by those skilled in the art that Figure 1A The blocks shown therein may be performed in any order that achieves the technical advantages of the present disclosure, and may include fewer or more blocks than shown. Figure 1A Each block shown in the diagram represents one or more processes, methods, or subroutines performed in the example method.

[0075] Figure 1AA flow chart of an example method 100 for identifying an arrhythmia target is shown. In a non-limiting example, the arrhythmia target can be a ventricular arrhythmia segment, a ventricular contour, an atrial segment, and / or an atrial contour. In some examples, the method 100 can be performed in real time at an image planning system. In other examples, the method 100 can be performed planned at an image planning system for future use.

[0076] Method 100 may begin at block 105. At block 105, at least one processor may receive one or more input maps (e.g., maps of corresponding ventricles, corresponding atria, etc.). In some examples, the one or more input maps may be previously acquired historical input maps and / or newly required input maps. In some examples, the one or more input maps may be one or more images from a single patient. For example, the one or more input maps may be one or more electrophysiological maps, anatomical maps (e.g., fibrosis), functional maps (e.g., perfusion, motion, etc.), computer simulation maps (e.g., electrical and / or mechanical whole heart models), clinical maps, etc. In some examples, the electrophysiological maps may be, but are not limited to: EKG, 12-lead ECG of VT (e.g., exit site of VT / PVC, etc.), 12-lead ECG of sinus rhythm (SR) (e.g., Q waves for pre-infarction, PVC location comparison, etc.), ECGI of VT (e.g., exit site of ventricular tachycardia, reentrant propagation, diastolic potential, etc.) and / or ECGI of SR (e.g., slow conduction / block area, late potential area, etc.). In some examples, anatomical mapping can be, but is not limited to: CT scans (e.g., myocardial thinning (% wall thickness), calcifications, etc.) and / or MRI scans, including but not limited to gadolinium and / or pixel intensity maps (e.g., location of fibrosis (which segment; endocardium, medial myocardium, epithelium), thickness of fibrosis (% wall thickness), volume of fibrosis, heterogeneity of fibrosis, etc.). In some examples, functional mapping can be, but is not limited to: SPECT scans (e.g., viability, infarction / ischemia, etc.), PET scans (e.g., metabolism, inflammation, etc.), MRI scans (e.g., abnormal wall motion, etc.), Echo (e.g., abnormal wall motion, etc.), and / or heart and lung motion data (e.g., motion-sensitive sequences (e.g., 4D-CT and 4D-MRI) can construct an expected motion envelope for optimal targeting and delivery). In some examples, the clinical map can be, but is not limited to: demographics (e.g., age, sex, NYHA, CKD, lung, PVD, Charlson vs Seattle HF model, etc.), surgical history (e.g., cardiac surgery, etc.), knowledge of clinical VT (e.g., MMVT or multiple VTs), and / or prior electroanatomical mapping / ablation. In some examples, the clinical map is optional. In some examples, the clinical map is used as supplemental information (e.g., during risk analysis). In some examples, the electroanatomical map can be displayed separately from the created image map (e.g., output from method 100) for comparison purposes (e.g., the electroanatomical map will not be used for segment identification). In some examples, the computer simulated map can include in silico models.

[0077] At block 110, method 100 may determine or identify one or more abnormalities in one or more input maps. Abnormalities may be identified, for example, by scar location in MRI, abnormalities in PET / SPECT may be inactive areas, abnormalities in electrical maps may be where VT is generated, etc. In various examples, abnormalities may be defined by selecting abnormalities on each map, manually or automatically segmenting the map and manually selecting abnormalities, automatically contouring the map and manually selecting abnormalities, automatically contouring the map and automatically selecting abnormalities, or a combination thereof. In some examples, identification of abnormalities may be used, for example, to train a neural network by using supervised learning or reinforcement learning. For example, a physician may locate abnormalities in different image types. Abnormalities, as well as images and metadata for the images, may be used to teach a neural network how to automatically (or autonomously) locate abnormalities.

[0078] Defining or identifying one or more anomalies in one or more input mappings may include segmenting one or more input mappings. In some examples, one or more input mappings may be segmented using a segmentation model. The at least one processor may determine one or more anomalies in one or more arrhythmia target segments. In various examples, the input mapping may be divided into at least 2 segments, at least 4 segments, at least 6 segments, at least 8 segments, at least 10 segments, at least 12 segments, at least 14 segments, at least 16 segments, at least 18 segments, or at least 20 segments. In one example, an input mapping (e.g., a 17-lead ECG) may be divided into 17 segments. The segments may be of the same or similar size, of different sizes, or a combination thereof. After the input mapping has been segmented, the segments containing the anomalies may be determined.

[0079] In some examples, 3D models of 17 segments can be generated. In various examples, models can be generated for the left ventricle, right ventricle, and / or atrium. The 3D model can be generated using an elliptical cone, but any ventricle-like or atrial-like shape can be used. In at least one example, for each input mapping, a 3D model can be overlaid on the mapping using a deformable registration of the model to the left ventricle contour. In some examples, a free form (b-spline) registration can be used for alignment. Because the segmented model is symmetrical, anatomical landmarks such as the apex, anterior atrial sulcus, posterior atrial sulcus, and mitral valve plane can be identified and used as anchors to align the correct segment in the model to the correct anatomical location. In some examples, the 3D model can be located on the epicardial surface and extended volumetrically to the endocardial surface.

[0080] Defining or identifying one or more anomalies in one or more input maps may include outlining one or more input maps. In various examples, outlines on one or more input maps may be identified. The input map may be a 3D input map. For example, a user or physician may physically draw an outline on a 3D input map, or a processor may identify an outline containing an anomaly on a 3D input map. Anomalies may be automatically defined in an input map. In some examples, more than one anomaly may be identified. For example, an anomaly defined in an input map may be a single VT, multiple VTs, a single VT outlet, and / or multiple VT outlets.

[0081] In one example, a segmentation model can be used to manually identify anomalies on a first input map and manually draw contours on another input map (e.g., ECGI, MRI, CT, and / or PET). In another example, a segmentation model can be used to manually identify anomalies on a first input map and automatically draw contours on another input map (e.g., ECGI, MRI, CT, and / or PET). In yet another example, anomalies can be manually defined on a 12-lead ECG using a segmentation model, and anomalies can be manually identified on ECGI, MRI, CT, and / or PET maps. In one example, anomalies can be automatically defined on a 12-lead ECG, and anomalies can be manually identified on ECGI, MRI, CT, and / or PET maps. In some examples, an expert can review each input map separately and can score the likelihood of each segment's contribution to VT. These scores can be stored in a database. A target probability for each segment s can then be generated, defined as a weighted average (by weight w) on all input maps i for each patient p.

[0082] In yet another example, anomalies can be automatically defined on all input maps. In one example, anomalies can be automatically defined on one or more input maps by machine learning. For each input map, the image content in each segment can be extracted. Since each segment can be a conical section, it can be unfolded into a rectilinear 3D image volume. The process can generate Ns,p,i segments, where s is the number of segments per image (e.g., up to 17, but can be any number), i is the number of input maps per patient, and p is the number of patients in the database. Each segment can be labeled as VT, no VT, and cannot be evaluated. For two types of modeling problems, segments that cannot be evaluated can be ignored during modeling. Experts can also assign a weighting parameter w of a set of numbers between 0 and 1 to represent the confidence of each segment.

[0083] Any modeling approach can be used to learn image features that predict the location of anomalies (e.g., VT) on each input map. For example, a deep convolutional neural network can be used. Since the model only includes two categories (VT / no VT), a deep network may not be required. For example, a model can use approximately 5 layer blocks, including convolution blocks, convolution layers, and fully connected layers. In some examples, the network can be trained using training data to predict VT / no VT. In some examples, preprocessing including whitening and normalization can be used. In other examples, the model can be executed on new input maps. Segments can then be extracted and inference can be performed using the above network. In one example, a VT / no VT label can then be generated for each segment of each new input map.

[0084] In some examples, the probability of each abnormality (target probability) can be learned by adding additional layers to the neural network. These can be fully connected layers that attempt to indirectly learn the values ​​of the weights w. The labeled data can be the actual target probabilities. In one example, the neural network can contain additional layers that combine the individual segment categories into a weighted average and attempt to indirectly learn the weighting algorithm.

[0085] In other examples, a target probability for each segment s can be first generated, defined as a weighted average (by weight w) over all images i for each patient p. The individual segment images can be first normalized and then concatenated into 4D depth images of depth i, which can then be trained to predict continuous target probabilities. In various examples, the continuous target probabilities range from 0 to 1.

[0086] At block 115, method 100 may combine one or more input maps with the identified abnormalities. For example, the input maps may be combined by overlapping segmentation models, combining segmentation models and 3D geometry, or a combination thereof. In some examples, an overlap of contours may be selected from the abnormalities identified in the 3D geometry. In another example, an input map with a segmentation model and contours of at least one 3D input map may be combined. For example, a segmentation model from a 12-lead ECG may be co-registered with contours identified from 3D geometry (e.g., ECGI, MRI, CT, PET).

[0087] At block 120, method 100 may define targets for ablation based on overlap of data from the combined mapping. An arrhythmia target may be a segment, multiple segments, or a 3D volume. In a segmentation model, an arrhythmia target may be defined by identifying areas of greatest overlap or highest intensity. For 3D contours, an arrhythmia target may be defined by identifying overlap of contours in the 3D geometry. 3D anatomically defined targets (manually or automatically) may be smaller in area than segments and more patient specific. In other examples, a segmentation model may be co-registered with at least one 3D contour so that overlap between segments and contours may identify a target. In some examples, multiple independently defined targets for ablation or prioritized targets for ablation may be identified by looking at the overlap of individual VTs (for each VT / VT exit site mapped).

[0088] In some examples, one or more segments to be ablated can be identified using one or more types of ablative energy techniques (e.g., SBRT, photons, carbon ions, protons, helium, ultrasound, etc.). In some examples, one or more segments can be identified for a non-invasive arrhythmia guidance system. For non-invasive stereotactic cardiac ablation, the method can be integrated with, transmit data to, or exist in a radiation therapy planning system or a decision support module. In other examples, one or more segments can be determined for invasive ablation. For invasive catheter ablation, the method can be integrated with, transmit data to, or exist in a catheter ablation system.

[0089] At least one processor may determine one or more targets for ablation, for example, the target for ablation may include one or more target segments of arrhythmia. In various examples, the target for ablation may include at least 1 segment, at least 2 segments, at least 4 segments, at least 6 segments, at least 8 segments, at least 10 segments, at least 12 segments, at least 14 segments, at least 16 segments, at least 18 segments, or at least 20 segments. These segments may be of the same or similar size, different sizes, or a combination thereof. For example, the target for ablation may be determined and formatted as an image map (e.g., an image map input into a treatment planning system). In some examples, the target for ablation may be determined based on overlapping segments and risk conditions of the input map. In some examples, the overlap of segments from the input map may include an initial list of segments for ablation. The initial list may then traverse the risk conditions to determine the potential risk of ablation of the segments from the initial list. Based on the risk conditions, one or more segments for ablation may be determined.

[0090] In one example, in Figure 3A , 3B One or more segments for ablation are shown in and 3C. Figure 3A , 3Band 3C show a cross-section of a ventricular segment. For example, the ventricle can be divided into 17 different segments - 1. basalanterior; 2. basal anterioseptal; 3. basal inferoseptal; 4. basal inferior; 5. basal inferolateral; 6. basal anterolateral, 7. mid anterior; 8. mid anteroseptal; 9. mid inferoseptal; 10. mid inferior; 11. midinferolateral; 12. mid anterolateral; 13. apical anterior; 14. apical septal; 15. apical inferior; 16. apicallateral; and 17. apex.

[0091] Figure 3B Further shown are one or more segments for ablation based on the output from method 100. For example, segments: 1. basal anterior; 2. basal anterior septal; 6. basal anterolateral; and 7. mid anterior have been identified as segments for ablation. In one example, the above segments have been identified for a 61 year old patient with non-ischemic cardiomyopathy (NYHA class 4 heart failure symptoms (LVEF 37%; LVEDD 6.1 cm), repeated ICD shocks for VT despite treatment with amiodarone and mexiletine). Prior to those medications, sotalol had failed treatment for the patient. ICD interrogation showed at least two different VT cycle lengths. SBRT targeted four different VT circuits in the left ventricle at the basal anterior septum and anterolateral sides, who had undergone endocardial ablation one year prior.

[0092] exist Figure 3B In the example of , the corresponding input maps can be: electrical maps of two VT-ECG (2, 6), ECGI (1, 6); anatomical maps of MRI (1, 2, 6, 7) and CT (not sparse) and functional maps of Echo (global HK). Each map can identify one or more potential ablation segments. When these maps are combined, segments 1, 2, 6, and 7 are recommended for ablation. The decision support module output of this example is as follows Fig. 6A as shown in .

[0093] Figure 3CFurther shown are one or more segments for ablation based on the output from method 200. For example, segments: 13. anterior apex, 14. apical septal, 15. infraapical, 16. lateral apex, and 17. apex have been identified as high priority segments designated for ablation, and segments: 7. mid anterior, 8. mid anterior septal, 9. mid infraseptal, and 10. mid infraseptal have been identified as medium priority segments designated for ablation. In one example, a 75 year old patient with ischemic cardiomyopathy, NYHA class 4 heart failure symptoms, (LVEF 20%; LVEDD 6.9 cm), had repeated ICD shocks for VP despite treatment with amiodarone and mexiletine (both currently intolerant). Sotalol was excluded for advanced CKD. ICD interrogation primarily revealed one VT cycle length. No previous endocardial ablation (high risk for complications).

[0094] exist Figure 3C In the example of , the corresponding input maps can be electrical maps of ECG (14, 15, 16, 17), ECGI (SR LOP 15, 16, 17), ECGI (VT 15, 17); anatomical maps of MRI (7, 8, 9, 10, 13, 14, 16, 17), CT (7, 8, 9, 13, 14, 15, 16, 17); and functional maps of PET (7, 8, 9, 13, 14, 15, 17), MRI (2, 3, 4, 7, 8, 9, 10, 13, 14, 15, 17). Each map can identify one or more segments that may be ablated, and when these maps are combined, segments 13-17 can be high priority recommendations for ablation, while segments 7-10 can be medium priority recommendations for ablation. The decision support module output of this example is as follows Figure 6B shown.

[0095] In some examples, the arrhythmia target for ablation can be adjusted. For example, the arrhythmia target can be adjusted based on various parameters, including but not limited to risk, clinical data, demonstration data, prior knowledge of a specific patient, prior knowledge of previous patients, toxicity, efficacy, data quality, data importance, data reproducibility, physiology of the heart, size of the scar, number of VTs, type of cardiomyopathy (e.g., ischemic or non-ischemic), transmurality (e.g., thick or thin scars), location of the abnormality (e.g., certain segments may be more or less successful), age, sex, heart size, ejection fraction, heart thickness (e.g., weak or healthy heart), medications (e.g., antiarrhythmic drugs), comorbidities, and combinations thereof. In various examples, adjustment of the target can be performed manually or automatically. As other patients are treated, previous treatment methods can inform future treatment methods and can be used to adjust the target for the current patient (e.g., by training a neural network). In another example, the target can be adjusted based on the weights of the input mappings or the confidence scores of the combined mappings.

[0096] In some examples, the prior knowledge of previous patients can come from medical literature and / or data from individual patients entering the database. In various examples, the data can be manually entered into the database from providers, researchers, etc., can be automatically entered from discrete data automatically obtained from natural language processing (NLP) from electronic medical records (EMR), or can be automatically entered by the patient from the device. Non-limiting examples of patient devices that can automatically store data in the database include ICDs, phones, or wearable devices. In one example, the ICD can be registered in a reporting program to facilitate automatic entry into the database. In another example, the patient's phone can include an application that captures the results of the patient's report. In yet another example, the wearable device can include a smart watch, a heart rate monitor, or an activity tracker that can transmit any captured data to the database.

[0097] In some examples, the determined segments can be assigned priorities and / or probabilities. For example, based on a combination of the input mapping and the risk profile, the determined segments can be assigned high, medium, or low priorities. In some examples, all arrhythmia target segments can be assigned priorities (e.g., high, medium, or low). In other examples, segments assigned high and medium priorities are provided in the determination.

[0098] In some examples, the determined segments can be provided on an image map for use in ablation procedures. Image maps can be integrated with invasive and non-invasive delivery platforms. For example, an image map can be a volume for treatment planning (e.g., in a treatment planning system / software). In some examples, an image map can be used as an input to a treatment planning system (e.g., ablation can be performed). Image maps can provide a personalized treatment plan for a patient (e.g., not all patients require or provide all known multimodal data).

[0099] The weighting of the input mappings can occur at any time in the method 100. The weighting of the input mappings can improve the quality or accuracy of target identification by assigning higher weights to input mappings that have higher quality, clinical relevance, or importance relative to other mappings for the patient. For example, the weighting of the input mappings can occur before combining the mappings, after abnormal identification, after combination, after defining the target, or after adjusting the target. Each input mapping can be weighted according to one or more factors, and the multiple factors are, for example, scan quality (e.g., ICD artifacts on MRI, etc.), the number of input mappings (e.g., the number of modalities), clinical relevance (e.g., non-clinical CT induction, etc.), expert acceptance of a single technology, the importance of the data, or a combination thereof. In some examples, weights can be assigned between input mappings, for example, the relative strength between each input mapping finding, consistency between groups, etc. For example, when the number of input mapping modalities with high-quality scans is large and the degree of overlap between the input mappings is high, a higher weight can be determined. In another example, when the number of input mapping modalities with various qualities is small and the degree of overlap between the input mappings is low, a lower weight can be assigned.

[0100] In some examples, a confidence score for a combined input map or a determined arrhythmia target can be determined. In one example, the confidence score can be an evaluation of the combined input map. The confidence score can also include clinical data or any data used to adjust the target. For example, input map weights, weights between input maps, data consistency between different input maps, or the amount of overlap of results (e.g., segments for ablation, etc.) can be used to determine a confidence score. For example, a confidence score can be determined to quantify the quality and reproducibility of one or more segments or contours used for ablation. In some examples, the confidence score can increase with a higher degree of overlap between imaging groups (e.g., segments for ablation, etc.). In other examples, the confidence score may decrease if the input map is incomplete, the input map quality is poor, the number of VTs increases, the scar size is larger (e.g., the scar is larger than the ablation), etc.

[0101] Method 100 may further include determining a risk profile. In some examples, a risk profile may be determined based on historical input mappings, side effects, segment volume limitations, and / or previously recommended ablation segment results. For example, data associated with a patient who has previously undergone method 100 (or a similar method) and the results of any process performed using data or mappings obtained through method 100 may be stored to determine the patient's risk profile. In some examples, the risk profile may be used to determine potential risks or effects on the patient associated with a recommended ablation site. In some examples, the patient's demographic information may be used to determine the risk profile (e.g., age, health status, previous surgery, etc.). In some examples, historical input mappings, corresponding segment determinations, and ablation results (from determinations) may be used to determine which arrhythmia target segments can be treated without high risk and / or complications, and which segments are typically high risk and / or have the most complications. In at least one example, the risk profile is reported to the user / physician for therapy selection or toxicity counseling.

[0102] In one example, the initial target list may include segments 1, 2, 6, 7, 8, 12, and 13. The risk profile may determine that the segment volume limit should be equal to four (4) based on the history of the current patient and / or the history of similar patients. The resulting determined ablation segments may then be 1, 2, 6, and 7. In another example, the initial list may include segments 1, 2, 6, and 7. The risk profile may determine that these segments are at higher risk because, for example, coronary arteries are located in these segments. The segments may then be determined to be 1, 2, 6, and 7, with additional planning information for follow-up or monitoring of the coronary arteries after ablation. In another example, the initial list may include segments 1, 2, and 6. The risk profile may determine that based on the history of similar patients (e.g., similar history, surgery, etc.), it is advantageous for segments 1, 2, and 6 to be ablated, while segment 7 is also ablated. The resulting segments for ablation may then be 1, 2, 6, and 7. The examples provided herein are exemplary only and are non-limiting.

[0103] In some examples, the determined segments can be provided on the image map for ablation procedures. Image maps can be integrated with invasive and non-invasive delivery platforms. For example, the image map can be a volume for treatment planning (e.g., in a treatment planning system / software). In some examples, the image map can be used as an input to a treatment planning system (e.g., ablation can be performed). Image mapping can provide a personalized treatment plan for the patient (e.g., not all patients require or provide all known multimodal data). In one example, the target segment or 3D contour can be used to identify the planned target volume in the treatment planning software. In some examples, the planning target volume and / or segmentation / contour can be used to simulate treatment to ensure internal quality.

[0104] As an example, it is provided Figure 1B In addition, although the example method is shown in a specific order of blocks, it will be understood by those skilled in the art that Figure 1B The blocks shown therein may be performed in any order that achieves the technical advantages of the present disclosure, and may include fewer or more blocks than shown. Figure 1B Each block shown represents one or more processes, methods, or subroutines performed in the example method.

[0105] Figure 1B A flow chart of an exemplary method 101 for determining one or more segments for ablation is shown. In some examples, the one or more segments for ablation may be arrhythmia targets. In a non-limiting example, the arrhythmia target may be a ventricular arrhythmia segment or an atrial segment. In some examples, method 101 may be performed in real time at an image planning system. In other examples, method 101 may be performed planned in an image planning system for future use. Method 101 includes the steps of method 100, and further includes weighting one or more input mappings at block 135, determining a confidence score for the combined mapping at block 150, and determining a risk condition at block 155.

[0106] At block 125, at least one processor may receive one or more input maps (e.g., corresponding ventricles, corresponding atria, etc.). In some examples, the one or more input maps may be previously acquired historical input maps and / or newly required input maps. In some examples, the one or more input maps may be one or more images from a single patient. For example, the one or more input maps may be one or more electrophysiological maps, anatomical maps (e.g., fibrosis), functional maps (e.g., perfusion, motion, etc.), computer simulation maps (e.g., electrical and / or mechanical whole heart models), or clinical maps. In some examples, the electrophysiological maps may be, but are not limited to: EKG, 12-lead ECG of VT (e.g., exit site of VT / PVC, etc.), 12-lead ECG of sinus rhythm (SR) (e.g., Q waves of previous infarction, comparison of PVC location, etc.), ECGI of VT (e.g., exit site of VT, reentrant propagation, diastolic potential, etc.) and / or ECGI of SR (e.g., slow conduction / block area, late potential, etc.). In some examples, anatomical mapping can be, but is not limited to: CT scans (e.g., myocardial thinning (% wall thickness), calcifications, etc.) and / or MRI scans, including but not limited to gadolinium and / or pixel intensity maps (e.g., location of fibrosis (which part; endocardium, medial myocardium, epidermis), thickness of fibrosis (% wall thickness), volume of fibrosis, heterogeneity of fibrosis, etc.). In some examples, functional mapping can be, but is not limited to: SPECT scans (e.g., viability, infarction / ischemia, etc.), PET scans (e.g., metabolism, inflammation, etc.), MRI scans (e.g., abnormal wall motion, etc.), echo (e.g., abnormal wall motion, etc.), and / or cardiac and lung motion data (e.g., motion sensitive sequences (e.g., 4D-CT and 4D-MRI) to construct the expected motion envelope for optimal targeting and delivery). In some examples, the clinical map can be, but is not limited to: demographics (e.g., age, sex, NYHA, CKD, lungs, PVD, Charlson vs. Seattle HF models, etc.), surgical history (e.g., cardiac surgery, etc.), knowledge of clinical VT (e.g., MMVT or multiple VTs), and / or previous electroanatomical mapping / ablation. In some examples, the clinical map is optional. In some examples, the clinical map is used as supplemental information (e.g., during risk assessment). In some examples, the electroanatomical map can be displayed separately from the created image map (e.g., the output of method 200) for comparison (e.g., the electroanatomical map will not be used for segment determination). In some examples, the computer simulated map can include a model in the computer.

[0107] At block 130, method 101 may determine or identify one or more abnormalities in one or more input maps. For example, abnormalities may be identified by scar location, such as in MRI, abnormalities in PET / SPECT may be areas of inactivity, and abnormalities in electrical maps may be where VTs are generated. In various examples, abnormalities may be defined by selecting abnormalities on each map, manually or automatically segmenting maps and manually selecting abnormalities, automatically contouring maps and manually selecting abnormalities, automatically contouring maps and automatically selecting abnormalities, or a combination thereof. In some examples, identification of abnormalities may be used to train a neural network, for example, by using supervised learning or reinforcement learning. For example, a physician may locate abnormalities in different image types. Abnormalities, as well as images and metadata for the images, may be used to teach a neural network how to automatically (or autonomously) locate abnormalities.

[0108] At block 135, the input mappings may be weighted. The weighting of the input mappings may occur at any time in method 101. For example, the weighting of the input mappings may occur before combining the mappings, after abnormality identification, after combination, after defining the target, or after adjusting the target. Each input mapping may be weighted based on one or more factors, such as scan quality (e.g., ICD artifacts on MRI, etc.), the number of input mappings (e.g., number of modalities), clinical relevance (e.g., non-clinical CT induction, etc.), expert acceptance of individual techniques, data importance, etc. In some examples, weights may be assigned between input mappings, such as the relative strength between each input mapping finding, consistency between groups, etc. For example, when the number of input mapping modalities for high-quality scans is large and the degree of overlap between the input mappings is high, a higher weight may be determined. In another example, when the number of input mapping modalities of various qualities is small and the degree of overlap between the input mappings is low, a lower weight may be assigned.

[0109] At block 140, at least one processor may convert the received input map into a compatible format. For example, the input map (e.g., CT, MRI, PET, SPECT, ECGI, etc.) may exist in multiple file formats. Multiple file formats may be converted to a compatible format, for example, to enable association of overlapping data points. In some examples, the input map may require repositioning of the image. For example, a standard process for orienting and converting the input map based on the modality may be implemented. In some examples, non-image data (e.g., 12-V EKG, ECGI system, etc.) may be integrated with the input map. In some examples, the non-image data (e.g., 12-lead map) may be displayed to the user, and the user may interact with or click on the arrhythmia target. In other examples, the non-image data may be automatically integrated with the input map. In some examples, the input map may be converted into a point cloud.

[0110] At block 145, at least one processor may combine compatible format input maps. For example, the input maps may be combined to provide a more reliable illustration of a mapped image (e.g., a corresponding ventricle or atrium). In some examples, the input maps may overlap (e.g., based on common data points between common format maps). For example, each input map may be used to individually determine one or more arrhythmia target segments that should be ablated. The one or more arrhythmia target segments may then be combined together for each input map to determine that the determined segments overlap. For example, when each input map contains segments 1 and 2 (for ablation) and only one map contains segment 4, the determination may be segments 1 and 2. In some examples, the input maps may be converted to point clouds, and the point clouds may be combined.

[0111] At block 150, a confidence score for one or more input mappings may be determined. In some examples, block 150 may be located before block 140 or may be located anywhere in method 101. For example, a confidence score may be determined for each input mapping, a combined input mapping, or a determined arrhythmia target. In one example, the confidence score may be an evaluation of the combined input mapping. The confidence score may also contain clinical data or any data used to adjust the target. For example, input mapping weights, weights between input mappings, data consistency between different input mappings, or the amount of result overlap (e.g., segments used for ablation, etc.) may be used to determine a confidence score. For example, a confidence score may be determined to quantify the quality and reproducibility of one or more segments or contours used for ablation. In some examples, the confidence score may increase with a higher degree of overlap between imaging groups (e.g., segments used for ablation, etc.). In other examples, the confidence score may decrease if the input mapping is incomplete, the input mapping quality is poor, the number of VTs increases, the scar size is larger (e.g., the scar is larger than the ablation), etc.

[0112] At block 155, a risk profile can be determined. In some examples, a risk profile can be determined (e.g., at block 160) from historical input mappings, side effects, segment volume limitations, and / or results of previously recommended segments for ablation. For example, data associated with a patient who has previously undergone method 101 (or a similar method) and the results of any procedures performed using data or mappings obtained through method 101 can be stored to determine the patient's risk profile. In some examples, the patient's demographic information can be used to determine the risk profile (e.g., age, health status, previous surgeries, etc.). In some examples, historical input mappings, corresponding segment determinations, and ablation results (from the determination results) can be used to determine which arrhythmia target segments can be treated without high risk and / or complications, and which segments are typically high risk and / or have the most complications. In at least one example, the risk profile is reported to the user / physician for therapy selection or toxicity counseling.

[0113] In some examples, methods 100 and 101 can be used in a machine learning environment (e.g. Figure 5 ). Machine learning can be used to automatically adjust the target, such as Figure 2 and Figure 5 As shown. For example, known clinical data and arrhythmia targets defined by the user over time can be used to suggest subsequent targeting for new similar clinical and multimodal data scenarios, or to suggest therapy options for new similar clinical and multimodal data scenarios. For example, one or more inputs (at blocks 105, 125) and one or more abnormal determinations (at blocks 110, 130) and one or more arrhythmia targets (at blocks 120, 160) can be used as inputs to a neural network (e.g., an artificial neural network, a convolutional neural network, etc.). The inputs to the neural network may also include other learning techniques, such as supervised learning, reinforcement learning, etc. The neural network may continue to receive inputs (e.g., training data) over a period of time until the neural network is trained. The inputs to the neural network may also include data related to treatment success, such as successful outcomes, survival rates, side effects, etc. For example, when a new input is received (e.g., never received) and one or more segments for ablation can be generated without other information (e.g., historical information, etc.), the neural network is trained. A trained neural network can be considered "autonomous". In one example, a machine learning environment may be used to predict the success rate of an ablation procedure using past information, such as patient success rates associated with their confidence scores.

[0114] In some examples, the determined segments can be provided on the image map for ablation procedures. Image maps can be integrated with invasive and non-invasive delivery platforms. For example, the image map can be a volume for treatment planning (e.g., in a treatment planning system / software). In some examples, the image map can be used as an input to a treatment planning system (e.g., ablation can be performed). Image mapping can provide a personalized treatment plan for the patient (e.g., not all patients require or provide all known multimodal data). In one example, the target segment or 3D contour can be used to identify the planned target volume in the treatment planning software. In some examples, the planned target volume and / or segment / contour can be used to simulate treatment to ensure internal quality.

[0115] Image mapping can facilitate targeting of arrhythmias, thereby facilitating pre-treatment targeting for both non-invasive and invasive ablations. In some examples, one or more segments to be ablated can be determined using one or more types of ablative energy technologies (e.g., SBRT, photons, carbon ions, protons, helium, ultrasound, etc.). In some examples, one or more segments can be determined for a non-invasive arrhythmia guidance system. For non-invasive stereotactic cardiac ablation, the method can be integrated with a radiation therapy planning system, data can be transmitted to it, or it can be present in it. In some examples, supporting indicators can be provided by one or more segments for ablation (e.g., as shown in Tables 1 and 2). In other examples, one or more segments can be determined for non-invasive ablation. For invasive catheter ablation, the method can be integrated with a catheter ablation system, data can be transmitted to the catheter ablation system, or it can be present in the catheter ablation system. The method can be used to facilitate the selection of treatment for a patient between non-invasive and invasive ablation. For example, a well-defined small target can be identified that is easy to catheter ablate, or a larger epicardial target can be identified that is more suitable for non-invasive ablation.

[0116] In some examples, the method can be directly interfaced with a non-invasive radiotherapy ablation delivery platform to facilitate real-time or near-instantaneous targeting and delivery. For example, a patient presenting with a known arrhythmia (e.g., VT or atrial fibrillation) is transported to a non-invasive ablation unit, where prior clinical and imaging data are entered. The ECGI vest can then be placed on the patient, who then undergoes volumetric imaging on the unit. For example, this can be accomplished with a cone-beam CT with any of the new linear accelerators or a volumetric MRI with an MR-guided unit. The ECGI vest can communicate directly with the unit. Communication with the unit can display steady-state arrhythmia data (e.g., if the patient is actively experiencing an arrhythmia) or electrophysiological data indicating the location or origin of the arrhythmia in real time. Additionally, localization of the signal in 3D space may be another way to target the location of the target in real time during the patient's internal and external motion. Targeting of arrhythmias can be performed on or near the unit, and planning of non-invasive treatments can be performed on or near the unit. Non-invasive treatment can be performed shortly thereafter, with the target tracked and / or gated by two typical methods (e.g., on-board kV / MV, CBCT, MR, external / internal fiducial tracking) and supplemented by real-time electrophysiological data provided by ECGI.

[0117] Data representing outcomes for previously treated patients may be used to enhance the generated image mapping method. In at least one example, the patient's risk profile, segments selected for ablation, and / or treatment success may be stored in a database to inform future patient risk profiles. In some examples, the risk profile may be used to determine potential risks or impacts to the patient associated with a recommended ablation site. In one example, over time, the method may predict the continued utilization of invasive catheter ablation for left apical small endosteal lesions in elderly males. Subsequent outcome data showed high toxicity rates for this population, with patients treated with non-invasive ablation having similar outcomes showing low toxicity rates. While the method would initially recommend further catheter ablation based on prior treatment, the enhanced model would recommend non-invasive ablation. In various examples, the outcome data may come from local data inputs, multi-institutional data, or a combination of both.

[0118] In one example, method 100 or 101 may be deployed as a remote service, with centers potentially submitting their patients' data for analysis, targeting, and treatment recommendations without having to deploy their own local method.

[0119] II. Decision Support Module

[0120] A decision support module is also provided herein to provide the physician with an informed ablation plan and to provide metrics to support the success and risk of various treatment options and the opportunity to improve patient outcomes. After the target is determined, the determined target can be presented to the physician in the form of a decision support module. In addition, the decision support module may include one or more of the following descriptions: a description of scar pattern, a description of the size and location of scar burden, a suggested amount of ablation to achieve (e.g., full thickness ablation, partial thickness ablation, etc.), a confidence score for the combined mapping, a list of structures at risk, general recommendations, expected success of SBRT, expected success of alternative treatment modalities (e.g., catheter RF, arrhythmia drugs (e.g., amiodarone), etc.), or a combination thereof. In some examples, the decision support module may also include a risk profile. In other examples, the decision support module can include patient demographic information (e.g., age, gender, renal function, pulmonary function, etc.), cardiac history (e.g., NYHA class, LVEF, CA priority, cardiomyopathy type, previous use of antiarrhythmic medications, etc.), cardiac imaging (e.g., heart size, scar size, scar location, number of ventricular tachycardias, etc.), and / or consistency of results (e.g., confidence score, etc.).

[0121] The information in the decision support module can be generated based on the patient's suggested ablation targets and historical data from previous patients to inform the physician of all relevant information related to SBRT treatment of the patient's arrhythmia.

[0122] Provided by way of example Figure 2In addition, although the example method is shown in a specific order of blocks, it will be understood by those skilled in the art that Figure 2 The blocks shown therein may be performed in any order that achieves the technical advantages of the present disclosure, and may include fewer or more blocks than shown. Figure 2 Each block shown represents one or more processes, methods, or subroutines performed in the example method.

[0123] Figure 2 A flow chart of an exemplary method 200 is shown for generating a decision support module and determining one or more segments for ablation. In some examples, method 200 may be performed in real time at an image planning system. In other examples, method 200 may be performed in an image planning system in a planned manner for future use. Method 200 may include method 100 and / or method 101. At block 205, at least one processor may receive one or more input maps as described herein above for method 100 and method 101. At block 210, at least one processor may identify an abnormality in one or more input maps as described herein above for method 100 and method 101. In some examples, the abnormality is a myocardial matrix abnormality.

[0124] At block 215, at least one processor may create an informed ablation plan. Any available data may be used to create an ablation plan, for example, the ablation plan may include abnormal segments, selected segments for ablation, input mapping weights, confidence scores, and / or risk conditions. The data and / or ablation plan for each patient may be stored in a database to inform future patient treatments. The database may also be populated with patient demographic information.

[0125] At block 220, at least one processor may generate a decision support module to inform the physician of the ablation plan, risks, and other treatment information. In some examples, the decision support module may be referred to as a decision support tool. The decision support module may include a description of scar pattern, a description of scar burden size and location, a suggested target location, a suggested volume to achieve ablation (e.g., full thickness ablation, partial thickness ablation, etc.), a confidence score for the combined mapping, a list of structures at risk, general recommendations, expected success using SBRT, expected success using alternative treatment modalities (catheter RF, amiodarone, etc.), or a combination thereof. In one example, the decision support module may be displayed, printed, or provided to the physician in any form capable of providing information. Fig. 6A , 6B and 6C are example outputs of the decision support module.

[0126] At block 225, at least one processor may determine one or more segments for ablation, as described herein above for method 100 or method 101. After treating the patient, method 200 may also determine the success of the treatment. The database may then be updated with the success information (including any side effects or problems encountered). The success information may adjust treatment recommendations or the selection of ablation segments for future patients.

[0127] In some examples, method 200 can be used in a machine learning environment (e.g., Figure 5 As shown). Machine learning can be used to automatically adjust the target. Machine learning can be used to automatically adjust future ablation plans and decision support modules. For example, blocks 215, 220, 225 and / or 230 can be automatically adjusted using machine learning. Machine learning tools and predictive analysis can be integrated in method 200 to create a clinical decision support infrastructure, such as a decision support module. For example, known clinical data and arrhythmia targets defined by users over time can be used to suggest subsequent targeting for new similar clinical and multimodal data scenarios, or to suggest therapy options for new similar clinical and multimodal data scenarios. For example, one or more inputs (in block 205) and the determination of one or more segments (in block 225) can be used as inputs to a neural network (e.g., an artificial neural network, a convolutional neural network, etc.) or a learning algorithm. The input to the neural network can also include other learning techniques, such as supervised learning, reinforcement learning, etc. The neural network can continue to receive inputs (e.g., training data) over a period of time until the neural network is trained. The input to the neural network can also include data related to treatment success, such as successful outcomes, survival rates, side effects, etc. For example, a neural network is trained when new input is received (e.g., never received before) and can generate one or more segments and / or decision support modules for ablation without other information (e.g., historical information, etc.). A trained neural network can be considered "autonomous."

[0128] The decision support module may include a written description of the scar pattern. In some examples, the decision support module may also include a measure of scar burden, such as Fig. 6A , as shown in FIGS. 6B and 6C. Measures of scar burden may include the volume of scar (eg, in cc) and / or the percentage of ventricular myocardium covered by scar.

[0129] The decision support module may include a list of suggested ablation locations based on the identified targets for ablation, such as Fig. 6A, as shown in 6B and 6C. In some examples, the determined segments can be assigned priorities. For example, based on a combination of the input mapping and the risk status, the determined segments can be assigned high, medium, or low priority. In some examples, priorities (e.g., high, medium, or low) can be assigned to all arrhythmia target segments. In other examples, numerical priorities are assigned to some or all arrhythmia target segments, or the segments are listed in order of priority. In other examples, segments assigned high and medium priorities are provided in the determination. In one example, the decision support module can provide priorities for the determined target segments, or list the target segments for ablation in order of priority, such as Figure 6B shown.

[0130] In some examples, the decision support module can further include a determined segment provided on the image map for use in the ablation process. The image map can be integrated with invasive and non-invasive delivery platforms. For example, the image map can be a volume for treatment planning (e.g., in a treatment planning system / software). In some examples, the image map can be used as an input to a treatment planning system (e.g., ablation can be performed). The image map can provide a personalized treatment plan for the patient (e.g., not all patients require or provide all known multimodal data). In some examples, an image map with determined segments is provided in the decision support module.

[0131] The decision support module can facilitate targeting of arrhythmias to facilitate pre-treatment targeting for both non-invasive and invasive ablation. For non-invasive stereotactic cardiac ablation, the decision support module can be integrated with, transmit data to, or reside in a radiation therapy planning system. In other examples, one or more segments can be determined for invasive ablation. For invasive catheter ablation, the decision support module can be integrated with, transmit data to, or reside in a catheter ablation system. The decision support module can be used to facilitate selecting a treatment approach for a patient between non-invasive and invasive ablation. For example, a small, well-defined target can be identified that is easily amenable to catheter ablation, or a larger epicardial target can be identified that is more amenable to non-invasive ablation.

[0132] In some examples, the decision support module can interface directly into the non-invasive radiation therapy ablation delivery platform to facilitate real-time or near-instantaneous targeting and delivery. The decision support module can also include a recommended amount of ablation to achieve (e.g., full thickness ablation, partial thickness ablation, etc.), for example, Fig. 6A , 6B and 6C. Volumes may be expressed in cc and / or as a percentage of the ventricular myocardium.

[0133] The decision support module may also include a confidence score. In some examples, a confidence score for a combined input map and / or an identified arrhythmia target may be determined. In one example, the confidence score may be an evaluation of the combined input map. The confidence score may also include clinical data or any data for adjusting the target. For example, input map weights, weights between input maps, data consistency between different input maps, or the amount of overlap of results (e.g., segments used for ablation, etc.) may be used to determine a confidence score. For example, a confidence score may be determined to quantify the quality and reproducibility of one or more segments or contours used for ablation. In some examples, the confidence score may increase with a higher degree of overlap between imaging groups (e.g., segments used for ablation, etc.). In other examples, if the input map is incomplete, the input map quality is poor, the number of VTs increases, the scar size is larger (e.g., the scar is larger than the ablation), etc., the confidence score may decrease. In some examples, the confidence score can be displayed as a numerical value or percentage (e.g., X / 10, X / 100) and / or a categorical display (e.g., low, medium / medium, or high), such as Fig. 6A , 6B and 6C. The decision support module may also include comments or annotations as to why the confidence score is within a particular range, for example, Figure 6B shown.

[0134] The decision support module may also include a list of structures at risk that may be affected by SBRT treatment of the identified ablation locations. Fig. 6A , as shown in 6B and 6C, the decision support module can list organs at potential risk, or abbreviations of organs or tissues at potential risk. In some examples, the structures at risk may be informed by treatment of previous patients. For example, the ablation treatment results of previous patients may populate the database and be used to suggest structures at risk in future patients with the same or similar ablation locations or clinical data.

[0135] like Fig. 6A, as shown in the examples in 6B and 6C, the decision support module may include recommendations regarding additional procedures, follow-up / evaluation, further imaging, etc. The recommendations provided in the decision support module may be provided from a database populated with possible recommendations. The recommendations may be provided by a previous patient. For example, data representing the outcomes of previously treated patients may be used to enhance the generated image mapping method and the generated decision support module. In one example, over time, the method may predict the continued utilization of invasive catheter ablation for left apical endocardial lesions in elderly males. Subsequent outcome data showed a high toxicity rate for this population, while similar outcomes for patients treated with non-invasive ablation showed a low toxicity rate. Although the decision support module may initially recommend further catheter ablation based on multiple previous treatment approaches, the updated decision support module will recommend non-invasive ablation. In various examples, the outcome data may come from local data inputs, multi-institutional data, or a combination of both.

[0136] The decision support module may also include the expected success rate of using various treatments. In one example, the decision support module may include the expected success rate of using SBRT, catheter RF and / or amiodarone. In one example, the decision support module may also include risk conditions for therapy selection or toxicity consultation.

[0137] In one example, the method 200 and / or the decision support module can be deployed as a remote service, and the center may be able to submit data about the patient to the patient for analysis, target determination, and treatment recommendations without having to deploy its own local method instance. In other examples, the method 200 and / or the decision support module can be integrated into a non-invasive treatment system. In some examples, the decision support module can be displayed, printed, or provided to the physician in any form that can provide information.

[0138] Now, the present disclosure turns to Figure 4 Example system shown. Figure 4 An example of a computing system 400 is shown in which components of the system communicate with each other using connection 405. Connection 405 may be a physical connection via a bus, or a connection directly into processor 410, such as in a chipset or system-on-chip architecture. Connection 405 may also be a virtual connection, a network connection, or a logical connection.

[0139] In some examples, computing system 400 is a distributed system, where the functionality described in the present disclosure can be distributed across a data center, multiple data centers, a peer-to-peer network, an entire layer of a fog network, etc. In some examples, one or more of the described system components represent a number of such components, each performing some or all of the functionality described for that component. In some examples, these components can be physical or virtual devices.

[0140] The example system 400 includes at least one processing unit (CPU or processor) 410 and connections 405 that connect various system components including system memory 415, read only memory (ROM) 42, or random access memory (RAM) 425 to the processor 410. The computing system 400 may include a cache 412 of high-speed memory that is directly connected to, closely adjacent to, or integrated as part of the processor 410.

[0141] Processor 410 may include any general purpose processor and hardware or software services, such as services 432, 434, and 436 stored in storage device 430, configured to control processor 410 as well as a dedicated processor where software instructions are incorporated into the actual processor design. Processor 410 may essentially be a completely independent computing system including multiple cores or processors, buses, memory controllers, caches, etc. Multi-core processors may be symmetric or asymmetric.

[0142] To enable user interaction, the computing system 400 includes an input device 445, which can represent any number of input mechanisms, such as a microphone for voice, a touch-sensitive screen for gesture or graphic input, a keyboard, a mouse, motion input, voice, etc. The computing system 400 may also include an output device 435, which may be one or more of a variety of output mechanisms known to those skilled in the art. In some cases, a multimodal system may enable a user to provide multiple types of input / output to communicate with the computing system 400. The computing system 400 may include a communication interface 440, which may generally govern and manage user input and system output, and may also connect the computing system 400 to other nodes in a network. There is no restriction on operation on any particular hardware device, so the basic functionality here can be easily replaced with improved hardware or firmware devices as they are developed.

[0143] The storage device 430 may be a non-volatile storage device and may be a hard disk or other type of computer-readable medium that can store data accessible by a computer, such as a magnetic tape, a flash memory card, a solid-state storage device, a digital versatile disk, a cartridge, a battery-backed random access memory (RAM), a read-only memory (ROM), and / or some combination of these devices.

[0144] Storage device 430 may include software services, servers, services, etc., which cause the system to perform functions when code defining such software is executed by processor 410. In some examples, hardware services that perform a particular function may include software components stored in a computer-readable medium and software components associated with hardware components required to perform the function (e.g., processor 410, connection 405, output device 435, etc.).

[0145] Now, the present disclosure turns to Figure 5 , which shows an example machine learning environment 500. The machine learning environment can be implemented on one or more computing devices 502A-N (e.g., cloud computing servers, virtual services, distributed computing, one or more servers, etc.). The computing device 502 may contain training data 504 (e.g., one or more databases or data storage devices, including cloud-based storage, storage networks, local storage, etc.). The training data 504 of the computing device 502 can be populated by one or more data sources 506 (e.g., data source 1, data source 2, data source n, etc.) over a period of time (e.g., t, t+1, t+n, etc.). In some examples, the training data 504 can be labeled as data (e.g., one or more labels associated with the data). For example, the training data can be one or more images, and a label (e.g., VT or no VT) can be associated with each image. In some examples, the labeled data can be an actual target probability (e.g., each segment as a weighted average of all received images for each patent). The computing device 502 can continue to receive data from one or more data sources 506 until a neural network 508 (e.g., a convolutional neural network, a deep convolutional neural network, an artificial neural network, a learning algorithm, etc.) of the computing device 502 is trained (e.g., has sufficient unbiased data to respond to new incoming data requests and provide autonomous or near-autonomous recommended action plans and / or actually provide input to execute action plans). In some examples, the neural network can be a convolutional neural network, for example, using a five-layer block, including a convolutional block, a convolutional layer, and a fully connected layer. Other layers can be added to the neural network, for example, to learn target probabilities (e.g., the probability of one or more arrhythmia targets, etc.). For example, the additional layer can combine the segments into a weighted average (e.g., indirectly learn a weighted average). Although an exemplary neural network is implemented, the neural network 508 can be one or more of various types of neural networks and is not specifically limited to one type of neural network or learning algorithm.

[0146] In other examples, target probabilities can be generated (e.g., each segment as a weighted average of images provided by the patient receiving data). In these cases, the individual segments can be normalized and concatenated into a 4D depth image at a specific depth. A neural network can be trained based on the depth of the images, for example, to predict a continuous target probability (e.g., ranging from 0 to 1). The training data may require an equal number of images per patient, so that if there are missing images, substitute images can be generated based on existing images (e.g., to enable sufficient training data without biasing the training data).

[0147] In some examples, although not shown here, the training data 504 can be checked for bias, for example, by checking the data source 506 (and corresponding user input) relative to previously known unbiased data. Other techniques for checking data bias are also implemented. The data source can be any data source (e.g., MRI, CT, 3D modeling, etc.) described above in the present disclosure for providing input images.

[0148] The computing device 502 can receive user (e.g., physician) input 510 related to a data source. The user input 510 and the data source 506 can be related in time (e.g., by time t, t+1, t+n, etc.). That is, the user input 510 and the data source 506 can be synchronized because the user input 510 corresponds to and supplements the data source 506 in a supervised learning or reinforcement learning manner. For example, the data source 506 can provide an MRI image at time t, and the corresponding user input 510 can input an ablation of the MRI image at time t. Although the time t may actually be different in the real world, they are synchronized in time relative to the data provided to the training data. In other examples, as described herein, the user input can classify a segment as VT or no VT.

[0149] The training data 504 can be used to train a neural network 508 or a learning algorithm (e.g., a convolutional neural network, an artificial neural network, etc.). The neural network 508 can be trained to automatically (e.g., autonomously) determine what the user input 510 is based solely on the received data 512 (e.g., imaging data, etc.) over a period of time. For example, by receiving a plurality of unbiased data and / or corresponding user inputs over a sufficiently long period of time, the neural network will be able to determine what the user input would be if only the data were provided. For example, a trained neural network 508 will be able to receive an MRI image (e.g., 512) and determine, based on the MRI image, an ablation that a physician would manually identify (and would provide as user input 510 during training). In some examples, this can be based on a label associated with the data as described above. The output from the trained neural network can be provided to a treatment unit 514 for treating a patient. In some examples, the output from the trained neural network can be directly input into the treatment unit to perform surgery on the patient.

[0150] The trained neural network system 516 may include the trained neural network 508, the received data 512, and the treatment unit 514. The received data 512 may be information related to the patient as described above. The received data 512 may be used as an input to the trained neural network 508. The trained neural network 508 may then, based on the received data 512, label the received data (e.g., VT or no VT) and / or determine a recommended course of action for treating the patient based on how the neural network was trained (as described above). The recommended course of action or the output of the trained neural network 508 may be used as an input to the treatment unit 514 (e.g., to perform a procedure on the patient corresponding to the received data 512). In other cases, the output of the trained neural network may be provided in a human-readable form, for example, for review by a physician to determine a course of action (e.g., less avoidance, validation, etc.).

[0151] For clarity of explanation, in some cases, the present technology may be represented as including various functional blocks including functional blocks including devices, device components, steps or routines in methods embodied in software, or a combination of hardware and software.

[0152] In some embodiments, computer readable storage devices, media and memories may include cable or wireless signals of bit streams etc. However, when referred to, non-transitory computer readable storage media explicitly excludes media such as energy, carrier signals, electromagnetic waves and signals themselves.

[0153] The methods according to the above examples can be implemented using computer executable instructions stored in or available from a computer readable medium. Such instructions may include, for example, instructions and data that cause or configure a general-purpose computer, a special-purpose computer, or a special-purpose processing device to perform certain functions or a set of functions. Some of the computer resources used may be accessed over a network. Computer executable instructions may be, for example, binary files, intermediate format instructions, such as assembly language, firmware, or source code. Examples of computer readable media that can be used to store instructions, information used, and / or information created during the methods according to the examples include disks or optical disks, flash memory, USB devices equipped with non-volatile memory, network storage devices, and the like.

[0154] Devices implementing methods according to these disclosures may include hardware, firmware, and / or software, and may take any of a variety of form factors. Typical examples of such form factors include laptops, smartphones, small personal computers, personal digital assistants, rack-mounted devices, stand-alone devices, etc. The functionality described herein may also be embodied in peripheral devices or add-in cards. By way of further example, such functionality may also be implemented on circuit boards between different chips or different processes executed in a single chip.

[0155] Instructions, media for communicating such instructions, computing resources for executing them, and other structures for supporting such computing resources are means for providing the functions described in these disclosures.

[0156] Although various examples and other information are used to explain aspects within the scope of the appended claims, limitations on the claims should not be implied based on specific features or arrangements in such examples, as a person of ordinary skill in the art will be able to use these examples to derive a variety of implementations. In addition, although some subject matter may have been described in language specific to examples of structural features and / or method steps, it should be understood that the subject matter defined in the appended claims is not necessarily limited to these described features or actions. For example, such functions may be distributed differently or performed in components other than those identified herein. Instead, the described features and steps are disclosed as examples of components of systems and methods within the scope of the appended claims. In addition, claim language that refers to "at least one" in a group indicates that one member of the group or multiple members of the group satisfy the claim.

[0157] Example

[0158] Example 1:

[0159] The treatment described in this example was performed based on the patient's clinical situation, without specific testing of the study hypotheses. All patients received a detailed explanation of the risks of treatment from the treating electrophysiologist and radiation oncologist; all patients provided written informed consent for treatment. Institutional review board approval for the use of electrocardiographic imaging had been previously provided. At the time of the study, the SBRT device had received 510(k) premarket approval from the U.S. Food and Drug Administration (FDA), but its use in the patients reported here was considered nonroutine clinical use; this information was communicated to the patients included in this study.

[0160] Patients with structural heart disease, implantable cardioverter-defibrillator (ICD) placement, and refractory ventricular tachycardia who had limited conventional treatment options for noninvasive cardiac ablation of ventricular tachycardia (noninvasive radioablation) were evaluated on a case-by-case basis. Patients were considered for noninvasive radioablation if they had at least 3 ICD-treated ventricular tachycardias in the previous 3 months despite treatment with at least 2 antiarrhythmic medications and had undergone at least one catheter ablation procedure (or had a contraindication to catheter ablation). Evaluation of patients for cardiac transplantation was encouraged according to the institutional standard of care, but eligibility for transplantation was not an absolute condition for consideration. Patients with placed left ventricular assist devices were not evaluated for study inclusion.

[0161] The procedural workflow of non-invasive radioablation is as follows Figure 7 As shown. Prior to treatment, patients underwent noninvasive electrocardiographic imaging during induced ventricular tachycardia to accurately map the ventricular tachycardia circuit. For electrocardiographic imaging, patients wore a 256-electrode vest (BioSemi), and a chest CT scan was performed. The patient was then taken to the electrophysiology laboratory and noninvasive programmed stimulation was performed by inducing ventricular tachycardia using an indwelling ICD. Data for the electrocardiographic imaging map were obtained, and the ventricular tachycardia was terminated with a brief overdrive pacing maneuver using the ICD. An electrocardiographic imaging map was created to identify the earliest site of electrical activation in the ventricular tachycardia (the “exit site”).

[0162] When clinically available, the patient was randomly assigned to a 4-year-old male using standard techniques ( Figure 7 ), using additional cardiac imaging to identify areas of anatomical scarring, either by resting single-photon emission CT (SPECT) or contrast-enhanced cardiac MRI. Electrical information from electrocardiographic imaging and information from the anatomical scarring are combined to establish a volumetric target for ablation that targets the area of ​​the first 10 milliseconds of ventricular tachycardia (exit site) and the entire myocardial thickness of the associated ventricular scar.

[0163] Additionally, before treatment, patients underwent a planning CT scan, which included immobilization of the entire body from chest to legs using a vacuum-assisted device (BodyFIX, Elekta) and a breathing-related CT scan (4D CT) to assess the sum of cardiac and pulmonary motion. The final target (planning target volume) was developed by expanding the target (defined above) to account for motion, setup uncertainty, and delivery uncertainty.

[0164] A total dose of 25 Gy in a single fraction was prescribed to the planning target volume in order to achieve maximum dose coverage while avoiding doses to surrounding organs (including the esophagus, stomach, lungs, and spinal cord) exceeding calculated dose limits. Prior to delivery, all plans were accepted on a calibrated phantom and passed standard in-house physical quality assurance.

[0165] SBRT was delivered using a linear accelerator (TrueBeam, Varian Medical Systems) equipped with image-guided radiation therapy that uses cone-beam CT to acquire chest images that can be directly registered to the planning CT. This procedure results in precise alignment of the heart and target volume without the need for invasive placement of fiducial markers. During treatment, patients were placed in a custom-made immobilization device that was aligned with the cone-beam CT, and alignment was verified by fluoroscopy. All patients did not use any additional imaging during treatment and did not receive sedation or anesthesia.

[0166] After treatment, patients who underwent VT ablation were followed according to standard of care. All ICDs were reprogrammed with a monitoring-only zone of 100 bpm to assess for ventricular tachycardia bradycardia. Patients were enrolled in a remote monitoring program for the device to enhance rapid recognition and interpretation of any arrhythmias detected by the ICD. Patients visited the clinic for ICD interrogations every 2 weeks for 2 months after treatment, monthly for 4 months, and then 1 year after treatment.

[0167] At each visit, attempts were made to wean patients off antiarrhythmic medications to mitigate the known short- and long-term toxic effects of these drugs. If no further ventricular arrhythmias were identified, antiarrhythmic medications were reduced or discontinued, with the goal of discontinuing all antiarrhythmic medications after the 6-week visit. Patients continued to receive medications (including beta-blockers) for heart failure before and after treatment.

[0168] Episodes of ventricular tachycardia were calculated as the sum of appropriate ICD shocks, appropriate ICD antitachycardia pacing, and sustained (>30 seconds) untreated ventricular tachycardia. The attending electrophysiologist adjudicated all ICD interrogations. Patients underwent echocardiography at baseline and 1, 6, and 12 months after treatment to assess for cardiac adverse events. Patients also underwent chest CT at baseline and at 3 and 12 months to assess for thoracic adverse events per the routine SBRT thoracic standard of care.

[0169] From April to November 2015, nine patients were evaluated for noninvasive radioablation; five of them were treated. Of the four patients who were not treated, two declined to participate (one chose to enter hospice care and died 1 week later from complications of ventricular tachycardia, and the other chose to undergo an invasive procedure for ventricular tachycardia), one patient died of progressive cardiogenic shock before treatment, and the other patient had a left ventricular assist device implanted and had recurrent ventricular tachycardia after surgery.

[0170] Table 1 lists the demographic and clinical data of each patient.

[0171]

[0172] The mean age of the 5 treated patients was 66 years (range, 60 to 83). The mean number of ventricular tachycardia episodes per patient in the 3 months before treatment was 1315 (range, 5 to 4312). All patients were taking two antiarrhythmic medications at the time of evaluation. Previous invasive catheter ablation procedures had failed in 3 patients. Two patients had contraindications to invasive catheter ablation: one (patient 2) had a new mechanical prosthetic mitral valve and evidence of epicardial tachycardia, and the other (patient 5) was considered too fragile for any invasive procedure. All 5 patients had New York Heart Association class III or IV heart failure symptoms. The mean left ventricular ejection fraction was 23% (range, 15 to 37).

[0173] All patients underwent noninvasive electrocardiographic imaging to reflect their ventricular tachycardia. Inducible ventricular tachycardia occurred in four patients (mean number of loops, 3; range, 1 to 6). Electrocardiographic imaging was performed during all induced episodes of ventricular tachycardia. In one patient (patient 3), ventricular tachycardia could not be induced and electrocardiographic imaging was not available, so the creation of a volume target was guided by the use of a 12-lead electrocardiogram and the results of previous invasive catheter mapping. Treatment characteristics are provided in Table 1. The ablation target volume ranged from 17 to 81 ml (mean, 49). The time for on-table treatment ranged from 11 to 18 minutes (mean, 14 minutes).

[0174] At a mean follow-up of 12 months, the burden of ventricular tachycardia after treatment was significantly reduced ( Table 1 and Fig.9A -9C). In the 15 patient-months before treatment, a total of 6577 episodes of ventricular tachycardia occurred. In the 6 weeks after ablation (the "observation period," because post-ablation inflammation may cause arrhythmias), 680 episodes of ventricular tachycardia occurred. After the 6-week observation period, 4 episodes of ventricular tachycardia occurred in the next 46 patient-months, a 99.9% reduction from baseline.

[0175] Fig. 8A The number of ventricular tachycardia episodes per patient per month is shown. The burden of ventricular tachycardia was reduced in each patient. Of the 4 patients who survived for 12 months, 3 did not receive any antiarrhythmic medication. Patient 3 restarted amiodarone treatment 9 months after the first tachycardia pacing treatment. Patient 4 underwent an additional invasive catheter ablation at week 4 after treatment due to incomplete ventricular tachycardia arrest and has not had any episodes since then. The number of ICD shocks observed (55 ICD shocks before treatment vs 1 shock after treatment) ( Figure 8B ) and ICD anti-tachycardia pacing (6577 times before treatment vs 3 times after treatment) were improved ( Figure 8C ).

[0176] No complications occurred during treatment or during the first hospitalization. Three patients reported fatigue the day after treatment. No acute exacerbation of heart failure occurred immediately after treatment. Patients were discharged from the hospital 1 to 3 days after treatment.

[0177] No adverse effects on ICD system performance, lead thresholds, or lead impedance were observed at any time after treatment. Serial echocardiograms showed no pericardial effusion. At the last follow-up, the mean change in left ventricular ejection fraction was an absolute increase of 6 points (range, 0.2 to 22) ( Fig.9A No pulmonary symptoms occurred after treatment. A CT scan performed at 3 months showed inflammatory changes in adjacent lung tissue consistent with typical chest SBRT, which had almost completely resolved at 12 months ( Fig. 9B At 12 months, there was no chest pain or significant myocardial or coronary artery changes on CT in the treated target area.

[0178] During the follow-up period, after amiodarone was discontinued, sinus node function recovered in three patients. The heart rate of sinus rhythm occasionally exceeded the detection cutoff rate (100 bpm) set by the ICD. This situation required ICD reprogramming to avoid inappropriate treatment.

[0179] One patient (Patient 5) had a fatal stroke 3 weeks after treatment. This 83-year-old woman had a history of atrial fibrillation, severe cardiomyopathy, and other risk factors for stroke. Oral anticoagulation for stroke prevention was not recommended because of the risk of frailty-related bleeding. Her ventricular tachycardia burden was reduced by 82% 3 weeks after treatment (from 1777 antitachycardia pacings in the month before treatment to 322 after treatment). Her left ventricular ejection fraction increased from 15% to 30%. No intracardiac thrombi were seen on echocardiography or pathological evaluation. It is unclear whether the stroke was related to SBRT or to a preexisting medical condition that placed her at high risk for stroke.

[0180] Written consent for postmortem cardiac pathology evaluation was obtained from patient 5. Distinct rectal vessels ( Fig. 10A This pattern has been described as part of acute vascular injury and is often observed within the first few weeks after radiation exposure. In this setting, the pattern of injury is often accompanied by endothelial cell swelling, vacuum formation, and perivascular tissue edema. However, in this patient, the endothelium of these vessels appeared normal and nonreactive, with no signs of acute vasculitis or tissue edema ( Fig. 10B ). No signs of acute myocardial necrosis, hemorrhage, or acute inflammation were observed. The relative contributions of distal myocardial infarction and acute cardiac SBRT to dense scar formation are unclear.

[0181] Example 2:

[0182] The ENCORE-VT trial was a prospective, single-arm, Phase I / II trial conducted at a single center. The local Institutional Review Board (IRB) approved the study without an Investigational Device Exemption (IDE) based on an interpretation of the indications and risks for the use of a linear accelerator and electrocardiographic imaging technology. All participants provided informed consent and independent data, and the Safety Monitoring Board reviewed the data semi-annually and provided guidance for continued study. Following study enrollment, the FDA reviewed the protocol and classified the study as a significant risk study requiring approval by an IDE. The investigators, IRB, and FDA worked together to ensure that appropriate human subject protections were in place.

[0183] Eligible patients were ≥18 years of age and had a) ≥3 sustained monomorphic VT, or b) cardiomyopathy (left ventricular ejection fraction (LVEF) <50%) associated with monomorphic PVCs (PVC>20%), and required ≥1 antiarrhythmic medication and failed ≥1 catheter ablation (or had a contraindication to catheter ablation). Patients were unlikely to have received radiation therapy to the intended treatment area in the past. Patients were considered ineligible if they had heart failure due to positive inotropy and / or left ventricular assist devices, or were considered unlikely to survive 12 months without VT. Patients were also ineligible if they had polymorphic VT or ventricular fibrillation on non-invasive programmed stimulation (NIPS) testing, more than 3 different clinical VT morphologies, or more than 5 inducible VT morphologies.

[0184] Targeting

[0185] Protocol-specified targeted baseline studies included cardiac CT, cardiac MRI, PET / CT, 12-lead ECG, and ECGI obtained during inducible VT during NIPS testing. The NIPS and ECGI methods have been described previously. A combination of imaging studies and electrophysiologic mapping was used to guide SBRT in each patient, with the rationale being to target all areas of the ventricular scar proximal to the VT exit site with associated circuits.

[0186] Treatment and follow-up

[0187] Patients received a single dose of 25 Gy delivered to the above-mentioned arrhythmogenic targets using SBRT. After treatment completion, a prespecified ICD programming plan containing a detection zone at least 20 milliseconds slower than the slowest clinical or induced VT was implemented in all patients. The ICD was remotely monitored as part of clinical care. Oral anticoagulation was prescribed during the first month after treatment if not contraindicated. Study visits occurred at day 3, 2, 4, and 6 weeks, 6, and 12 months, and annually thereafter. Adverse events were continuously assessed, and ICD interrogation was performed at each study visit. A 12-lead ECG was obtained at day 3, week 6, and 3, 6, and 12 months. For patients with PVCs, 24-hour Holter monitoring was performed at week 6, 3, 6, and 12 months, and chest CT and ECGI (without NIPS) were performed at 3 and 12 months.

[0188] Outcome indicators and statistical analysis

[0189] The ENCORE-VT trial was designed with co-primary endpoints of (1) safety and (2) efficacy. The primary safety endpoint (using Common Terminology Criteria for Adverse Events (CTCAE, v4.0) that was considered treatment-related (possibly, probably, or definitely related to study treatment)) was defined as the rate of serious adverse events (SAEs) ≤ 90 days. SAEs were defined as any grade 3 toxicity requiring hospitalization, or any grade 4-5 toxicity. An early stopping rule was set to terminate protocol enrollment if 5 or more of the first 10 patients had an SAE.

[0190] The primary efficacy endpoint was defined as the number of subjects with a reduction in the number of ICD treatments for VT or the number of 24-hour PVC burden, comparing pre- and post-SBRT 6 months (with a 6-week observation period after treatment to measure efficacy). ICD treatments included ICD shocks and anti-tachycardia pacing (ATP). Other selected pre-specified secondary objectives described herein included more stringent efficacy endpoints (50% reduction, 95% reduction), patient-derived endpoints (reduction in shocks in patients with VT, improvement in cardiac function with PVCs), overall survival, late (>90 days to 1 year) adverse events, and quality of life (measured by the SF-36 questionnaire).

[0191] The study was designed to balance a high probability of safety with a preliminary assessment of efficacy. The study population was expected to be at higher risk, given that all patients had failed prior therapy to stop VT, and it was assumed that an SAE rate as high as 20% and an efficacy as low as 40% would be clinically acceptable. Using a one-sided, one-sample proportion test, 19 patients provided 75.4% power to determine that the SAE rate was indeed no higher than 20% (range, 5-20%, alpha = 0.0829) and 81.5% power to determine that the efficacy was no less than 40% (range, 40-65%, alpha = 0.0885).

[0192] Continuous variables are reported as medians and ranges. The number of VT events, ICD shocks, and ATP events between the baseline and 6-month time points were compared using the Wilcoxon signed-rank test. McNemar's paired test was used to assess changes in the proportion of antiarrhythmic use. For quality of life analyses, the means of the short form-36 measures at baseline, 6 weeks, and 6 months were compared using repeated measures analysis of variance (ANOVA) with the Greenhouse-Geisser correction. Median follow-up was calculated from the date of treatment to the date of last planned follow-up or death. Kaplan-Meier analysis was used to estimate survival function. All statistics were performed in IBM SPSS Statistics for Windows, version 24.0, Armonk, NY, 2016.

[0193] Patients and treatment

[0194] Table 2 lists the demographic data of this population, which was characterized by a median age of 66 years, 89.5% males (n=17), ischemic cardiomyopathy (n=11, 57.9%), median LVEF of 25% (range 15-58), and New York Heart Association (NYHA) class III / IV (73.7%).

[0195] Table 2

[0196]

[0197]

[0198] The median number of previous catheter ablations before enrollment was 1 (range, 0-4). Three patients did not undergo catheter ablation for the following reasons: mechanical AVR and MVR (n=1), severely reduced ejection fraction with medical comorbidities that hampered hemodynamic support (n=1), and the presence of active left ventricular thrombus (n=1). The median follow-up was 13 months, and no patient was lost to follow-up.

[0199] Patients enrolled had ICD-treated VT (n=17) or PVC-related cardiomyopathy (n=2). Of the 17 patients with ICD-treated VT, 10 were considered to have VT storm (3 or more VT episodes within 24 hours), 2 of whom were in sustained VT at the time of treatment. More than half (n=11, 57.9%) of the patients were taking >1 antiarrhythmic medication and were taking ≥300 mg amiodarone per day at the time of treatment.

[0200] Target and treatment characteristics are shown in Table 3. Ten patients were excluded from cardiac MRI because of abandoned ICD leads. All patients underwent CT, nuclear imaging, and VT induction, followed by 12-lead ECG and ECGI images. Patients had 2 (median) VT inductions (range, 1-5).

[0201] Table 3

[0202]

[0203] The median total target volume was 25.4 cc (range: 6.4-88.6). Patients with PVC had the smallest GTV (6.4 cc, 11.5 cc). Taking into account motion and conservative margins for setup and delivery, the median planned target volume was 98.9 cc (range, 60.9-298.8). Most patients (n=17, 89.5%) were treated with volumetric modulated arc therapy technology. The median beam delivery time was 15.3 minutes (5.4-32.3).

[0204] Security

[0205] No acute toxicities were observed during or after SBRT. No adverse reactions to the ICD were observed during or after SBRT. Two patients (10.5%) experienced SAEs that were grade 3 treatment-related (possibly, probably, or definitely related). One patient was hospitalized for worsening heart failure (grade 3) 65 days after treatment and was conservatively rated as possible. Another patient was hospitalized for pericarditis on day 80 but improved with prednisone and was rated as probable. No grade 4 toxicities were recorded. Another patient died 17 days after treatment in a nursing home from a cause that was recorded as an unrelated incident, and this patient did not contribute to the primary efficacy endpoint at 6 months. Early stopping rules were not met and the DSMB recommended completion of the study.

[0206] Adverse events probably or definitely related to treatment were generally grade 1-2 (8 events in 4 patients, (22%)). Transient grade 1 fatigue and hypotension were common. Three patients required adjustment of antihypertensive medications due to hypotension within 2 weeks. Other common grade 1-2 adverse events included dizziness, dyspnea, and nausea. Two patients (11.1%) developed grade 2 radiation pneumonitis, which resolved with steroids. Pericardial effusions were documented in 6 of 5 patients (28%); 3 were asymptomatic, 1 resolved with medical management, and 2 were of higher grade (1 possibly related grade 3 after epicardial puncture, 1 unrelated grade 4). The 4 patients with symptomatic pericarditis or pneumonitis were treated with prednisone 1 mg / kg daily (maximum 560 mg), tapered by 10-20 mg weekly depending on symptoms. Six patients (33%) were hospitalized for heart failure at any time during follow-up at least once (6).

[0207] effect

[0208] Of the 18 patients who survived for 6 months, 17 / 18 (94%) patients achieved the primary efficacy endpoint of reduction in VT episodes or reduction in PVC burden. Fig.11 The frequency of VT episodes and 24-hour PVC burden are shown for all 18 patients before and after noninvasive cardiac radioablation.

[0209] For the 16 evaluable patients with VT treated with ICD, there were a total of 1778 VT episodes in the 6 months before treatment. During the 6-week observation period, there were a total of 149. For the next 4.5 months, there were 111 VT episodes (94% reduction in total VT episodes). The median number of VT episodes decreased from 6 months before ablation (119, range 4-292) to 6 months after ablation (3, range 0-31, p<0.001). The median number of ICD shocks was significantly reduced (4 before ablation, range 0-30 vs 0 after ablation, range 0-7, p=0.002) and was observed for ICD ATP (81 before ablation, range 0-292 vs 3.5 after ablation, range 0-29, p=0.001).

[0210] In the two patients with PVC-related cardiomyopathy, the 24-hour PVC burden decreased from 24% to 2% and from 26% to 9%. Fig.12 As shown in Figure 2, LVEF increased by 13% and 8%, respectively.

[0211] Prespecified secondary endpoints included a 50% and 95% reduction in VT episodes or 24-hour PVC burden. This endpoint was achieved in 94% and 61% of patients, respectively. In 89% of patients, the frequency of VT episodes or PVC burden was reduced by 75%. Prespecified secondary endpoints were elimination of ICD shocks and / or improvement in LVEF, which was achieved in 72% of patients. Despite a near-universal reduction in VT burden, many (11 / 16, 69%) had some degree of VT recurrence between the end of the 6-week observation period and 6 months. Overall survival was 89% at 6 months and 72% at 12 months ( Fig.13A ).

[0212] Fig. 13B The distribution of antiarrhythmic medication use before treatment and at 6 months is shown. The use of dual antiarrhythmic medication decreased from 58% to 11% (p=0.008). The use of high-dose amiodarone (>300 mg per day) decreased from 58% to 11%. The use of level 1 agents decreased from 67% to 11%. Four patients stopped antiarrhythmic medication completely.

[0213] After treatment, selected patient-reported quality of life scores are shown in Fig. 13C There were no changes in the general health domain. Quality of life scores did not decrease in any domain.

[0214] Example 3:

[0215] This example provides sample patient data and example output of a method for determining segments for ablation. The patient is a 50-year-old with non-ischemic cardiomyopathy (normal coronary angiography). The patient had an endocardial LV mapping / ablation procedure 2.5 years ago and is taking antiarrhythmic medications. The patient has monomorphic VT with a cycle length (CL) of 300-320ms and 183 episodes of tachycardia.

[0216] As in Figure 3A The left ventricle is shown segmented. A previous map / ablation identified minimal abnormal signal in segment 10, and an empirical catheter ablation of segment 10 was performed.

[0217] exist Fig.14 The patient's ECG is shown in , showing VT1 at segment 10, with a CL of 300 ms and a HR of 200 bpm. This indicates that the EGM within the heart matches the clinical VT.

[0218] MRI identifies wall dyskinesia and finds hypokinesia in segments 11, 12, 5, and 6. MRI with gadolinium identifies dense scarring in segments 5, 6, 11, 12, and 10 and intramyocardial scarring in segments 1 and 2, e.g. Fig.15A and Fig. 15B CT scan identified a thin area in segment 5, e.g. Fig.16A and Fig. 16B PET viability scan (non-inflammatory) identified areas of reduced uptake in segments 5 and 11, e.g., as in Fig.17 Shown in.

[0219] As in Fig.18A , 18B and 18C show that the earliest segments identified by ECGI are 10 and 6, so endocardial segment 11 is the likely origin because, as shown in Figure 3A As shown in FIG, segment 11 is located between two epicardial synapses of segments 10 and 6. Local EGMs with small R waves are identified. Sinus arrhythmias are identified in segments 10, 11, 6, and 5.

[0220] Contour Fig.19 The exit site is identified as segment 10. The nearby scars are identified as segments 11 and 12. The ablation is extended back to segment 5. The scars in segments 1 and 2 are not targeted. Segment 6 is not targeted because the ECG is completely prioritized.

[0221] Typically, this patient has a complex scar, including an almost transmural inferolateral scar, a patchy scar transition at the inferior and anterolateral edges, and a midmyocardial "band" in the anterior septal base. Clinical VT is easily initiated, and ECGI and ECG map to an inferomedial left LV outlet site (probably endocardial, with two epicardial outlets on ECGI). ECG and ECGI made targeting possible because previous cardiac mapping was inconsistent with the MRI scar.

[0222] From Fig.14 , 15A , 15B, 16A, 16C, 17, 18A, 18B, 18C and 19 are input maps in a method for determining one or more arrhythmia targets for ablation. Fig. 20A , 20B , 20C, 20D, 20E, and 20F show segmental abnormalities identified from each input map (EKG, MRI motion, MRI scar, CT, PET, ECGI, respectively). Figure 20G is the output of the method for defining one or more arrhythmia targets, showing the probability of each target. For example, higher probability targets are red, medium targets are orange, and lower targets are yellow. In this example, the final targets are segments 5, 10, 11, and 12.

[0223] Example 4:

[0224] This example provides patient sample data and exemplary output of a method for determining segments for ablation. The patient is a 56-year-old man with ischemic cardiomyopathy (LVEF < 25%). The patient has previously undergone 4 catheter ablations and is receiving dual antiarrhythmic therapy.

[0225] like Figure 3A As shown, the left ventricle is segmented. Fig.21A The patient's ECG is shown in Figure 1, with VT1 shown in segment 7 and Fig. 21B , VT2 is shown in segment 14.

[0226] MRI mapping was not available because the patient's pacemaker was in an elective replacement interval (ERI). CT scan identified apical aneurysmal thinning at segments 7, 13, 14, 15, 16, and 17, e.g. Fig. 22 PET scans identified large anterior apical aneurysms at segments 7, 13, 14, 15, 16, and 17, such as Fig.23A and 23B shown.

[0227] As in Fig.24A , 24B , 24C and 24D show that ECGI identified VT1 in segments 7, 12, 13 and 16, and VT2 in segments 15 and 16. Fig.24E and 24F As shown in FIG, ECGI identified large apical aneurysms in sinus rhythm in segments 13, 14, 15, 16, and 17, as well as isochronous late activation in segments 13 and 16.

[0228] Contour Fig.25A and 25B Shown in . Typically, the patient has a large anterior apical scar with two distinct VTs, both exiting from two different edges of the scar, and the sinus rhythm activation of interest aligns with the VT site. The ECGI is consistent with the VT1 ECG but different from the VT2 ECG. The targeted therapy approach balances scar homogenization (large) with the more favored approach. The VT on 12 leads is two distinct areas that are far apart and require more extensive ablation. Complete scar homogenization would require segments 7, 13, 14, 15, 16, and 17, while ECGI alone would recommend the outer edge of the scar with only segments 13, 15, 16.

[0229] From Fig.21A , 21BThe maps of 22, 23A, 23B, 24A, 24B, 24C, 24D, 24E, 24F, 25A and 25B are input maps into a method for identifying one or more arrhythmia targets for ablation. Fig.26A , 26B , 26C, 26D, 26E, 26F, 26G, and 26H show the segmental abnormalities identified from each input map (EKG for VT1, EKG for VT2, CT, PET, ECGI for VT1, ECGI for VT2, ECGI for sinus rhythm scar, and ECGI for sinus rhythm latest activation). Fig.26I is the output of the method for defining one or more arrhythmia targets, showing the probability of each target. For example, higher probability targets are red, medium targets are orange, and lower targets are yellow. In this example, the final targets are segments 7, 13, 14, 15, 16, and 17.

Claims

1. A method for determining an arrhythmia target for ablation, the method comprising: receiving a plurality of maps of a patient, the plurality of maps selected from electrical maps, anatomical maps, and functional maps; identifying a predetermined number of segments on each of the plurality of maps, each segment corresponding to a different anatomical location; identifying a first anomaly in a first mapping of the plurality of mappings; selecting any segment in the first map where the first anomaly is located; identifying a second anomaly in a second mapping in the plurality of mappings; selecting any segment in the second map where the second anomaly is located; combining the plurality of mappings; and The arrhythmia target is selected based on segments in the combined plurality of maps, wherein the selected segment containing the identified first abnormality and the selected segment containing the identified second abnormality overlap with each other.

2. The method according to claim 1, wherein: Identifying the first anomaly comprises identifying one or more contours on the first of the plurality of maps that contain the anomaly, or a combination thereof; as well as Identifying the second anomaly includes identifying one or more contours on the second map of the plurality of maps that contain the anomaly, or a combination thereof.

3. The method according to claim 2, wherein: Identifying the first anomaly includes receiving a manual identification of the anomaly from a user, and identifying the second anomaly includes receiving a manual identification of the second anomaly from a user.

4. The method according to claim 1, wherein: Selecting the arrhythmia target includes selecting the arrhythmia target based on segments in the combined plurality of maps and additional data about the patient, wherein the selected segment containing the identified first abnormality and the selected segment containing the identified second abnormality overlap with each other. The method of claim 1 , further comprising ablating the arrhythmia target.

6. The method according to claim 5, wherein: Ablating the arrhythmia target includes non-invasive ablation of the arrhythmia target.

7. The method of claim 1 , further comprising weighting each of the plurality of received mappings, and wherein: Selecting the arrhythmia target includes selecting the arrhythmia target based on segments in the combined multiple mappings and weights of the received multiple mappings, wherein the selected segment containing the identified first abnormality and the selected segment containing the identified second abnormality overlap with each other.

8. The method of claim 1 further comprising determining a confidence score for the combined mapping, and wherein: Selecting the arrhythmia target includes selecting the arrhythmia target based on segments and confidence scores in the combined plurality of maps, wherein the selected segment containing the identified first abnormality and the selected segment containing the identified second abnormality overlap with each other.

9. The method according to claim 1, wherein: The electrical map is an electrocardiogram image, the anatomical map is at least one of a computed tomography image or a magnetic resonance image, and the functional map is at least one of an optical emission computed tomography image, a positron emission tomography image, or an echocardiogram image.

10. The method of claim 1, further comprising generating a decision support module.

11. The method according to claim 10, wherein: The decision support module includes the output of the arrhythmia target and one or more of: a description of scar pattern, a description of scar burden size and location, a recommended amount of ablation to achieve, a confidence score for the combined mapping, a list of structures at risk, general recommendations, and an expected success rate for non-invasive therapies or an expected success rate for alternative treatment modalities.

12. At least one non-transitory computer-readable medium storing instructions that, when executed by at least one processor, cause the at least one processor to: receiving a plurality of maps of a patient, the plurality of maps selected from electrical maps, anatomical maps, and functional maps; identifying a predetermined number of segments on each of the plurality of maps, each segment corresponding to a different anatomical location; identifying a first anomaly in a first mapping of the plurality of mappings; selecting any segment in the first map where the first anomaly is located; identifying a second anomaly in a second mapping of the plurality of mappings; selecting any segment in the second map where the second anomaly is located; combining the plurality of mappings; as well as An arrhythmia target is selected for ablation based on segments in the combined plurality of maps, wherein the selected segment containing the identified first abnormality and the selected segment containing the identified second abnormality overlap with each other.

13. The at least one non-transitory computer-readable medium of claim 12, further comprising instructions that, when executed by the at least one processor, cause the at least one processor to select the arrhythmia target for ablation based on the segments in the combined plurality of maps and additional data about the patient, wherein The selected segment containing the identified first anomaly and the selected segment containing the identified second anomaly overlap with each other.

14. The at least one non-transitory computer-readable medium of claim 12, further comprising instructions that, when executed by the at least one processor, cause the at least one processor to weight each of the received plurality of maps and select the arrhythmia target based on segments in the combined plurality of maps and the weights of the received plurality of maps, wherein The selected segment containing the identified first anomaly and the selected segment containing the identified second anomaly overlap with each other.

15. The at least one non-transitory computer-readable medium of claim 12, further comprising instructions that, when executed by the at least one processor, cause the at least one processor to determine a confidence score for the combined mapping, and select the arrhythmia target based on segments in the combined plurality of mappings and the confidence score, wherein The selected segment containing the identified first anomaly and the selected segment containing the identified second anomaly overlap with each other.

16. The at least one non-transitory computer-readable medium of claim 12, wherein: The electrical map is an electrocardiogram image, the anatomical map is at least one of a computed tomography image or a magnetic resonance image, and the functional map is at least one of an optical emission computed tomography image, a positron emission tomography image, or an echocardiogram image.

17. The at least one non-transitory computer-readable medium of claim 12, further comprising instructions that, when executed by the at least one processor, cause the at least one processor to: receiving one or more clinical maps comprising at least one of demographics, surgical history, information regarding ventricular tachycardia, and prior electroanatomical mapping; as well as An updated arrhythmia target for ablation is determined based on the combined plurality of maps and the one or more clinical maps.

18. The at least one non-transitory computer-readable medium of claim 12, further comprising instructions that, when executed by the at least one processor, cause the at least one processor to generate a decision support module.

19. The at least one non-transitory computer-readable medium of claim 18, wherein: The decision support module includes the output of the arrhythmia target and one or more of: a description of scar pattern, a description of the size and location of scar burden, a recommended amount of ablation to achieve, a confidence score for the combined mapping, a list of structures at risk, general recommendations, and an expected success rate for non-invasive therapies or other treatment modalities.

20. The method according to claim 1, wherein: Selecting the arrhythmia target includes selecting the arrhythmia target based on segments in the combined plurality of maps and additional data from previous patients, wherein the selected segment containing the identified first abnormality and the selected segment containing the identified second abnormality overlap with each other.

Images