A system and method for guiding abnormal biological rhythms toward a target and treating that target.

Abstract

An ablation catheter for treating electrical rhythm disorders includes an array of sensor electrodes for detecting electrical signals to determine the location of a target region for treatment. If the catheter is not optimally positioned at the target region, a controller uses the detected signals to guide movement of the catheter toward the target region. Once proper positioning is confirmed, the controller activates an ablation component within the catheter to supply energy to modify tissue at the target region.

Description

[Technical Field]

[0001] Cross-reference of related applications This application claims priority to U.S. Provisional Patent Application No. 62 / 979,367, filed on 20 February 2020, the contents of which are entirely incorporated herein by reference.

[0002] Government rights This invention was implemented with government support under authorization numbers HL83359 and HL103800 granted by the National Institutes of Health (NIH). The government has certain rights in this invention.

[0003] Field of Invention The present invention generally relates to personalized identification and treatment for electrical rhythm disorders, and more particularly to systems and methods for facilitating personalized treatment. [Background technology]

[0004] Background of the Invention Medical therapies can be improved through personalization. Generally accepted therapies that work often do little to no work. Even in patients for whom a therapy works, there are often differences in response among individuals. Typically, there are few a priori clues as to whether a particular therapy will or will not work in a given patient. “Predictors” of response or failure are often based on retrospective observations, and present-day, insightful predictors offer only modest incremental benefits.

[0005] Current medical strategies explicitly prioritize the majority of individuals based on their described condition, but implicitly ignore statistical minorities. A crucial, yet overlooked, challenge is that this minority of individuals, even with the same diagnosis, may respond differently to therapies than those used for the majority. While this minority can comprise a large number of individuals, identifying them (their phenotype) can be difficult because they may already be separated into distinct subcategories from the others. [Overview of the project] [Problems that the invention aims to solve]

[0006] There is a need to personalize therapy in order to a priori identify patients who are likely to benefit from a therapy and those who are unlikely to benefit from it, and thereby to tailor the therapy to each individual. To satisfy these objectives, personalized medicine is being increasingly researched.

[0007] Personalized medicine is often advocated for conditions resulting from genetic causes ("mechanisms") in order to phenotype individuals and then adapt therapies accordingly. Unfortunately, many very common diseases have no apparent genetic cause. In the heart, for example, genetic cases can be identified in coronary diseases due to hereditary familial hypercholesterolemia or hereditary atrial fibrillation (AF), but these cases are minority. The majority of cardiac conditions are considered to have no clear genetic cause and are thought to be multifactorial, resulting from multiple factors. In fact, recent studies have failed to demonstrate genetic abnormalities even in conditions traditionally considered hereditary, including hereditary sudden cardiac arrest in young people, i.e., sudden arrhythmic death syndrome ("SADS").

[0008] Other conditions may manifest as partially hereditary, or may have hereditary causes with "incomplete penetrance." The causes of such variability in disease manifestation or response to therapy are unknown and occur, for example, with many therapies for atrial fibrillation. Such variability is often attributed to the "environment" and may manifest as fluctuations in the cellular "proteome" or "metabolome," but these can be difficult to identify, often unproven, and rarely used to guide therapy.

[0009] In a normal heart rhythm, the sinoatrial node maintains the heart's sinus rhythm. Heart rhythm disorders are common and are a significant cause of disease and death worldwide. The most common forms of heart rhythm disorders do not have a clear genetic cause.

[0010] Malfunctions of electrical systems or abnormal propagation of radio waves are common causes of rhythm disorders in the heart, brain, and other organs that generate electrical impulses ("excitable tissues"). Cardiac rhythm disorders can be classified as simple or complex. Simple rhythms have a stable and well-defined circuit over time, as detected by most methods of analysis. Examples include sinus rhythm (SR), inappropriate sinus tachycardia (IST) or rapid activation of the normal sinoatrial node leading to sinoatrial node reentry, atrial tachycardia (AT) or flutter (AFL), atrial-ventricular node reentrant tachycardia (AVNRT), and atrial-ventricular reentrant tachycardia (AVRT). Complex rhythm disorders have a relatively less distinct circuit that can change over time, such as atrial fibrillation (AF), ventricular fibrillation (VF), or polymorphic ventricular tachycardia (PMVT). Other rhythm disorders may have simple activation patterns and may be difficult to treat because they are transient or difficult to ablate, such as atrial premature contractions (PACs) or multiple ventricular premature contractions (PVCs), and this includes atypical forms of atrial flutter or ventricular tachycardia (VT).

[0011] Treating cardiac rhythm disorders can be challenging, especially in cases of AF, VF, and VT. Pharmacological therapies for complex rhythm disorders are not optimal, with success rates of only 40-60% in the medium to long term. Ablation for cardiac rhythm disorders is increasingly used and involves manipulating sensors / probes to the heart via blood vessels or directly in surgery, and supplying energy to the source region to alleviate or eliminate the rhythm disorder. In cases of complex rhythm disorders, ablation is often difficult because conventional systems for identifying and pinpointing the cause (source) are inadequate, lacking accuracy, precision, and / or time efficiency, which hinders attempts to supply energy to eliminate the disorder. For example, the success rate of a single ablation procedure for "paroxysmal" AF, considered the simplest form, is only 65% ​​at 1.5 years and further decreases over time. For patients with relatively complex persistent AF, the success rate of a single procedure using "absolute criteria" techniques is approximately 40-50% one year after treatment.

[0012] There are several unmet needs that, if addressed, could improve the success of the therapy. First, why does the same ablation method work in some patients, but not in others, even after multiple attempts? Second, what are the interpersonal mechanisms of similar or different rhythm disorders, and are these distinguishable beforehand? The current disease classification is not ideal for this purpose, because pulmonary vein isolation fails in 35-50% of cases of "simple" paroxysmal AF within 1-2 years, but works in 40-50% of cases of "advanced" persistent AF.

[0013] One proposed mechanism (cause) for AF is a localized source region or driver (referred to as rotor, rotational activity, repetitive activity, or focal site) that can drive surrounding disordered activity. It is unclear how well the source can be identified. It is unclear why some patients recover their health after ablation of AF sources while others do not. It is unclear why some individuals have a small number of source regions even in complex AF, while others have several. It is not defined why source regions are related to structural anomalies such as low voltage or magnetic resonance imaging anomalies in some individuals, but not in others.

[0014] Electrical rhythm disorders are classified by their electrical patterns. This often involves introducing a catheter with multiple sensors / probes into the heart via the patient's blood vessels. These sensors detect the heart's electrical activity (ECG) at multiple locations, which is used to identify the cause of conditions such as AVRT or AVNRT and to define distinct therapies, although the ECG appearance is similar. In the case of simple arrhythmias such as atrial tachycardia, the source can be identified by tracing the activation back to the earliest location, which is then ablated to treat the disorder. This can be difficult even in the case of simple cardiac rhythm disorders.

[0015] Identifying the source or other target region for treating complex rhythm disorders is relatively difficult. Firstly, the signal at each sensor can change beat by beat in terms of shape and the number of deflections. When the signal in AF has 5, 7, 11 or more deflections, it is difficult to identify what is local (i.e., below the sensor), and these are often from adjacent regions (i.e., far-field activity) or noise. Secondly, the relative pathways of activation between adjacent sensor sites can change over time, as in AF or VF. Overall, this makes it difficult to correctly map the activity in complex arrhythmias to identify its source.

[0016] The causes of cardiac rhythm disorders can be identified by several methods, but none of them are yet perfect. It is difficult to a priori distinguish between patients with and without localized sources. Some identified sources may be false positives that do not require treatment (even when the source is verified by optical imaging). Methods for identifying sources, including the use of clunky, low-resolution systems, are cumbersome and time-consuming. Because sources can be located anywhere, conventional methods often involve mapping the entire chamber non-invasively from the body surface, either using multipolar catheters. These types of global mapping systems are difficult to use, have low spatial resolution, and are variable.

[0017] Further conventional treatment methods for complex arrhythmias often require different tools for mapping the arrhythmia and separate tools for delivering therapies, thus introducing practical disconnections when switching tools. When systems are swapped to detect critical areas in order to treat critical areas, registration errors may reduce the accuracy of precisely treating the same site and may add time. Furthermore, conventional methods do not clearly indicate which source areas are most important and when they are detected. Thus, all sources are treated in common, but some of these sites may not be important in any given patient, and treating all these sources adds time, difficulty, and potential risk to the procedure. [Means for solving the problem]

[0018] Summary of the Invention The present invention provides a system and method for identifying and locating source regions or other target regions to treat biological rhythm disorders using a personalized digital medical approach. The system utilizes a probe or catheter to detect electrical signals from biological tissue and provides navigation guidance to the rhythm source or target region based on the detected electrical signals. The system can then directly treat these regions without moving or replacing the probe or catheter. All steps can be automatically adapted to the individual based on a quantified artificial intelligence-based algorithm in which patients with similar data patterns respond to treatment.

[0019] The systems and methods described herein provide a modality of quantitatively personalized therapy via one or a combination of lifestyle changes, medication, electrical or mechanical therapies, surgical or minimally invasive ablation, genetic or stem cell therapies. The invention disclosed herein is partially related to the subject matter of International Patent Application No. PCT / US2019 / 029004 filed on July 22, 2019, and the disclosure of this patent document is hereby incorporated by reference in its entirety.

[0020] An exemplary embodiment uses tools to identify individuals in whom ablation therapy for complex rhythm disorders is likely to be successful. These tools may be non-invasive or invasive. In patients suitable for ablation therapy, another embodiment includes a device that provides direction guidance for recording the electrical pattern within the heart and for moving the device three-dimensionally within the biological organ towards the optimal location for therapy. Another embodiment provides the ability to deliver therapy directly to the tissue at this location.

[0021] In some embodiments, the system of the invention provides a personalized diagnosis of complex rhythm disorders, navigation guidance to the target site of the rhythm disorder, and a "single-shot" detection and therapy tool for said rhythm disorder.

[0022] An advantage of the present invention lies in its ability to personalize therapy by comparing a stream of data from an individual at the present time with streams from other individuals having similar or dissimilar profiles using a digital taxonomy that can be updated using strategies such as crowdsourcing.

[0023] The examples described herein are directed to cardiac rhythm disorders, mechanical contractions, or heart failure, but other exemplary uses of the inventive approach include epileptic disorders of the brain, gastrointestinal rhythm diseases such as irritable bowel syndrome, and bladder disorders including detrusor instability. Generally, the inventive approach is applicable not only to atrial fibrillation within the heart or generalized seizures within the brain, but also to disordered disorders within organs such as simple rhythm disorders. Thus, the examples provided herein are not intended to be limiting. The personalized aspect of the present invention is suitable for disorders that are heterogeneous syndromes rather than simple disease entities.

[0024] The present invention discriminates between patients in whom the therapeutic target is a localized source for a rhythm disorder and patients in whom no source exists. An example of this embodiment is to identify patients with atrial fibrillation who are likely to benefit only from pulmonary vein isolation ablation. Other patients may require ablation of a localized source for success. Others may require ablation of other targets, such as those targeted by the maze procedure. Similarly, the inventive approach can identify patients with ventricular tachycardia in whom ablation will succeed or fail.

[0025] The source region is a subset of the targets for the rhythm disorder and is identified as a patch or region of tissue activity that (a) occurs within a disorder such as atrial fibrillation within the heart or (b) from which activation emanates to drive a tissue disorder such as atrial or ventricular tachycardia. The inventive approach uses analytical tools including machine learning to detect tissue patches. Sources located in the vicinity of regions targeted by standard therapies such as the pulmonary veins in atrial fibrillation, the scar isthmus in ventricular tachycardia, or local brain lesions in epileptic disorders may not require specific additional therapy. This information is communicated to the operator.

[0026] Furthermore, the method of the present invention informs of the most important targets for rhythm disorders. Without this information, the method would often involve treating all detected targets in atrial fibrillation, which would involve detecting and treating multiple sources, scar tissue areas, or complex signals. However, some of these areas may not be important, and this method can be time-consuming, complicating, and potentially having adverse effects. Therefore, the present invention identifies patients with targets located in areas already treated by standard therapy, or those that are less clear, but none of which require further treatment.

[0027] In one embodiment, the present invention quantifies the importance of a target area by quantifying the size or area of ​​a tissue region or patch of disturbing activity, such as atrial fibrillation in the heart or generalized persistent / chronic seizures in the brain. A target hierarchy, from most dominant to least dominant, is communicated to the operator, which can be used for treatment planning.

[0028] The present invention uniquely detects therapeutic targets for biological rhythm disorders, such as localized sources, without the need for wide-area global mapping. Global mapping can be cumbersome, may not cover entire organs, and typically requires the use of large probes or catheters that are not ideally suited for treatment or are unable to deliver therapy; therefore, detection and therapy require the use of separate probes. In one embodiment, the system of the present invention uses a mapping spade that is physically large enough to cover source areas for simple or complex rhythm disorders, or to cover other targets such as channels in viable tissue within fibrous areas small enough to provide high-density recording.

[0029] A mapping tool or spade contains multiple electrodes, which can be numbered from approximately 4 to 256. Each electrode has a size ranging from 0.1 to 4.0 mm, and in this case, the size selection depends at least in part on the characteristics of the suspected disorder. For complex rhythms such as atrial fibrillation, typical electrodes have a size range of 0.5 to 1.0 mm to provide good signal fidelity and to detect complex signal types that may be therapeutic targets. For ventricular tachycardia, typical electrodes have a size range of 1 to 2 mm. For simpler rhythms such as tachycardia mediated by accessory pathways, typical electrode sizes are 0.5 to 1 mm to recognize the potential of the accessory pathway. The selection of appropriate electrode sizes for other applications falls within the realm of the art.

[0030] The spacing between electrodes varies within the range of 0.5 to 5.0 mm. In atrial fibrillation, the usual electrode spacing is 1 to 2 mm. In ventricular tachycardia, the usual electrode spacing is 2 to 4 mm. When extremely fine details need to be elucidated, the usual electrode spacing is 0.5 to 0.75 mm.

[0031] The size of the Spade is personalized not only for the number and spacing of electrodes, but also for the type of rhythm and the patient's profile. Personalization is performed using tools such as machine learning calibrated for patients with similar types and data (Personalized Digital Phenotype, PDP). The Spade therapy tool makes contact with the organ by conforming its surface to the same multiple locations where the target or source is recorded.

[0032] Contact can be improved by using various conforming materials in the structure, depending on the intended location within the target organ. For example, Nitinol (nickel-titanium alloy) in 34-36 gauge is one such material that can provide sufficient structural stability and flexibility. This can be used to construct devices for thermal ablation such as radiofrequency or light-emitting diode ablation, cryoablation such as cryoablation, or non-thermal ablation such as pulsed-field ablation. One embodiment uses a conforming chamber for mapping and cryoablation, in which case the therapeutic device adheres to the tissue during energy delivery for rapid, accurate, and safe ablation. This can be effective in cases of atrial fibrillation and atrial tachycardia sources in the heart and in cases of seizure foci in the brain.

[0033] In one embodiment, having both the detector and the therapeutic element within the same physical device eliminates the need to use separate tools for each. This can reduce time and complexity, and even improve accuracy, because the location of the desired target area is saved or recorded and does not need to be re-found using separate tools thereafter.

[0034] In one embodiment, the present invention provides navigation guidance for a sensor tool without the need to collect data globally using cumbersome, large catheters. The invention processes data at the current sensor site and calculates the direction in which to move the sensor to navigate to the source. This is similar to a car's global positioning system, which uses the current location to navigate to a desired location without surveying the entire globe map. This approach enables relatively high-resolution mapping in the vicinity of a target area compared to what is currently used in wide-area or global mapping systems within the heart.

[0035] This invention personalizes detection and therapy using personal digital phenotypes (PDPs) of health and disease. PDPs digitally implement "personalized medicine" or "high-precision medicine" with or without cellular or genetic data. Generally, -omic data may be unavailable for many individuals, or its contribution to aging or environmental disease may be relatively low. Input data (e.g., data streams from sensors, stored data from electronic health records, imaging data) is linked to observed labels such as changes in surrogate markers or elimination of disease by specific therapies. The PDP then partitions the input for that individual into those related to health and those related to deviations (possible disease). Thus, this invention does not address only the statistical majority of individuals.

[0036] PDP can combine data streams individually or in combination (e.g., networked) from various sensors, medical devices, or consumer devices. Data streams may originate from dedicated equipment such as imaging systems, from novel wearable sensors, or from multiple individuals, for crowdsourced population data. Data from existing systems may include unidentified data, diverse contributing patients, practice patterns, and data from multiple hospitals within a large digital registry of outcome data from different therapies. Such a scheme may require blockchain technology to ensure data security, traceable logs of data transactions, and data access across multiple physical storage systems.

[0037] PDP informs of the relevance of biological and clinical data for rhythm disorders within an individual, which may not be apparent to professionals using systems and methods trained on pre-labeled datasets indicating whether a particular therapy was successful or unsuccessful. This enables the identification of individuals who have and do not have treatable forms of the disorder, such as predicting the location of sources for biological rhythm disorders, providing assistance in guiding navigation to such sources, predicting the type and size of such sources, and identifying individuals who are likely to respond to personalized therapy for that individual.

[0038] A PDP is generated from a digital taxonomy of patients with a given disease or health condition. The taxonomy is constructed from multiple data streams and stratified by desirable or undesirable outcomes. Input data can be simple, such as heart rate, weight, and other easily accessible data, or elements stored in electronic health records, and / or complex or sophisticated data that can change dramatically over time (e.g., proteomics and biomarkers) or not change over time (e.g., genetic data). Other phenotypes may be clinical labels not tracked by biomarkers, or may have loose statistical definitions such as race or racial susceptibility. The more detailed and extensive the population data elements, i.e., the "richer" they become, the more extensive the digital taxonomy will also become.

[0039] A Personal Digital Phenotype (PDP) is a quantified pathophysiological network representing data clusters, including those from signal processing, associative algorithms, and unsupervised machine learning, as well as indices from supervised networks trained on labeled events in similar and dissimilar individuals. The data is partitioned into data labeled as “health vs. disease” or “response vs. non-response to therapy,” analyzed by one or more of the following: supervised machine learning, neural networks, unsupervised machine learning, cluster analysis, correlation analysis, logistic regression analysis, decision trees, time-domain analysis, frequency-domain analysis, trigonometric transformations, and logarithmic transformations.

[0040] The patient's tissues may include the heart, nerves supplying the heart region, brain regions controlling the nerves, blood vessels supplying the heart region, and tissues adjacent to the heart. In some embodiments, the disease may be a cardiac rhythm disorder having one or more of the following conditions: atrial fibrillation, ventricular fibrillation, atrial tachycardia, atrial flutter, polymorphic or monomorphic ventricular tachycardia, ventricular flutter, or other electrical abnormalities within the heart.

[0041] Analysis based on PDP can decipher patterns of cardiac rhythm disorders that are difficult for experts to understand. This is especially true in cases of complex disorders that may include rotational circuits, focal circuits, repetitive patterns, partial rotations or focal circuits, "random" activity, electrical propagation around scar areas, or specific anatomical sites within an individual. These patterns are difficult to sort. A digital taxonomy links specific patterns in a given patient of PDP to the success or failure of drug therapy, ablation, Maze surgery, or other therapies. A patient-specific PDP based on the patient's electrical, structural, and clinical data is "fitted" into the taxonomy to identify appropriate targets for therapy. This personalized identification of diagnostic or therapeutic targets is novel and based on the integration of data across biological scales.

[0042] A PDP for cardiac rhythm may include a data stream of invasive recordings of electrical activity (ECG), blood flow and blood pressure (hemodynamics), wall tension (cardiac contraction and relaxation), and related indices. More detailed data may include three-dimensional anatomical and structural abnormalities. Clinical data may be extracted from history and physical examinations, indices of pathophysiological comorbidities, blood and tissue biomarkers, and the individual's genetic and cellular composition. Non-invasively, the sensor may record the ECG, skin response scales of nerve activity, and skin reflectivity. Other types of detection signals that may be used will become apparent to those skilled in the art.

[0043] In cases of complex cardiac rhythm disorders, inflammation is a likely contributing factor, but is often not included in patient phenotyping. Inflammation can cause some arrhythmias or other conditions such as myocarditis after surgery. The link between obesity and atrial fibrillation may operate through inflammation in the extrapericardial fat, resulting in reactive oxygen species. The detection of inflammation may have undefined importance at a single point in time, over time, among people, or within any given person. While "inflammasomes" can measure the effects of inflammation from various pathological disorders at the cellular or tissue level, they are not generally performed, cannot assess diurnal variation, cannot have a clear relationship to inflammation throughout the body, and can vary among individuals. Therefore, there is no clear method for establishing a "nomogram" of normal or abnormal conditions.

[0044] Inflammatory biomarkers constitute a data stream. Personalized inflammatory states can be detected by inflammatory cells within the organ system experiencing inflammation, or in bodily fluids such as blood, urine, or cerebrospinal fluid. By-products of inflammation can be detected by elevated concentrations of biomarkers and cytokines such as interleukin-6, nerve growth factor, and matrix metalloproteinases. Conversely, several physiological markers are abnormal in inflammation (so-called "acute phase reactants"). Inflammation causes abnormalities in numerous acute phase reactants, including elevated white blood cell count, red blood cell count, hemoglobin concentration, and C-reactive protein, erythrocyte sedimentation rate, and white blood cell count. Within the heart, serum troponin, a marker of cardiac cell destruction, is well known to be an acute phase reactant whose levels decrease with inflammation ("reverse acute phase reactant").

[0045] Arrhythmias within a subgroup of patients with inflammatory causes can be targeted using anti-inflammatory therapies, including immunosuppression, with substances such as tacrolimus, which is a previously unrecognized form of therapy for complex arrhythmias such as atrial fibrillation. Other immunosuppressive therapies or cell therapies, such as steroids or nonsteroidal substances, may also be effective. One logical explanation is that patients who receive immunosuppressants after heart transplantation rarely develop AF. The benefit is attributable to surgical isolation of the pulmonary veins at transplantation, but such isolation in other populations provides relief from less than 50-70% of AF. The use of immunosuppression for complex rhythm disorders, including AF, is rarely used. Digital taxonomy and PDP in this invention will identify individuals with inflammation-mediated arrhythmias for which anti-inflammatory therapies, including immunosuppression, may be beneficial.

[0046] For other organ systems, detection signals from measurable body tissues may include electroencephalograms (EEGs) measured on the central and peripheral nervous systems or scalp, invasive electrode recordings, or peripheral sensors. Measurements may also include the respiratory system, skeletal muscle and skin, any index of electrical signals, hemodynamics, clinical factors, nerve signals, gene profiles, biomarkers of metabolic status, and patient movement. Other input data elements may originate from imaging, nuclear, genetic, laboratory, or other sources, and may be detected as streams (i.e., transmitted to the system) or input as values ​​at specific points in time.

[0047] Generally, the sensor may be in physical contact with the patient's body, and the sensor data stream is acquired by either wired or wireless transmission. The sensor may be one or more of the following: electrodes, optical sensors, piezoelectric sensors, acoustic sensors, electrical resistance sensors, thermal sensors, accelerometers, pressure sensors, flow sensors, and electrochemical sensors.

[0048] Personalized therapy may include steps to modify at least a portion of the patient's tissue by one or more of the following: energy delivery via contact devices, ablation by energy delivery via non-contact devices, electrotherapy, thermotherapy, mechanical therapy, drug therapy delivery, immunosuppression delivery, stem cell therapy delivery, and gene therapy delivery. The computing device may be further configured to generate updated personal history data having a PDP, one or more classified qualitative disease classifications, personalized interventions, and intervention outcomes.

[0049] In one embodiment, the system of the present invention includes a processor and a memory storing instructions that, when executed by the processor, perform actions including: detecting bodily signals associated with one or more bodily functions in one or more sensors associated with a human body; processing the bodily signals to generate one or more detected signatures; processing the signatures using digital objects to determine effector responses; distributing one or more effector responses to control a bodily task; and monitoring the responses.

[0050] In another embodiment of the present invention, an ablation catheter for treating electrical rhythm disorders includes an array of sensor electrodes for detecting electrical signals to determine the location of a target region for treatment. If the catheter is not optimally positioned in the target region, a controller uses the detection signals to guide the catheter's movement toward the target region. Once proper positioning is identified, the controller activates the ablation component within the catheter to supply energy for altering the tissue in the target region.

[0051] In summary, the present invention is capable of identifying individuals suitable for therapy for complex rhythm disorders, providing three-dimensional directional guidance to move a novel sensor device toward the optimal location for therapy, and providing the ability to deliver therapy directly to the tissue at this location. Thus, one embodiment is a system providing personalized diagnosis of complex rhythm disorders and a "single-shot" sensor / therapy tool. Some embodiments, not intended to be limiting, include cardiac applications in cardiac rhythm disorders, in ischemic heart disease, and in heart failure.

[0052] In one aspect of the present invention, a system for treating cardiac rhythm disorders includes a catheter configured to be positioned in contact with a tissue surface, having a flexible body with a contact surface; an array of sensor electrodes positioned within the flexible body, each sensor electrode having a conductive surface substantially coplanar with the contact surface, and each sensor electrode being configured to detect electrical signals from the tissue surface; and one or more therapeutic elements configured to supply energy to the tissue surface. Each conductor of the plurality of conductors has a far end connected to a sensor electrode and one of the one or more therapeutic elements. A controller in communication with the near ends of the plurality of conductors includes a processor configured to receive detected electrical signals, determine the location of a target region of cardiac rhythm disorder, determine whether the catheter is optimally positioned in the target region, calculate the orientation to the target region and generate a motion command to move the catheter toward the target region if it is not optimally positioned, and generate a therapeutic signal to activate one or more therapeutic elements to modify the tissue within the target region after it has been determined that the catheter is optimally positioned. In some embodiments, the flexible layer is generally planar and has a shape selected from the group consisting of rectangles, ellipses, and rings. An elongated hollow shaft having a distal end, a proximal end, and a length is provided with a catheter disposed at the distal end, and a controller disposed at the proximal end, and as a result, multiple conductors are held within the shaft and extend along its length, in which case the distal end of the shaft is operable from the proximal end. A shaft motor can be configured to steer the distal end of the shaft in response to motion commands generated by the controller.

[0053] A sheath, slidably mounted on the shaft, has an internal volume configured to hold the catheter in a folded state until it is deployed by sliding the sheath away from the distal end of the shaft.

[0054] In some embodiments, the catheter may further include an irrigation fluid port formed within a flexible body, the irrigation fluid port being in fluid communication with a controller and associated irrigation fluid source, in which case the irrigation fluid source is configured to supply irrigation fluid to tissue in a target region through the irrigation fluid port in connection with the activation of an array of therapeutic elements.

[0055] In some embodiments, one or more therapeutic elements have an array of ablation electrodes, and in this case, a subset of multiple conductors connected to one or more therapeutic elements is a conductor configured to supply electromagnetic energy to each ablation electrode. The array of sensor electrodes and the array of ablation electrodes may be uniformly dispersed around the contact surface, or the array of ablation electrodes may be evenly scattered among the array of sensor electrodes.

[0056] The processor may be further configured to determine the size of the target region based on the detected electrical signal, to identify one or more ablation electrodes in an array of ablation electrodes based at least on the size and location of the target region, and to activate the identified ablation electrodes.

[0057] The processor may generate a directional map of the heart rhythm based on the detected electrical signals, generate guidance directions for moving a flexible body toward the target region, and determine the location of the target region by integrating the directional map to determine the location of the target region, in which case the directional map describes the conduction path of the heart rhythm. The directional map can be generated by applying a trained machine learning model to the electrical signals, in which case the machine learning model is trained on training examples that have known target regions of human heart electrical signals and heart rhythm disorders.

[0058] In some embodiments, each ablation electrode is configured to emit a distinct waveform. The controller can be configured to address one or more subsets of the ablation electrodes in the array individually, in which case the treatment signal comprises a first signal to a first subset of ablation electrodes for emitting a first waveform and a second signal to a second subset of ablation electrodes for emitting a second waveform. The sensor electrode array has at least four electrodes. In some embodiments, the sensor electrodes can be configured to supply ablation energy such that one or more treatment elements have an array of sensor electrodes.

[0059] In certain embodiments, one or more therapeutic elements have one or more coolant chambers formed within a flexible body and configured to hold coolant, and in this case, the plurality of conductors have a subset of conductors configured to guide coolant fluid from a coolant source to one or more coolant chambers to supply freezing energy to tissue in a target region. The flexible body may have a thermally conductive material built inside to improve the conduction of freezing energy to tissue in contact with the contact surface. Alternatively, one or more therapeutic elements may be an array of cryoablation locus formed within a flexible body, and in this case, the plurality of conductors have a subset of conductors configured to guide coolant fluid from a coolant source to the cryoablation locus in response to therapeutic signals from a controller to supply freezing energy to tissue in a target region. In further embodiments, one or more therapeutic elements may be an array of targeting fiducials dispersed within a flexible body, in which case the targeting fiducials are configured to guide the supply of ablation energy from one or more external ablation energy sources.

[0060] In yet another aspect of the present invention, a method for treating cardiac rhythm disorders includes: detecting electrical signals of the heart using an array of sensor electrodes; generating a directional map describing the conduction pathway of the cardiac rhythm based on the detected electrical signals; integrating the directional maps to determine (i) the location of a target region of cardiac rhythm disorder within the directional map and (ii) one of the guidance directions to the target region of cardiac rhythm disorder outside the directional map; determining whether a flexible body is optimally positioned in the target region according to the directional map; and activating one or more therapeutic elements of the system to supply energy for tissue modification at the determined location in the target region in response to the determination of optimal positioning. Generating the directional map may involve applying a trained machine learning model to the detected electrical signals, in which case the machine learning model is trained on training examples having known locations of human cardiac electrical signals and one or more target regions of cardiac rhythm disorders.

[0061] In some embodiments, the method may further include maneuvering the device to a subsequent position in the guidance direction to a target region in response to a determination of the guidance direction to a target region outside the direction map. Furthermore, the method may include detecting subsequent electrical signals of the heart by a plurality of sensing electrodes after maneuvering the device to a subsequent position; generating a subsequent direction map describing the conduction path of the heart rhythm based on the subsequent electrical signals; integrating the subsequent direction map to determine (i) the location of the target region of the heart rhythm disorder within the subsequent direction map and (ii) one of the subsequent guidance directions to the target region of the heart rhythm disorder outside the subsequent direction map; and maneuvering a flexible body to a third position in the subsequent guidance direction to a target region in response to a determination of the subsequent guidance direction to a source region outside the direction map. In addition, the method may include providing a notification on an electronic display to move the device in the direction of the target region in response to a determination of the direction of the target region outside the direction map. Other methods may include one or more of the following: identifying one or more ablation components within a threshold proximity to the location of the target region in response to the determination of the location of the target region, and determining the size of the target region based on a direction map, in which case the identification of one or more ablation components is at least further based on the size of the target region.

[0062] In further embodiments, the method may include: generating a subsequent directional map describing the conduction pathway of the cardiac rhythm based on subsequent electrical signals in response to a determination that a cardiac rhythm disorder persists; integrating the subsequent directional maps to determine the location of a second target region of the cardiac rhythm disorder within the subsequent directional map and one of the guidance directions to the second target region of the cardiac rhythm disorder outside the subsequent directional map; and instructing one or more ablation components on the device to modify tissue in the second target region at the determined location of the second target region in response to the determination of the location of the second target region.

[0063] Brief explanation of the drawing Some embodiments are shown in the following attached drawings as examples, not as limitations. [Brief explanation of the drawing]

[0064] [Figure 1] This is a block diagram illustrating the use of a personal digital phenotype (PDP) for clinical purposes in individuals, for comparison with a digital taxonomy, to enable personalized diagnosis and to provide personalized therapy, in one or more embodiments. [Figure 2] This describes the use of a personal digital phenotype in a system for the heart of the present invention, which integrates streaming data from the heart or other organs with input data, having an output designed to diagnose or treat a region of the heart, according to one or more embodiments. [Figure 3] This document illustrates the common use of a personal digital phenotype in one or more embodiments for performing a diagnosis, delivering therapy, and tracking / displaying an individual therapy response. [Figure 4] This is a flowchart illustrating the generation of a personal digital phenotype in comparison with a digital taxonomy, according to one or more embodiments. [Figure 5] In one or more embodiments, a personal digital phenotype is shown to be compared to a stored normal value for the individual or to a population value in order to notify of health or disease. [Figure 6] This document summarizes a process flow for describing data elements that are important for distinguishing health or disease in a personal digital phenotype, according to one or more embodiments. [Figure 7] This is a flowchart for managing complex arrhythmias based on PDP, according to one or more embodiments. [Figure 8] One embodiment of a system for mapping cardiac arrhythmias and notifying the time when a sensor reaches a target source region is shown, by a display unit of detected signals that notifies directional guidance for the sensor to move toward the target source region, according to one or more embodiments. [Figure 9] This shows sample regions (ROIs) that could be potential targets for therapeutic purposes in treating rhythm disorders. [Figure 10] This provides an overview of the directional guidance in the present invention, which uses detected data within an individual to direct the guidance toward a target region for rhythm disorders, according to one or more embodiments. [Figure 11A] This is a flowchart illustrating the steps for directional analysis according to one or more embodiments. [Figure 11B] This is a flowchart illustrating the steps for directional analysis and treatment according to one or more embodiments. [Figure 12] This is a flowchart showing a process for treating electrical rhythm disorders according to one or more embodiments. [Figure 13] The following are exemplary ablation catheters for treating electrical rhythm disorders according to one or more embodiments. [Figure 14A] An example of an alternative spade configuration is shown. [Figure 14B] This demonstrates the adaptation of a spade configuration to a source or other target region in electrical rhythm disorders. [Figure 15A]This is a perspective view of one embodiment of an ablation catheter configured to supply electromagnetic energy to tissue. [Figure 15B] This is a cross-sectional view of one embodiment of an ablation catheter configured to supply electromagnetic energy to tissue. [Figure 16A] This is a perspective view of one embodiment of an ablation catheter configured to deliver irrigation fluid to tissue. [Figure 16B] This is a cross-sectional view of one embodiment of an ablation catheter configured to deliver irrigation fluid to tissue. [Figure 17A] This is a perspective view of one embodiment of an ablation catheter having one or more cryoablation components configured to apply freezing energy. [Figure 17B] This is a cross-sectional view of one embodiment of an ablation catheter having one or more cryoablation components configured to apply freezing energy. [Figure 17C] This is an alternative cross-sectional view of one embodiment of an ablation catheter having one or more cryoablation components configured to apply freezing energy. [Figure 18A] This is a perspective view of one embodiment of an ablation catheter having one or more cryoablation components configured to apply freezing energy. [Figure 18B] This is a cross-sectional view of one embodiment of an ablation catheter having one or more cryoablation components configured to apply freezing energy. [Figure 19] This is a perspective view of one embodiment of an ablation catheter having a targeting fiducial. [Figure 20] This is a block diagram of an exemplary computing environment for implementing an embodiment of the present invention. [Modes for carrying out the invention]

[0065] Detailed description of exemplary embodiments For the purposes of this disclosure, the following definitions apply.

[0066] "Ablation energy" refers to the energy used to alter tissue. The altered tissue may correspond to a source region or other target region for electrical rhythm disorders. The alteration of tissue affects one or more electrical rhythms generated in the source region. The intended effect of providing ablation energy to a target region is to treat the electrical rhythm disorder. Ablation energy includes electromagnetic energy (e.g., in the form of high-frequency energy provided by ablation electrodes), freezing energy (e.g., removal of heat from tissue by a coolant, which is generally rapid removal or rapid cooling), and any other form of energy that has the ability to alter tissue.

[0067] Associative learning refers to the process of linking input data to measurable physiological or clinical outcomes. Associative learning may be iterative, which allows the associations to be modified ("learned") based on patterns of change between the input and the measured output (physiological or clinical endpoint).

[0068] "Biological signals" are signals produced by the body and may reflect one or more bodily systems. For example, heart rate reflects cardiac function, autonomic nervous system tension, and other factors. See also "non-biological signals."

[0069] "Biometric signals" refer to signals that provide metrics of human characteristics. Biometric identifiers may be physiological or behavioral. Physiological biometrics include, without limitation, an individual's DNA, fingerprints or palm prints, oral swabs, tissue or urine samples, retinal images, facial recognition, hand or foot shape, iris recognition, or scent / fragrance. This may also apply to signals such as vital signs, ECG, EEG, EMG, etc. Behavioral biometrics include patterns such as gait or typing rhythm. Embodiments of the present invention utilize dynamic patterns of combined physiological and behavioral biometrics over time, which adapt to changes within the individual and are therefore stable against forgery from previous "versions" of the person's signature.

[0070] "Body" refers to the physical structure of single-celled organisms, multicellular organisms, viruses, and prions. "Organism" includes animals (such as humans and other mammals, without limitation), plants, bacteria, and so on.

[0071] "Consumer devices" refer to devices that are directly available to consumers without a medical prescription. In the past, such devices were not typically regulated by medical regulatory authorities or organizations, such as the U.S. Food and Drug Administration or other similar regulatory authorities in other countries, although recently some devices have received FDA approval. Consumer devices may include hardware, software, or a combination thereof. This is not typically a medical device, which is defined as an instrument, apparatus, implementation, machine, device, implant, in vivo reagent, or other similar or related article, including component parts or accessories, intended for use in the diagnosis of disease or other condition in human or other animals, or in the treatment, mitigation, therapy, or prevention of disease. The definition of a medical device excludes medical decision support software.

[0072] An "effector" is a means of performing a bodily task, such as a physical appliance, prosthesis, or mechanical or electrical device. A physical appliance can improve bodily function, such as one or more signals to stimulate bodily function, including devices to move limbs or diaphragms to improve breathing during sleep, splints to maintain an open airway during sleep, or electrical stimulation of the phrenic nerve to improve breathing during sleep; or a prosthesis, such as a cybernetic limb or an implanted circuit for the peripheral or central nervous system.

[0073] A “data stream” refers to biological data detected by one or more sensors that can provide real-time or near-real-time information about the biological process being detected. Sensors within the heart may provide a stream containing an electrocardiogram (ECG), heart rate, pulsation waveform, and therefore, cardiac hemodynamics. Other data streams may include heart sounds, including analysis of heart murmurs and advanced analysis of hemodynamics related to the heart. Lung function can be detected as chest movement, auscultation sounds, and nerve firings associated with respiration. Gastrointestinal diseases can be detected as sounds (borborygmus), movement on the abdominal wall, and electrical signals related to smooth visceral muscle activity. Central and peripheral nervous system activity can be detected as nerve activity from scalp (electroencephalogram, EEG), detached from the scalp, but still reflecting the EEG, and from peripheral nerve firings.

[0074] As used herein, “demographic” means personal information that may include, and may be clinically appropriate, age, sex, family history of disease, ethnicity, and the presence of comorbidities, without limitation.

[0075] A "digital taxonomy" refers to a partition of different disease or health conditions based on quantitative indices. Traditional disease classifications are qualitative, for example, "atrial fibrillation is relatively common in relatively older individuals, those with comorbid cardiac conditions such as valvular disorders or heart failure, and those with metabolic syndrome." Digital taxonomies are quantitative and describe an individual's health or risk of a specific disease in terms of quantifiable primary and secondary data elements (data vectors). Disease Entity D n The likelihood that it exists within a particular individual is given by probability p(D n ) is approximated by,

number

[0076] "Historical data" refers to stored data, which may include clinical demographics such as age, sex, family history of disease, and presence of comorbidities, as well as reports from medical imaging such as magnetic resonance imaging (MRI), computed tomography (CT), radiology, or other scans of organs, data from genetic testing and analysis (e.g., presence of one or more genomic mutations), pre-obtained ECG reports, pathology, cytology, and other laboratory reports. Historical data may further include additional personal history details that may be relevant to the development of PDP, such as psychosis, employment in a stressful occupation, number of pregnancies (for women), smoking, and engaging in high-risk behaviors such as drug or alcohol abuse.

[0077] "Input data" or "data input" means data that is not directly detected by the physical components of the system, but is used by the processing unit in relation to detected data in order to generate the PDP and digital taxonomy. Input data from data sources may include, for example, streams of data detected using other systems such as external ECG or EEG systems, clinical, laboratory, pathological, chemical, or other data, or data from medical imaging devices, and this data is transmitted to the processing unit.

[0078] "Indicator individual" means a patient or research or evaluation target from which a personal digital phenotype can be generated.

[0079] "Machine learning" refers to a set of analytical methods and algorithms that can learn from data by building models rather than following static programming instructions, and that can make predictions based on that data. Machine learning is often classified as a branch of artificial intelligence and focuses on developing computer programs that can change when exposed to new data. In this invention, machine learning is a tool used to generate a digital network that links tasks and detected data in each individual. Mathematically, some forms of machine learning can be approximated by statistical methods. Machine learning techniques include supervised learning, displacement learning, unsupervised learning, or reinforcement learning. Several other classifications may exist, but most can implement the following concepts.

[0080] "Unsupervised machine learning" potentially includes methods such as cluster analysis, which can be used to identify links between data, such as internal links between data, family history, data from physical examinations (irregular pulsations), data from sensors, electrical data (irregular atrial signals on ECG), structural imaging data (hypertrophied left atrium), biomarkers, and genetic and histological data.

[0081] "Supervised machine learning" includes methods that can classify a set of inputs that appear to be related or unrelated into one or more output classes without explicitly modeling the inputs, i.e., without adopting potentially inaccurate ("biased") mechanism hypotheses.

[0082] Reinforcement learning is a form of machine learning related to psychology that focuses on how software agents perform actions in a specific environment to maximize cumulative rewards. Reinforcement learning is often used in game theory, operations research, swarm intelligence, and genetic algorithms, and is also known by other names such as approximate dynamic programming. One implementation in machine learning is through the formulation of a Markov decision process (MDP). Reinforcement learning differs from supervised machine learning in that it does not require matched inputs and labeled outputs, and (unlike supervised learning, which can correct for suboptimal rewards through algorithms such as backpropagation in perceptrons) actions that result in suboptimal rewards are not explicitly corrected.

[0083] "Medical device" means an instrument, apparatus, implement, machine, device, implant, in vivo reagent, or other similar or related article, including component parts or accessories, intended for use in the diagnosis of disease or other condition in human or other animal, or in the treatment, mitigation, therapy, or prevention of disease.

[0084] A "neural network" refers to a self-learning network of interconnected nodes loosely modeled after the human brain, which can be used to recognize patterns. Artificial neural networks can be combined with heuristics, deterministic laws, and detailed databases.

[0085] A Personal Digital Phenotype ("PDP") is a digital representation of an individual's health or disease, which may or may not include cellular, genetic, or other omic data calibrated against the individual's observed response to therapy. A personal PDP is matched to the most similar PDP from a digital taxonomy of data from a larger group. Thus, PDPs enable personalized medicine that does not consider only the statistical majority of individuals. The data elements used to generate a PDP can represent an individual's health status, weighted by the likely contribution of similar age, sex, and survival disease to the individual's disease or health. PDPs are matched by algorithmic analysis that considers the calculated or documented probabilities of their impact on health or disease. This can be done using deterministic algorithms or machine learning. For example, a heart rate rhythm phenotype would primarily consider heart rate and electrographic signals (surface ECG and intracardiac). Relatively large mathematical weights would be assigned to these data elements. Other (indirect) organ system data streams may include changes in respiratory rate in association with heart rate (i.e., lung sensors) and changes in neuronal firing in association with heart rate (i.e., neuronal function). Other data elements include abnormal cardiac ejection fraction and the location and presence of structural abnormalities of the heart. In addition, historical data, including age, sex, and family history, may influence the overall digital personal phenotype.

[0086] As used herein, “population data” is a determinant of the accuracy of the method of the present invention. If an index individual differs significantly from the reference population, the digital taxonomy may not adequately represent that individual. In this case, the data would primarily be derived from conventional data on the individual at the time of determined health and determined illness. If the reference population is broad but has other limitations, such as not having clearly phenotyped or clearly labeled data elements, the taxonomy would also be unhelpful. Therefore, an ideal dataset would have clearly labeled data streams and individuals, such as index individuals, that can be partitioned to generate a digital taxonomy. Simply providing “large” or “massive” data is not sufficient.

[0087] The term "sensor" includes devices capable of detecting biological signals from an individual's body. A sensor may be in direct contact with the body or at a distance. When applied to a group of individuals, a sensor may represent all or part of a defined population. Electromagnetic sensors may detect electromagnetic signals related to electromyography (EMG), electroencephalography (EEG), electrocardiogram (ECG), nerve firing, or other emitters. The term "sensor" may be interchangeably used between "electrode," "electrode catheter," or "catheter," particularly when describing specific cardiac applications of the invention where electrical information is detected. Electrical sensors may also detect bioimpedances, such as conductance across the skin that decreases when a person sweats, which may occur when the sympathetic nervous system is dominant. Sensors may also detect other chemical changes via current flow. Furthermore, sensors may also include devices that detect temperature, such as thermistors or other thermal detectors. Sensors can detect light, such as changes in the color of reflected light from pulsating cardiac activity (photoplethysmography), or ambient oxygenation (e.g., cyanosis, anemia, vasodilation in the skin). Sensors can detect sound via microphones, which can be used to detect sounds from the heart, lungs, or other organs. Sensors can detect other vibrations or motions via piezoelectric elements. Sensors can directly detect chemicals using specialized sensors for hormones, drugs, bacteria, and other elements, which are usually converted into electrical signals on the device. Examples include chest wall motion from respiration or heartbeat, motion detection of chest wall vibrations from certain types of respiration (e.g., loud obstructive breath sounds), or cardiac sounds (e.g., the so-called "thrill" in medical literature). Respiratory sensors can detect movement of the chest wall, abdomen, or other body parts associated with ventilation, acoustic data (sounds) associated with respiration, or oxygenation associated with respiration. Chemical sensors can detect chemical signals on the skin or other membranes that reflect biochemical reactions familiar to those skilled in the field of biochemistry, such as oxygenation and deoxygenation, metabolic acidosis, stress, or other conditions.Furthermore, the sensors may also detect images using cameras or lenses that require contact from fingerprints or other body parts, or detect movement from specific muscles, or detect iris dilation or vibration from a photosensor in a contact lens. Position sensors can identify the position and changes over time (including gait) of various body parts, or the detection of contact of the position of a specific body part at a given point in time or over time (e.g., facial ptosis, facial tics, or other unusual movements). In exemplary embodiments of the system of the present invention, multiple sensors may be used in communication with a central processing unit, or they may form a network linked via BLUETOOTH®, WiFi®, or other protocols to form an Internet of Things (IoT) of biological sensors.

[0088] A “signal” includes electronic, electromagnetic, digital, or other information that can be detected or acquired. A detected signal is detected (i.e., recorded) in its natural form and invariant state without transformation. Detected signals are typically biological signals. Detected signals can be detected by humans (e.g., sound, vision, temperature), but also by machines such as microphones, audio recorders, cameras, and thermometers. Acquired signals are detected in a transformed state, such as an ECG recording. Such signals may be biological because cardiac bioelectricity generates an ECG, or non-biological signals such as vibrations detected after the application of sound or ultrasonic energy, or tactile signals transformed from detected electrical, sound, or other signals. Signals can be detected through physical contact with a sensor.

[0089] "Smart data" refers to application-specific information obtained from a source that can be used to identify normal or abnormal functioning in an application, and / or to function based on that identification. Therefore, smart data differs from the term "big data." Smart data is tailored to individuals and to address specific tasks or elements, such as maintaining health and agility, or detecting and treating diseases like sleep-disordered breathing. The tailoring is based on knowledge of what the system may influence the task at hand. Such knowledge may be based on psychology, engineering, or other principles. In contrast, "big data" often focuses on extremely large datasets with the aim of identifying statistical patterns or trends without individually tailored links. In machine learning terminology, smart data can be a consequence of supervised learning of datasets to known outputs, while big data simply refers to the volume of data without necessarily implying any knowledge of the significance of a particular dataset.

[0090] In this specification, “sources” for cardiac rhythm disorders are used to indicate therapeutic targets. In the biological literature, an electrical source or electrical driver indicates a focal point from which radio waves are emitted outward, or a reentrant, rotational, or rotor-like circuit from which activation is emitted. These electrical sources drive rhythms such as focal atrial tachycardia, reentry in ventricular tachycardia, or atrial flutter. Sources may also drive atrial fibrillation or ventricular flutter or ventricular fibrillation. In the clinical literature, different definitions may apply, and other targets that are effective therapeutic targets for cardiac rhythm disorders may be identified. These include small channels in viable tissue within low-voltage fibrotic or scar tissue, areas of complex signals, and areas of high frequency or high-rate activation (including high-dominant frequencies). Other electrical targets include areas of conduction slowing where outline lines of activation ("isoclones") that can be detected in sinus rhythms or at relatively rapid rates, including during pacing, are present.

[0091] Other biological terms, such as heart failure, tidal volume, sleep apnea syndrome, and obesity, have their own standard definitions.

[0092] The following description and accompanying figures provide examples of applications of the system and method of the present invention for generating personal digital phenotypes (PDPs) of health and disease in comparison to digital taxonomies, enabling personalized methods for detecting and treating areas of interest for biological rhythm disorders. The examples described herein are illustrative only. Further modifications and combinations can be formed by utilizing the principles of the present invention disclosed herein, as will be apparent to those skilled in the art.

[0093] Figure 1 shows an exemplary system for defining a personal digital phenotype (PDP) in comparison to a digital taxonomy for personalizing the determination of an individual's health or disease, including the identification of a target area for electrical rhythm disorders and the subsequent delivery of personalized therapy to such area. Input / output (I / O) data 100 includes input signals related to the individual, generated by one or more sensors 105 which may be located outside and / or inside the body. Therapeutic devices 110, such as therapeutic catheters, may be temporarily inserted or implanted. Implantable devices may be explicitly inserted to generate / maintain a PDP, to provide health maintenance, or to provide continuous therapy. Further inputs 125 include clinical data, patient history, physical data, and / or data from an electronic medical record system. The device may communicate via the Internet of Things (IoT), with timestamped data transmitted to an input unit 130 via wired or wireless means. Data can be transmitted continuously, nearly continuously, in real time, nearly real time, in some other format, or in a combination of signals acquired over time.

[0094] Several types of sensors may be used, including photosensors, piezoelectric, acoustic, electrical resistance, thermal, accelerometer, pressure, flow, electrochemical, or other sensor types, which can be used to measure chemical, optical, skin activity / moisture levels, pressure, motion, and other parameters related to PDP generation. The selection of appropriate sensors will be obvious to those skilled in the art. Sensors may be interchangeable or fixed in each embodiment. The selection of appropriate component values ​​(resistors, capacitors, etc.) and circuit performance characteristics, as well as the addition of supporting components / circuits (filters, amplifiers, etc.), are within the scope of the art in this field and are therefore omitted from this specification.

[0095] In this exemplary implementation of the system, signals are detected from the heart and may be of several types. Electrical activity of the heart can be detected directly using sensors that can be placed on the heart, either in contact or non-contact, that detect the magnetic field generated by the electrical activity of the heart, or on the body surface, near other body regions (e.g., esophagus, bronchi and airways, mediastinum), or non-contact sensors such as magnetocardiograms. Sensors can also measure cardiac motion or the presence of ischemic areas by detecting cardiac motion or the movement of blood through it. Cardiac motion can be detected from changes in electrical impedance, from movement on body regions, from cardiac motion (ballistic electrocardiogram), or from changes in electrical impedance, for example, by using non-electrical devices such as echocardiography or ultrasound. Blood flow can be detected using known methods such as Doppler echocardiography, 4D flow MRI, or imaging methods that tag carriers such as red blood cells. In various embodiments, these sensors can be used individually or in different combinations, and furthermore, these individual or different combinations can be used in combination with sensors implanted in the patient's heart. Signals can be detected without physical contact with the sensors. An example of such a method includes detecting a heartbeat from an electromagnetic field emitted from a magnetocardiogram (MCG) or from an infrared signature of cardiac motion. Other non-invasive detection signals may include auditory respiratory sounds or cardiac sounds from a highly sensitive external microphone. Signals from one or a combination of the described sensors are transmitted to the input unit 130 via wired or wireless communication.

[0096] Neural activity is another sensing signal that can be used in the present invention in conjunction with indices such as firing rate and period, diurnal and interday cycles, type and pattern of neural firing, and the spatial distribution of these measures. In one embodiment, non-invasive recording is produced from a skin patch, while other embodiments may use an electroneurogram (ENG) in which electrodes are inserted into the skin to record from nearby nerve tissue. Invasive methods may be suitable for hospital treatment but are not very suitable for continuous recording or consumer use. When sensors are positioned within different regions, it is possible to record from nerves in different regions; for example, electrodes on the chest may measure neural activity related to the heart or its nerves, and electrodes on the neck or head may measure neural signals including those controlling the heart or other locations familiar to those skilled in the art.

[0097] Pulmonary (lung) function activity is another type of signal that can be detected and can change independently of the heart or as a result of changes within the heart. This can be measured by sensors for respiratory sounds, chest wall motion, oxygenation, phrenic nerve electrical activity, or other sensors.

[0098] The therapeutic tool or effector 110 enables the maintenance of health or the treatment of disease. It may have an electrical stimulator, a thermal stimulator, an optical stimulator, a chemical release device (such as an infusion pump for drug therapy), or other stimulators. In one embodiment, it is an ablation catheter for treating cardiac rhythm disorders, introduced via a vascular access labeled 120.

[0099] The input data 125 is used for personalization and to associate the indicator individual with population data represented as a disease taxonomy. The input data 125 may include demographics, laboratory data, chemical reaction data, and image data. For example, some data inputs for a person may include "static" stored data such as date of birth (age), sex, and race. The input data may also include near real-time data such as patient exercise from a separate device (e.g., treadmill in a building, motion sensor), patient ECG or cardiac information from a separate device (e.g., hospital telemetry, ICU bed monitor), respiratory sensor data, time-course or time-series data from a separate device (e.g., periodic blood glucose values ​​from a glucomometer), or other data inputs. The input data may also include an index of familial predisposition to disease (Mendelian or non-Mendelian), identifiable gene loci, weight changes, or sensitivity to toxins such as tobacco or alcohol. The input data, along with a timestamp, is transmitted to the input unit 130 via a wired or wireless connection.

[0100] The input unit 130 is a data hub, which may be a physical device or cloud-based interface for multiple digital data streams transmitted to it. The data can be time-stamped and maintained separately as real-time data (streaming) or stored data (history).

[0101] Conventional cloud-based computing / storage 135 can optionally be used to store data as an additional or backup to what is stored on the device in 105 or 130 and / or to perform data processing. Raw data and analysis results stored or generated within the cloud-based computing / storage can be separately transmitted to external servers connected via the Internet. For example, independent recipients may include research facilities, clinical trial administrators, or other recipients authorized by patients.

[0102] The population database 140 provides a standard for data from individuals for this indicator and may include stored data from the population, time-varying streaming data, and optional streams that can be crowdsourced.

[0103] The process controller 145 is programmed to execute algorithms that include not only deterministic formulas but also neural networks (or other learning machines) and other distributed representations in order to generate a personal digital phenotype (PDP) 150 in comparison with a digital taxonomy 155 that classifies data for an indicator individual from a preceding point in time based on a population database 140. In one embodiment, machine learning is used to process input data, to generate and learn classifications that link complex physiological and clinical inputs to outcomes at the patient level (i.e., to generate PDPs), to compare these with quantitative traits (a digital taxonomy of health or disease) in the relevant population, and to proactively design "optimal" or "personalized" therapies based on specific individual characteristics in relation to preceding observations in that individual, preceding observations in a comparison population, or mathematically estimated predictions.

[0104] The comparison between the PDP 150 and the digital taxonomy 155 identifies and / or tracks the patient's state, i.e., health or disease. This is done by calculating deviations from normal in an index individual in a state compared to a predetermined "acceptable range limit" and comparing it to a different population. In one embodiment, this is achieved by detecting a data stream from sensor 105 or iteratively updated data. Data can be entered in a determined "healthy" period or a determined "disease" period for that organ system within that person. The accumulated data supports future learning to validate the PDP. Different states can be detected regarding changing states or grades of health or disease among individuals (e.g., exercise vs. rest). This method differs from current medical practice, in which the "population" ranges of "normal" and "disease" are applied across multiple patients, with little range available to adapt them to individuals. It is this aspect of the present invention that provides "personalized medicine" or "high-precision medicine."

[0105] The conditions identified through comparison between the PDP150 and the digital taxonomy result in personalized disease and health management 160 and therapy guidance 170, which are subsequently transmitted to the diagnostic and reporting unit 165. The diagnostic / reporting unit 165 may be a smartphone app, a dedicated device, or an existing medical device. Briefly referring to Figure 3, a custom-designed smartphone app 470 can indicate sites of termination from digitally acquired personal data from the imaging / mapping system. A sample display panel can display personal streaming data of AF maps generated by a freely available online method operating within the smartphone app. The display panel can provide interactive input with a physician to assist in identifying critical sites for personalized therapy.

[0106] In one embodiment used in cardiac rhythm disorders, the therapy unit controls electrical intervention (pacing) or destructive energy (ablation). The effector device 110 shown in Figure 1 can be activated to ablate tissue to treat the biological disorder in a manner adapted to the personal phenotype. In an alternative embodiment, the therapy unit may be supplied with antiarrhythmic or anti-inflammatory drugs by an infusion pump, or with gene or stem cell therapy. In another embodiment, the therapy can be a mechanical restraint, which can be supplied to improve a stretch that may trigger an arrhythmia.

[0107] Figure 2 schematically illustrates an exemplary system using a personal digital phenotype (PDP) in one embodiment for the treatment of cardiac rhythm disorders. Detected or stored data 200 may include clinical, laboratory, genetic, or other data. Examples of data may include markers of abnormal inflammatory or immunological conditions in the body and biological markers for atrial fibrillation or ventricular fibrillation. Direct measures of inflammatory / immunological equilibrium include, without limitation, the number of inflammatory cells or cytokine concentrations in body fluids or in affected organs. Indirect measures of inflammatory / immunological equilibrium represent the variable effects of inflammation on various organ system abnormalities in static and diurnal scales of body temperature, fluid composition, cardiac rhythm, neuronal firing rate, and electroencephalogram (EEG). Further data may include the detection of abnormal cardiac and bodily neural control that allows modulation of such conditions to maintain, improve, or correct biological rhythms, including atrial fibrillation or ventricular fibrillation.

[0108] The input data is iteratively evaluated in comparison to normal and abnormal values ​​not only for the individual but also for the population, and by controlling interventions and therapies to maintain a normal equilibrium. In this context, signal "detection" goes beyond the conventional collection of raw signals from detection devices and can include data generated from other testing procedures, such as clinical, laboratory, or chemical. As shown in Figure 2, the data 200 can be generated by a device that detects electrical signals, such as an ECG or bioimpedance sensor, combined with a transmitter for transmitting the detection signal to the process controller 285 by wired or wireless transmission. Furthermore, other repositories or repositories of streaming or input data obtained from clinical systems, hospital databases, hospital equipment, or laboratory equipment can interface with the process controller 285.

[0109] The heart 210 can be measured in many ways, including by ECG electrodes 250 applied to the body surface. Electrical or electromagnetic signal sensors, such as electrode catheters in the esophagus 225, electrodes in the right atrium 230, interatrial septum or left atrium 220, or electrodes via the great cardiac veins to the coronary sinuses (electrodes 235, 240, 245), or electrodes to the anterior cardiac veins (electrode 215), access the left or right ventricle or any of these chambers directly. Sensors 215-245 can also detect activation from other areas of the heart.

[0110] Cardiac sensors may be located externally or internally. In some embodiments, one or more sensors may be located externally to the patient's heart. For example, sensor 250 detects cardiac activation via the patient's surface (e.g., electrocardiogram - ECG). Sensors (not shown) can remotely detect cardiac activation without contact with the patient (e.g., magnetocardiogram). Also, as another example, some sensors can derive cardiac activation information from cardiac motion of non-electrical sensing devices (e.g., echocardiogram, Doppler signal of blood flow, red blood cell tagged scan). Such sensors can be classified as “external sensors” to distinguish themselves from catheters and electrodes inserted into or near the patient’s body, i.e., “internal sensors”. In various embodiments, various external sensors can be used individually or in different combinations. Furthermore, these individual or different combinations of external sensors can also be used in combination with one or more internal sensors.

[0111] The process controller 285 receives detection and input data. In some embodiments, the process controller 285 is configured to analyze unipolar signals, and in other embodiments, it analyzes bipolar signals. The process controller 285 uses these detection data streams to generate a personal digital phenotype (PDP) for the individual. This is compared to relevant pre-stored data 265, i.e., a digital taxonomy of arrhythmias. The process controller 285 may have access to population data in the population database 262 (which is also shown as database 140 in Figure 1) to generate the digital taxonomy. As a result, interventions are designed to diagnose or treat the individual based on that individual's personal digital phenotype 270.

[0112] In some embodiments, the process controller 285 analyzes input electrical data to generate a map representing the source or other target of cardiac rhythm disorders, which can then be displayed on an output device. A population database 262 can be used to store intermediate data. The population data 262 can support or assist signal analysis and can store maps of other personal potential target areas or source locations along with known personal digital phenotypes as part of a digital taxonomy.

[0113] Internally, the process controller 285 typically has a digital signal processor. It may also include a graphical or other processing unit for executing machine learning algorithms or other calculations of the PDP to compare against a digital taxonomy to guide the therapy. Other elements may include conventional computing devices, cloud computing, biological computing, or biological-cybernetic devices. Referring briefly to Figure 1, the functions corresponding to the input unit 130 and the process controller 145 would be included in the computational operations performed within the process controller 285.

[0114] The process controller 285 is programmed to implement a functional module (PDP module 260, corresponding to element 150 in Figure 1) for generating a personalized digital phenotype of the heart (rhythm) within an individual, to match or align the PDP with a digital taxonomy (taxonomy module 265, corresponding to step 155 in Figure 1) using data from the population database 262, to design a personalized intervention (design module 270), and to guide the delivery of the intervention 275.

[0115] Next, the personalized therapy designed within module 270 is supplied via supply element 275. Supply element 275 can utilize several effector devices, such as a sensing / ablate multi-electrode catheter 290 or an external energy source 295. Other therapy devices may include direct electrical output, piezoelectric devices, visual / infrared or other stimulation systems, nerve stimulation electrodes, or, in some cases, virtualized data such as avatars in a virtual world interface or elements in a large, queryable database, as well as other effector elements that would be apparent to those skilled in the art.

[0116] In the illustrated embodiment, the therapy may include pacing. For example, commands generated within module 270 cause process controller 285 to pace from pacing module 255. Pacing can be applied through electrodes 250, 215, 230-245, 290, or 295. The therapy may be ablation using energy generated by energy generator 280 to modify tissue. Internal electrodes (e.g., 215, 230-245), a dedicated ablation catheter 290, or an external energy source 295 can be ablated from energy generator 280. Other forms of energy include heating, cooling, ultrasound, and lasers, for example, using appropriate devices controlled by process controller 285 and other modules. The therapy is personalized by supplying to tissue that indicates the boundaries of electrical and / or structural targets determined by PDP. Other therapy units may provide antiarrhythmic agents using infusion pumps, anti-inflammatory therapy (since inflammation can be a proximate cause of arrhythmias, including fibrillation), gene therapy, or stem cell therapy. In another embodiment, therapy can be provided that provides mechanical constraint to improve stretches that may trigger arrhythmias. Therapy using an external energy source 295 can enable a completely non-invasive therapy in which key targets for arrhythmias are identified and then treated without involving an invasive method.

[0117] Figure 3 summarizes an exemplary workflow using a PDP to guide and monitor therapy. Personal inputs utilize two forms: a data stream 400 that can be updated over time and personal stored data 420. The data stream 400 may include detected data from novel or existing sensors 402 such as from wearable devices or consumer products, from implanted devices 404, from invasive sensing signals 406 from existing or dedicated devices including minimally invasive products such as skin, nose, cornea, cheek, anus, or auditory probes, from non-invasive sensors 408 that can provide data including motion and temperature from infrared probes, and from transmitted data 410 such as telemetry from existing medical devices.

[0118] Personalized stored data 420 may ideally include imaging data 422, including detailed coordinates of scar tissue, fibrosis, ischemia, reduced systolic function, and potential border zone tissue; laboratory values ​​424, including not only serum biochemistry but also genetic, proteomic, and metabolic data (when available); demographic data 426; and static data such as elements from patient history, such as the presence of diabetes or hypertension, and left atrial size from echocardiography. Further personalized stored data may include outcome data, such as subjective symptoms, such as whether the patient feels comfortable, such as "healthy" or "not very healthy." Outcome data may also include objective data, such as acute endpoints of therapy, such as fever reduction with antibiotics, or in one embodiment, termination of atrial fibrillation by ablation. Objective evidence may also include chronic endpoints, such as the absence of infection or the absence of recurrence of atrial fibrillation in long-term follow-up. The method of the present invention may also use a combination of population data, including population stored data 449, population data streams 452, and domain knowledge 455, to partition data classes based on health status in order to identify health and disease, and to define a disease taxonomy 446 in order to compare population classes with individuals.

[0119] Step 440 continuously updates the personal digital phenotype from the data stream 400 and the stored data 420. This can be performed periodically or continuously at predetermined points in time. Step 443 compares the personal digital phenotype against an externally determined digital disease taxonomy 446 generated from one or more of the population stored data 449, population data stream 452, and domain knowledge 455. These steps are further detailed in Figure 5.

[0120] In step 460, the therapy is adapted to a personal digital phenotype, and in step 463, the effectiveness of the therapy is iteratively monitored using a data stream in the context of data already stored in the previous step 440. According to the method of the present invention, the phenotype is based on whether a patient is ill, the manner in which they are ill, and a ground torus (label) of the best way to treat them to maintain their health or to cure the disease. These clinically or biologically relevant actions are included in steps 426, 446-455. The present invention acquires new data in steps 402-410 to generate a personalized phenotype by using data types that do not always correspond to data types in the stored data of that individual (420-426) or the comparison population 446-455. Such data types are actively acquired by the system so that the personal phenotype can guide the therapy relatively well and so include data types that include electrical information or cardiac structure.

[0121] Finally, in step 466, an interactive interface for reporting data is provided. Display 470 provides an example of many types of data that can be displayed via an application on a computer or mobile device (in the illustrated example, a smartphone is shown). One implementation of an app written in Swift via Xcode is generated for APPLE® iPhone®. Image 470 shows a sample map of arrhythmias with ablation targets and may also include one or more 3D cardiac images, numerous coordinates, text descriptions, and quantitative scores. The present invention displays the characteristics of each individual's personal digital phenotype with any notification of personalized management and therapy decisions. Step 466 generates a smartphone display 470 of this data, along with an illustration of a smartphone app.

[0122] Figure 4 provides a more detailed workflow for generating and comparing PDPs and digital taxonomies. Step 500 acquires personal data in a time vector [T] having one or more points in time. This includes a data stream 502 (corresponding to step 400 in Figure 3) and personal stored data from chart 504, similar to the personal stored data 420 in Figure 3. This data includes measures of biological or clinical significance, such as acute or chronic outcomes.

[0123] Population data stream 506 (block 452 in Figure 3) is designed to leverage increasingly available datasets. Such streams can include data from individuals with similar or different phenotypes in relation to indicator individuals. To give a few examples, this data could include telemetry data from patients in intensive care units, individuals using similar wearable devices, or soldiers on the battlefield being monitored for various vital signs.

[0124] Data from the stored population database 508 (database 449 in Figure 3) is embedded at a predetermined level of granularity based on similar digital phenotypes. This differs from conventional methods using statistical associations, which tend to weaken, such as a significant association between AF and obesity, failing to account for lean individuals who develop AF or obese individuals who do not. Other levels of patient granularity, such as heart disease or hypertension in patients with AF who may be lean or obese, may also be common among such different patients. Similarly, while atrial scarring in imaging is associated with AF, many individuals with AF have minimal scarring, while individuals without AF also have considerable scarring. Numerous patient matching based on PDP features improves the present invention's ability to adapt therapies for individuals based on known outcomes in similar patients. Statistical associations are also performed using multivariate analysis.

[0125] In step 510 (similar to step 446 in Figure 3), population data from database 508 is integrated, including an index of biological or clinical significance, such as acute or chronic outcomes, which serve as criteria for diagnostic or therapeutic usefulness. The first step is to feature the data in step 514 to address the curse of dimensionality in machine learning. Feature reduction and feature extraction techniques are well known in the art and may vary depending on the type of learning machine used, or may be implemented using different types or combinations of learning machines or other algorithms. Possible operations for feature extraction in step 526 include time-domain mathematical operations, including principal component analysis, averaging, merging, area analysis, and correlation. Frequency-domain analysis includes Fourier analysis, wavelet transform, and time-frequency analysis of fundamental frequency, harmonics, or other frequency components. Polynomial fitting can also be used to represent the data as polynomial coefficients. Other common featureization steps can be used, using widely available libraries such as TSFresh (Time Series Feature Extraction). In parallel, the population data is parameterized in step 534 by using similar or different behaviors.

[0126] Step 518 partitions the data into classes representing digital “disease phenotypes” within individuals or digital “disease taxonomies” within a population. The goal is to relatively well separate the data—clinical data, but also granular invasive data points and laboratory studies—into partitions of individuals with seemingly similar but different outcomes from a given therapy (success versus failure). Mathematically, this is accomplished by constructing a “hyperplane” in the k-parameter space that separates patients with one outcome from those without. For example, in the case of embodiments of arrhythmia, it remains unclear why “paroxysmal” AF in two patients with similar profiles may respond in entirely different ways to drug therapy or pulmonary vein ablation. Personal phenotypes encode observations from multiple patients into crowdsourced partitions ("digital taxonomies") about why AF in some patients reflects source or driver regions, structural abnormalities, neurological components, and metabolic comorbidities including obesity, while in others it does not. These factors are not predicted by the conventional taxonomy of “paroxysmal” or “persistent” AF. By using PDP, estimation methods, including statistical and machine learning approaches, can be used to compare the data to a baseline population in order to estimate best management.

[0127] Step 530 partitions the data by classification method. Partitioning can be performed using many techniques known in the art, without limitation, including cluster analysis, and other types of unsupervised or supervised learning methods, including support vector machines (SVMs), logistic regression, naive Bayes, decision trees, or other methods. This partitioning is performed for personal data (step 518) and, in parallel, for population data (step 538). Note that the partitioning technique used may differ for each step.

[0128] Cluster analysis, a known unsupervised learning technique, can be used in step 530 to group unlabeled data (e.g., data streams from multiple sensors, input data, etc.) into sets of items that are familiar to one cluster but unfamiliar to others. This can occur even when there is no obvious natural grouping, which is often true in these applications because typical phenotypes rarely include clinical data, imaging, and continuous data streams. While clustering is a powerful tool in this invention, any ambiguity in identifying the initial cluster pattern may lead to bias, as the final cluster pattern depends on the initial clusters. The results of the clustering are verified later in step 558.

[0129] Step 546 generates a disease taxonomy from population data by using a mathematical model to integrate the data stream from the population, stored data (database 508), a data reduction scheme (step 534), and data partitions (step 538). Data from the domain knowledge database 574 is incorporated to filter the mathematical relationships. For example, mathematical weighting can be increased for older men with common coronary artery disease, but minimized for rare cases such as breast cancer in men, or rare cases such as AF in young children. Such conventional domain knowledge can be obtained from epidemiological data and population statistics, and may also be obtained from stored population data, and can be easily converted into mathematical weightings.

[0130] Step 550 generates a population digital phenotype from the disease taxonomy. In other words, the population digital phenotype corresponds to data partitions that form self-consistent disease classes from quantitative data. These may be clinically obvious or clinically ambiguous, such as the link between low magnesium levels and atrial fibrillation in some studies. Each of these partitions will be statistically represented by a confidence interval and used for comparison against the personal digital phenotype.

[0131] Step 522 generates a prototype personal digital phenotype by using the personal disease partitions from Step 518 as a basis. Step 542 compares the personal digital phenotypes (PDPs) to find the best-matching population digital phenotype. Candidate personal phenotypes are assigned in Step 554 by a matrix of vectors [P], which have multiple data elements, data types, some ordinal number, some vector properties, and some time dependency.

[0132] In some embodiments, supervised learning is used to enhance the digital phenotype to predict defined outcomes. This requires feature selection, network architecture selection, and appropriate data for training and testing.

[0133] The key feature is that by intentionally generating sparse input "vectors," machine learning is identified and "tuned" to avoid overfitting (i.e., poor generalization to inputs that will not be observed in the future). The present invention removes redundant features and maximizes the diversity of input features in order to make the underlying input data more extensive.

[0134] In one embodiment, features are grouped into three types: (a) conventional clinical variables (demographics, comorbidities, biomarks), (b) electrical signals (12-read ECG and intracardiac signals, whose signal-processed parameters can separate AF phenotypes), and (c) without limitation, 2-D echocardiographic images (atrial shape), 3-D CT data (shape), 3-D MRI data (fibrosis area, shape), and 3-D electroanatomical shells of voltage and composite electrogram distributions generated in EP studies. Clusters (unsupervised learning) can be used to assist in data reduction and can be used as further input features. Filtering and regularization are used to understand the importance of features. The method according to the present invention removes variables that are not related to the response class during training. One method uses LASSO (least ablolute shrinkage and selection operator), which combines the advantages of filtering and wrapping methods to minimize prediction errors, and includes variables that contribute to regression analysis in the final model.

[0135] Missing data within each feature group will be processed by inserting (inputting) a) the median, b) a predicted value using multiple attributions (statistical techniques), c) a predicted value for that data type from the literature, and d) a constant representing the missing data. Each method will be compared during the training of various network models.

[0136] Input image and signal formatting. The present invention formats each 3D MRI, CT, or pseudo-colored atrial electroanatomical image (distribution of specific types of electrograms, such as voltage, representing anatomical structures) as a 3D matrix. Time-series data (12-read ECG, bipolar coronary sinus electrogram, unipolar intracardiac electrogram) will be processed before being input into the final network by using feature extraction, cluster analysis, and pre-processing networks.

[0137] The results used to train phenotypes will vary depending on the application. In the case of embodiments of cardiac rhythm disorders, several results can be used to train phenotypes. One result may be high-voltage versus low-voltage (<0.1mV, for example) electrogram signals, where phenotypes associated with high-voltage signals may have relatively superior therapeutic outcomes. Another potential result is the presence of a clean spatial map of AF showing a consistent rotor or local source / driver, where these sites may be effective therapeutic targets. Another desirable result is the termination of AF by drug therapy or ablation, or long-term success from drug therapy or ablation. All of these can be determined retrospectively within a reference population to form a digital taxonomy, and then used to identify matching personal phenotypes.

[0138] In one embodiment, supervised learning is used, typically implemented as an artificial neural network ("ANN"), to represent diverse input data and data streams for individual individuals and populations. An ANN typically has three elements. First, a pattern of connections between different layers of nodes (artificial neurons), each forming a network of a variable number of layers, each containing multiple nodes / layers. The implementation can be as simple as many other designs, including perceptrons, adaptive linear networks, or deep learning networks. The actual network design may be adaptive to the complexity of the particular task and data partition. Second, the connection weights between nodes can be varied and updated according to several learning rules. Third, an activation function that determines how each weighted input is transformed into its output. Typically, the activation function f(x) is a complex of other functions g(x), and these other functions g(x) can also be expressed as a complex of other functions, i.e., a nonlinear weighted sum, i.e., f(x) = K(Σ i wi g i (x) may be used, in which case K (activation function) may be a sigmoid, hyperbola, or other function.

[0139] Various connection patterns, weights, and mathematical activation functions can be selected, and various update functions are possible for any embodiment. A particular form is optimal for different disease states and tasks. For example, a device for detecting abnormal heart rates in known atrial fibrillation would not be as complex as a device for identifying ablation sites in atrial fibrillation to predict the onset of atrial fibrillation, to predict the worsening of heart failure, or to predict the onset of coronary artery ischemia.

[0140] Alternative forms of learning include supervised and unsupervised methods, including linear logistic regression, support vector machines, decision trees in "if-then-else" descriptions, random forests, and k-nearest neighbor analysis. Such formulations can be applied independently or as part of machine learning to augment or generate boundaries between desired decisions, such as the presence (or absence) of atrial fibrillation sources, or other associations linking input data to personal clinical or physiological outcomes. Several other forms of machine learning are applicable, and these will be apparent to those skilled in the art.

[0141] Various connection patterns, weightings, node activation functions, and update methods can be selected, and a particular form may be optimal for different applications depending on the data input. For example, imaging inputs and a continuous series of data (e.g., electrogram signals) may be represented by different networks, and in embodiments with large training data, they may be optimized for each dataset within a given reference population. Thus, depending on the application, the present invention can be adapted to best represent EEG data, cardiac and respiratory signatures, weights, skin impedance, respiratory rate, and cardiac output. A recurrent neural network is a data structure that enables analysis of the manner in which the network stores its trained conclusions. Manually designed scalar features (e.g., clinical data elements) can be embedded by using fully connected layers. Featured time series (i.e., 12-read ECG, or so-called "electrograms" from inside the heart) are processed via a convolutional neural network. Standard techniques of dropout, batch normalization, and hyperparameter tuning are used to avoid overfitting.

[0142] A key characteristic of machine learning approaches is that they do not require prior knowledge of the details of human pathophysiology; instead, they learn patterns in detected signals and input data in health, and deviations in disease. Therefore, they are highly suitable for personalized medicine where current mechanistic hypotheses are not optimal.

[0143] Step 558 determines one or more candidate phenotypes that can be validated, i.e., predict any reliable outcome measure. In the case of AF, this may be the site where ablation terminated AF. In the case of coronary disease, this may be a clinical constellation that predicts severe stenosis of the epicardial coronary vessels, i.e., an advanced coronary risk score. By extension, such candidates can be defined for non-cardiac diseases. If no matching is achieved, the process of generating PDP 522 is repeated, or the acceptable tolerance range X (block 562) is expanded. If a matching is achieved within the acceptable tolerance range (vector X), the candidate becomes the personal digital phenotype P within the tolerance range X at time T in step 566. The phenotype is then used to update the individual's personal history data in database 570 in light of the labeled results used to validate the phenotype. This step is used not only to validate the clusters defined in the preceding steps, but also for supervised training.

[0144] Figure 5 illustrates how personal digital phenotypes define the interface between health and disease. Step 600 acquires each PDP at time T and examines the dominant signal types within the phenotype 603. Mathematical and network analysis 606 is first used to identify anomalies in the state compared with the stored personal phenotype in database 609, for example, data from a determined time when the individual feels comfortable or uncomfortable (symptoms), or has objective evidence of disease, such as the presence or absence of AF. Step 612 generates a personal health or disease portrait based on this analysis.

[0145] Step 615 determines whether the portrait from Step 612 represents health 618 for the individual within an acceptable range. If the result is “outside the range” of the individual’s health, the individual may enter a potential disease state in Step 621, and mathematical and network analysis 624 is performed and, similarly, compared with the population disease taxonomy 627 to determine whether the individual’s abnormality falls within the “outside the health range” of the population. In comparison against the fixed and variable definitions of the population of statistically defined abnormalities (Step 630), the invention now asks in Step 633 whether a disease exists. If “yes,” a disease 639 is declared; otherwise, careful observation of the patient continues in Observation Step 636. In either case, the process is repeated for continued monitoring.

[0146] Figure 6 outlines an explainability analysis for the method of the present invention, which aims to identify the most appropriate data components for a particular disease or health aspect. This enables a “disease-specific PDP.” Explainability also addresses the criticism that some data science techniques, including machine learning, can be “black boxes.” As will be discussed later, several methods for explainability are used, including expert domain knowledge to feature data and techniques such as LIME and Grad-CAM. Step 700 analyzes diverse data types in the PDP. Then, steps 703, 706...709 examine each data element (input data or data stream) to determine what contributes to the decision-making about that disease process or health aspect in step 712. Several methods (step 715) can be used to determine one or more components that dominated or otherwise contributed to the “disease” vs. “health” classification.

[0147] Several explainability (or interpretability) techniques familiar to those skilled in the art can be used. One involves the use of attention layers in recurrent neural networks. Alternatively, LIME (Local Interpretable Model-agnostic Explanation) can be used to explain predictions by approximating an interpretable model. LIME is used for one-dimensional data such as ECG or electrical signals (electrograms) from within the heart, numerical features, or images. Another approach is Grad-CAM (Gradient-weighted Class Activation Mapping), which identifies the most important nodes as the maximum weight multiplied by the backpropagated pooled gradient downstream of the final convolutional layer. Another embodiment defines features that must or must be part of the model (e.g., the size of the atrial driver region, or ventricular conduction velocity, or the spatial extent of fibrosis in the human heart), including spatial domains in images, to ensure that the model does not converge on irrelevant concepts, enabling a fitted interpretation of domain electrophysiological "concepts." One example of this is the TCAV (Testing with Concept Activation Vector) approach. This involves examining certain features (e.g., the size of the AF driver region) that may or may not be part of the model, thereby enabling the invention to adapt explainability to an accepted “concept.” As another example, an interpretable model can be used to test predictions of AF outcomes (e.g., success or failure of ablation), such as the presence of fibrosis near the right atrium. This approach attempts to ensure that the numerical model is suitable for prediction and that the model does not converge on irrelevant concepts. Thus, explainable features that predict outcomes are quantitatively identified. Clinical theoretical explanations can be added later through domain knowledge, such as the determination that obesity predicts a negative outcome from ablation or drug therapy, while hair color may not predict a positive outcome.574Furthermore, data on the population for which class IC antiarrhythmic drugs (AADs) may be used may be included.

[0148] A key feature of digital taxonomies is their coding of inconsistent cases, i.e., situations where a neural network fails to predict the actual outcome. For example, in a patient with a failed ablation therapy whose profile includes atrial damage on MRI, the invention would be trained to link the location of the damage to the location of the ablation injury and the outcome. One potential output from the trained network would be that an ablation that misses the area of ​​damage may produce a poor outcome. Domain knowledge (physiological interpretation) is used to provide plausibility for any trained network in order to mathematically ensure that no unbelievable (or impossible) links are constructed, and therefore to modify the network. This combined mechanism / machine learning approach is a novel strength of the invention, often omitted from machine learning systems that do not check data representations against known domain knowledge. Errors to be avoided include adversarial examples, such as in image recognition, where a change in one pixel can change the classification from "cat" to "dog." This invention prevents such trained networks from generating errors in the medical field, where errors must be minimized.

[0149] Therefore, the methods of the present invention concern the development, testing, and modification of increasingly interpretable data structures. The models combine statistical analysis with expert interpretations of success / failure cases. Simple statistical tests and linear models can help identify correlations between different variables in a system, but they may not be able to capture the complexity and nonlinearity underlying these studied complex clinical problems. Decision trees, such as CART, can allow for a relatively large interpretability of the importance of each extracted feature from layers of the network. The input to the decision tree would be extracted features from images and time-series signals.

[0150] Another method in the present invention is referred to as "network clamping." In step 715, the network 712 is rerun to identify abnormal input combinations from the network's trained baseline "healthy" version, where the inputs are out of range, one by one or in batches, and therefore the network comes closest to summarizing a "disease state."

[0151] These steps contribute to a determination regarding the presence or progression of a particular disease or health aspect, i.e., they are evaluated in step 718 to identify the constellation of data elements best suited to this process in step 721. This “disease-specific PDP” is used for each specific embodiment and is updated within the personal database 724.

[0152] Figure 7 shows an exemplary workflow for using PDP to manage and treat atrial fibrillation. Steps 800, 810, and 820 constitute the first triage level and are personalized for the specific patient. Steps 830–870 personalize AF mapping, map interpretation, and ablation.

[0153] Step 800 inputs non-invasive signals of AF. These can include ECG data from a standard 12-lead ECG, surface potential mapping (ECG imaging, also known as ECGI), which may use up to 200 body surface leads, non-invasive structural imaging of magnetocardiography (MCG), and other features that may be acquired before invasive examination. Step 810 compares these non-invasive data with the elements of PDP associated with this arrhythmia in this individual, i.e., the “arrhythmic PDP” determined in Step 805, as outlined in Figure 6. Step 820 outputs the decision from this analysis.

[0154] Step 800 provides options for disease prediction, where the technique of the present invention identifies phenotypes that may be at risk due to a specific pattern of structural anomalies marked by low voltage, or potential anomalies in delayed, improved magnetic resonance imaging, although AF is not explicitly stated. In this example, the present invention provides AF prediction. Ideal input data in this example may have fine-grained imaging data or fine-grained data on low-voltage areas indicating MRI anomalies, so as to enable non-invasive detection of structural risk profiles by the network to provide potential targets for prediction or therapy. Therapy may include ablation to connect these areas of injury or fibrosis.

[0155] The output of step 820 is quantitatively determined within the individual by non-invasive data from step 800, disease-specific PDP (here, for arrhythmias), and digital taxonomy. In the case of a particular embodiment of AF therapy, the outputs include lifestyle changes, drug therapy, and ablation. The system of the present invention quantitatively assigns a score to each output using the steps outlined above in the sequence shown in Figures 4 and 5. The lifestyle changes output is assigned a relatively high score in patients with correctable factors such as a large body mass index, poorly treated diabetes, a sedentary lifestyle, and excessive alcohol consumption. The drug therapy output will be assigned a relatively high score in relatively older patients who do not have heart failure but have a previous failed AF ablation. These characteristics are based on epidemiological data, some of which are known to those skilled in the art in the context of PDP and digital taxonomy for AF. For example, if non-invasive data indicate a significant AF area near the pulmonary veins or in other areas suitable for ablation, ablation will be assigned a relatively high score. If PDP indicates a good candidate (small BMI, paroxysmal AF, no prior ablation), but non-invasive data do not indicate a significant area near the PV or in other areas suitable for ablation from the digital taxonomy, ablation will be assigned a relatively low score.

[0156] If step 810 identifies that the ablation is likely to be successful, invasive recording is initiated in step 830. Here, the first step is to acquire data from inside the heart, which is typically done during AF using an invasive catheter such as a catheter, as described later. Alternatively, non-invasive data 800 may also be used to provide this data. Step 840 determines the type of ablation that is likely to be successful. If a large score has not been assigned to a PDP-fitted ablation to a specific source region of interest, in step 845, the present invention increases the score to consider only anatomical ablations. This would include pulmonary vein isolation, as well as other anatomical targets, less frequently including posterior left atrial wall isolation, or, in a particular patient, mitral, superior, inferior vena cava, or other lines fitted by PDP.

[0157] If the ablation matched by the PDP is determined in step 840 to have a success probability exceeding a predetermined threshold, steps 850-870 are performed to guide and deliver the therapy.

[0158] First, step 850 examines each region of the target. Analysis of electrical signals based on PDP focuses on identifying areas of the target that may be drivers, such as areas of rotation or local activity, low voltage areas suggesting scarring, or other areas of the target. The size of these areas is also identified from intracardiac recordings (invasive or non-invasive) to appropriately fit the size of mapping and therapeutic tools.

[0159] In some embodiments, regions are identified one by one from a small mapping catheter that provides high-resolution recording. In step 855, the signal from the detection tool (AF mapping catheter) is analyzed to determine the direction in which it should move toward the target region (e.g., toward the source).

[0160] Step 860 determines whether the AF mapping catheter overlays the critical area of ​​interest. Catheter size is important for evaluating proper positioning and is selected using PDP to adapt the procedure to one or more expected sizes for the patient. If the mapping catheter does not overlay the critical area, the process is performed to guide navigation.

[0161] In step 865, if the mapping tool overlays a critical area, this area is now targeted for therapy. In some embodiments, the mapping catheter also has the capability to supply ablation energy so that this is performed seamlessly. In other embodiments, the deployment of a separate ablation tool will be required.

[0162] Step 870 assesses the response to the therapy, specifically whether the area of ​​concern has been eliminated. If not, the therapy is repeated.

[0163] The process returns to step 850, which navigates to the target area and ablates it until all of it is removed. The total number of areas to be treated, along with the predicted number of areas based on the PDP, is determined in real time by the electrical signals obtained from steps 800 and / or 830.

[0164] In another embodiment, all target regions are identified simultaneously using global mapping from a basket catheter or reverse approach, and navigation is applied only to the therapeutic tool and not to a wide-area mapping catheter.

[0165] Figure 8 summarizes a personalized therapy for one embodiment of ablation therapy. On the left side of the figure, a detection tool (mapping catheter) 880 is shown at a certain distance from the area of ​​target. The system analyzes radio waves to determine whether the mapping catheter is overlaying the area of ​​target, and in this case, it determines that it is not. The system then provides navigation information to guide the catheter toward the nearest area of ​​target. This can be displayed on a portable display 890, such as a repurposed smartphone or smartphone app, or on a dedicated medical display unit. Each display unit will include appropriate data security and privacy safeguards in place. The navigation process is repeated as indicated by arrow 892. On the right side of the figure, the mapping tool 895 is shown overlaying the area of ​​target. This is referred to as the “treatment position.” Display 898 now indicates “Optimal position, ablate.” Now, if the mapping catheter contains the ablation tool, ablation can be performed. Otherwise, a separate ablation catheter can be inserted. The process is repeated until the operator determines that a sufficient area of ​​the target has been treated. The number of areas to be treated will be determined by the PDP for this type of patient, in relation to the location and size of the areas.

[0166] Figure 9 schematically illustrates several types of regions of interest (ROIs) for cardiac rhythm disorders that can be identified and classified according to the present invention. These patterns cover the majority of electrical rhythms. ROI 900 indicates rotational activation without fibrillation, which can be observed in micro-reentry from a local site. ROI 905 corresponds to rotational activation within fibrillation, which can be observed in atrial or ventricular fibrillation. This can be detected using the procedure described herein for directional analysis. ROI 920 represents local activation without fibrillation and can be observed in local tachycardia or extra beats from the atrium or ventricle. Local activation in the center of fibrillation (ROI 925) can be observed in atrial or ventricular fibrillation and can be detected using directional analysis. Some activation patterns may not indicate conventional electrical rotation (rotor) or local source. In this case, atypical patterns (ROI 940), such as partial rotations or repetitive activations that may be found near low-voltage or abnormal sites on delayed-type improved magnetic resonance imaging, for example in patients with advanced disease or certain comorbidities, can still be therapeutic targets. Such target patterns can be identified as low-voltage zones or viable channels in tissue within low or borderline voltage regions. When such patterns are found, the system of the present invention will suggest navigation toward this detected target type. Note that ROI 900, 920, and 940 cover small areas of tissue and can be covered by small mapping tools, while ROI 905 and 925 cover relatively large areas of tissue (within fibrillation) and may require relatively large detection tools. ROI 960 represents activity with a disorder that does not have a clear pattern.If ROI960 is identified, no specific navigation guidance is provided, and the system recommends systematic mapping until another region of interest can be detected.

[0167] Figure 10 provides an overview of an exemplary directional analysis sequence. In step 1000, a complex electrocardiogram (unipolar signal) in AF is acquired. Note that the onset and offset of the signal in a complex arrhythmia are often indistinct, and the components may include activation, recovery (liporization), noise, or other features. The method of the present invention uses a number of methods calibrated for monophasic action potentials (MAPs) in step 1020. MAPs provide one of several methods for identifying the actual activation time (onset) and recovery time (offset) in a complex rhythm within the human heart. In any electrical rhythm within tissue, phase 0 (1022) of the MAP indicates the onset time, and phase 3 (1024) of the MAP indicates the offset time. MAP onsets of continuous beats are typically separated by duration, typically 100 ms to 250 ms in atrial fibrillation or ventricular fibrillation, and 200 to 500 ms in atrial tachycardia, atrial flutter, or ventricular tachycardia. Conversely, conventional electrical signals 1030 in complex rhythm disorders often contain multiple deviations from which it can be difficult to distinguish activation onsets (depolarization) or offsets (ripolarization). Such signals are conventionally analyzed using features such as sharp inflection points or large slopes of depolarization. However, signal 1030 shows that such rules often mislabel deviations of indistinct importance as activation onsets. The method of the present invention utilizes analytical techniques such as machine learning to identify activation onset and offset timings from an electrocardiogram calibrated against ground torso, annotated in the MAP recording shown as line 1028. Accordingly, the analytical tools used in the method of the present invention can distinguish onset ("O"), offset (also referred to as termination "E"), and other far-field noise components ("F") from conventional noisy electrocardiograms, such as signal 1030, when MAP is no longer present.

[0168] For any given array of electrodes, a trained system can accurately identify activation onset and offset, thereby enabling accurate mapping of the activation pathway in step 1040 even in cases of complex rhythms. The array of electrodes measures electrical signals from the electrodes on the array. These electrocardiograms have many features including noise. Machine learning models are applied to each electrocardiogram to identify the onset and offset times of each cycle (or beat) and far-field noise. The analysis ultimately results in a sequence of accurate activation (onset and offset) times at each electrode. These activation times can be described as an activation front defined by the following equation,

Number

[0169] The analysis of the propagation flow in step 1060 provides information regarding directionality. The direction of electrical flow in a rhythm can be calculated across the multi-sensor array tool 1064. Retracing the direction of electrical flow provides a path toward the source of rhythmic disturbance, if detected, or toward other electrical targets. In step 1080, even in the case of complex fibrillation rhythms, the type of source is determined, for example, local source 1084 and rotational source 1088, in order to identify where the activation is emitted outward and the direction toward which the path will be. Obtaining this result does not require a global mapping of the entire chamber, which is often not possible and, when possible, only provides a low-resolution mapping. The present invention's method of mapping conduction flow over time can be applied to complex rhythms such as fibrillation, which can be difficult to interpret by standard vector analysis due to the fact that waves tend to change rapidly in space and time.

[0170] Figure 11A shows the steps in a sequence for directional guidance. The algorithm flow begins at time step 1120, which starts at time t. In step 1124, adjacent electrodes are identified as physically adjacent with a known electrode spacing. The sensing device with adjacent electrodes can take many forms. Some examples shown in the figure include the multipole device 1104 - a high-resolution multipolar spade catheter and the basket device 1108 - a multipolar basket catheter. Other multi-electrode devices are known, and the selection of a suitable sensor with a known electrode spacing will be obvious to those skilled in the art. In step 1128, the flow is calculated using the electrode signals integrated at time step t (as previously shown in Figure 10). First, the system spatially interpolates the wavefront Φ with electrodes at known intervals on the array. For each point i along this interpolated wavefront Φ at time t, the system searches within a circle for a point j at the next time step that has the most similar gradient. The system estimates that the activation wavefront has moved from point i to point j at this point and marks this flow by its instantaneous flow vector (propagation over time). Step 1132 iterates the flow (direction) calculation across the region of the electrode array to generate multiple electrocardiograms on a 150 ms window to generate a collection of electrocardiograms. As an exemplary example, array 1145 is shown, in which case the window configured as a pattern corresponds to the positions of electrodes in a multi-electrode catheter. The directionality is integrated across the entire number of available electrodes on the array to determine the average direction of the electrical flow, indicated by a large arrow labeled "integrated direction" in step 1136. The dashed line 1148 indicates the flow used to determine the average direction, which has the ability to describe complex spatiotemporalally changing fibrillation. By guiding the sensor in the opposite direction from the average direction, it will move closer to the nearest source region or other target region.This method improves the accuracy that can be achieved when using a single electrode, which has previously been unable to locate the critical area of ​​the target for fibrillation. The time step is incremented in step 1140 (+t1), and the process is repeated for one or more subsequent time steps over a predetermined number of time steps, or continuously until terminated by the user.

[0171] Figure 11B provides an overview of the process flow for identifying possible arrhythmias. In step 1150, patient data, including physical signals such as intracardiac signals, is integrated with ECG data and clinical variables such as age and sex. In step 1165, multiple mathematical methods can be used to integrate the patient data signals, including correlation coefficients from supervised machine learning models such as multivariate regression or convolutional neural networks (CNNs) or support vector machines (SVMs) trained on specific output labels of long-term results in AF termination or algorithm development. In step 1170, the integrated signals are input to a PDP-based arrhythmia prediction to estimate the source region. In step 1175, directional analysis is used to guide the ablation catheter to the target region, such as the source of the arrhythmia. The ablation catheter is then analyzed in step 1180 to determine the ratio (percentage) of the number of electrodes covered by the region of interest. This is achieved by determining the area of ​​sensors covering the predicted region of interest. In step 1185, it is determined whether the area ratio exceeds the predicted ratio. If the ratio is exceeded, in step 1190, the therapy is applied to this site. If the ratio is not exceeded, in step 1188, the catheter is guided in directions such as right, left, forward, etc., by using the available controls to move the catheter toward a predetermined position until the ratio predicted in step 1185 is met or exceeded.

[0172] Candidate AF ablation targets include electrocardiogram features and comorbidities (e.g., body mass index, diabetes, hypertension), demographics (e.g., age, sex, presence or absence of prior ablation), and, where available, a mathematical combination of genetic, metabolic, and biomarker information. Novel electrocardiogram targets are analyzed beyond “conventional” targets. For example, studies suggest that targets such as repetitive patterns, or transient or localized rotations, or disrupted rotations or localized patterns may be important for maintaining arrhythmias in some individuals. This embodiment of the invention defines these electrogram features by determining which within individual patients may be associated with a favorable outcome. As a result, data from a relatively large number of individuals are labeled and accumulated, which becomes a numerical classification within a digital taxonomy.

[0173] Depending on the patient, the therapeutic target may be rotational or local source / driver or other electrical target, regardless of structure. Intermediate phenotypes may exist in the phenotype of a particular individual (which may be electrical and structural and may change dynamically, for example, with changes in health status). In this case as well, multiple forms of electrical patterns may coexist with such structural elements, and the present invention will store the electrical signals associated with these sites to update personal and population databases. Therapies may include tissue destruction for modulation via electrical or mechanical pacing by surgical or minimally invasive ablation, or using genes, stem cells, or drug therapies. Dosages may include class I substances to reduce atrial conduction velocity and class III substances to extend fire resistance. AF ablation can remove not only tissue but also areas of borderline fibrosis or electrical fragility in the target area. Therapies may also target the related tissues, their nerve supply, or other modulating biological systems in these areas.

[0174] Figure 12 is a flowchart illustrating a process 1200 for treating electrical rhythm disorders according to one or more embodiments. Process 1200 is performed using a device for detecting and treating electrical rhythm disorders. In some embodiments, the device is an ablation catheter, such as a basket catheter, like the basket catheter 1108 in Figure 11A, or other multi-electrode catheters, as described later with reference to Figures 13 to 19. The ablation catheter is configured to perform both the detection of electrical signals in tissue and the treatment of electrical rhythm disorders by ablation energy. The ablation catheter can provide ablation energy in one or more forms, such as electromagnetic energy, freezing energy, etc.

[0175] In step 1210, the device detects multiple signals in the tissue using multiple sensing electrodes on the ablation catheter. In step 1220, the electrical signals detected from the electrodes are used to generate a directional map describing the conduction pathway of the electrical rhythm. The directional map can be generated by inputting the detected electrical signals into a trained machine learning model. In some embodiments, a supervised training method is used, in which case the learning machine is trained based on training examples that have known locations of one or more sources or other target regions of electrical signals and cardiac rhythm disorders in the human heart.

[0176] In step 1230, the directional maps are integrated to determine (a) the location of a source or other target area of ​​cardiac rhythm disturbance within the directional map and (b) one of the guidance directions to a source or other target area of ​​cardiac rhythm disturbance located outside the directional map.

[0177] If the model is successful in determining the location of the source or other target region, i.e., the directional map is substantially aligned with the target, then in step 1240, one or more ablation components on the catheter are activated to modify the tissue at the determined target location. The number of ablation components activated may be based on threshold proximity to the location of the source region used to supply the ablation energy. The directional map can be used to determine the size of the target region to be ablated and to select an appropriate combination of ablation components to be used to modify the determined size of the target. In some embodiments, the ablation electrode may be configured to generate electromagnetic waves that modify the tissue at the target. The ablation energy applied to the electrode may be modulated to generate distinct waveforms. Along with other ablation signal characteristics, the selection of a particular ablation waveform will generally be part of the physician's skill level in the field of art. The selection may be assisted by a lookup table or guided by knowledge gained through the use of machine learning models trained on data obtained from population data. The device may release more irrigation fluid onto the tissue through one or more irrigation holes on the device to prevent overheating of surrounding tissue or other parts of the tissue that are not intended to be altered.

[0178] The irrigation of the electrode grid can be adjusted according to the number, size, and spacing of the electrodes, as well as the expected power supply and the heat generated during use. The primary purpose of the irrigation fluid is to cool the tissue and limit the rise in heat during energy supply. Cooling allows for increased power supply without approaching temperature levels that could evaporate tissue or blood with gas generation or form char or clots. The irrigation fluid also directly washes away small clots or char clumps before they can accumulate. Thus, irrigation improves safety by reducing the likelihood of these problems. A suitable irrigation fluid may include saline with an osmotic pressure similar to that of plasma (i.e., "saline"), half that of plasma ("half saline"), or higher than that of plasma ("super saline"). Alternatives include glucose solution or other electrolyte solutions familiar to those skilled in the art. The flow rate of the irrigation fluid is typically in the range of 2 to 50 ml / min on the catheter during ablation. Increasing the flow rate increases cooling and allows for a relatively greater power supply, while reducing the flow rate results in relatively greater heating. The typical flow rate range for the apparatus and method of the present invention is 5–10 ml / min for the atria of the heart and 15–30 ml / min for treating the ventricles of the heart.

[0179] In further embodiments, the ablation component may consist of a plurality of cryoablation chambers configured to be filled with a suitable coolant. When the outer surface of the coolant-filled chamber is in contact with the tissue, this allows for rapid cooling of the target area for cardiac rhythm disorders. Coolants that may be used in embodiments of cryoablation include nitrogen oxide (N2O) with a boiling point of -89°C, carbon dioxide (CO2) with a boiling point of -79°C, and liquid nitrogen (N2) with a boiling point of -188°C. Other coolants that may be used will be apparent to those skilled in the art. Cardiac and nerve tissues are typically dead at about -100°C, and therefore, temperatures lower than this can ensure even more complete tissue ablation. Disadvantages of very low temperatures, such as liquid nitrogen, are the technical difficulty of maintaining the irrigation fluid in a cooled state and the risk of unintended damage to surrounding areas, including structures adjacent to the heart.

[0180] If, in step 1230, the process has not yet determined the actual target location but determines a guidance direction to the target outside the directional map, then in step 1250, the catheter is maneuvered toward the subsequent location along the guidance direction. Once the catheter has moved to the next location, step 1210 is repeated to detect subsequent electrical signals from the tissue using multiple sensing electrodes. Steps 1220 and 1230 are repeated to continuously determine the guidance direction. If the location has not yet been determined in step 1230, the catheter continues incrementally maneuvering toward the source or other target region in step 1250 until notification is provided that the target has been found. In some embodiments, the positioning of the catheter is controlled by a physician, in which case the system may provide some form of notification, such as a visual display on a display device, an audible tone, or a combination thereof, to inform the physician of the location to move the catheter toward the target in the guidance direction.

[0181] The system of the present invention can be used to confirm whether an electrical rhythm disorder has been successfully treated. Successful treatment inevitably involves correction of the electrical rhythm and removal of a target area that would affect the electrical rhythm. In operation step 1260, the system can be used to detect electrical signals in the treated area using multiple sensing electrodes after ablation of the source or target area. Upon confirmation of success, the procedure may be terminated by a physician or automatically terminated by the system based on confirmation that no further indicators of cardiac rhythm disorder are present. The system can be used to determine whether the cardiac rhythm disorder persists based on subsequent electrical signals. If the disorder persists, steps 1210-1250 can be repeated to find a second source or target area that may be contributing to the persistent disorder. In some cases, if the ablation was unsuccessful in treating the target, persistence of the disorder can be determined based on subsequently captured electrical signals. The system can be used to apply additional ablation energy to the same area to ensure successful treatment.

[0182] Figure 13 shows one embodiment of an ablation catheter 1300 for treating electrical rhythm disorders. The ablation catheter 1300 combines sensing and therapy delivery functions within a single tool including a spade 1310, a shaft 1320, and a controller 1360. The spade 1310 includes a thin, flexible body 1302 supporting an array of sensing electrodes 1340 to guide the ablation catheter 1300 to one or more sources or other target areas. In most embodiments, the body 1302 also supports one or more ablation components 1350 configured for the delivery of tissue alteration energy (e.g., electromagnetic or thermal), as further described with reference to Figure 13. The body 1302 is generally planar in its relaxed (undeployed) state and is formed from a sufficiently flexible elastic material so that the body collapses and folds within a retaining volume, as well as so that, once deployed, the contact surface of the catheter conforms to an adjacent tissue surface. The spade 1310 may further include components such as one or more irrigation ports for guiding the irrigation fluid to the surrounding tissue or to imaging. The near end of the spade 1310 is coupled to the shaft 1320, which is steerable by the controller 1360. The shaft 1320 houses wiring to various electrodes on the spade 1310, along with channels for supplying fluids such as coolant, irrigation fluid, etc., to the spade 1310. The shaft 1320 extends concentrically through a sheath 1330 having an internal volume configured to hold the spade 1310 when folded. The shaft 1320 may also include one or more contact sensors 1325 for detecting whether the spade 1310 is in contact with tissue. The controller 1360 is coupled to the near end of the shaft 1320. The spade 1310 may be mounted on the shaft 1320 at any point around its circumference. In other words, this does not need to be aligned symmetrically with the center line of the spade as shown in the diagram; it may be offset from the center.

[0183] The controller 1360 is configured to receive and analyze electrical signals detected by the sensing electrode 1340 to determine the location of the source or other target region for arrhythmia and / or the guidance direction. The controller 1360 provides control signals to the shaft 1320 to guide the movement of the spade 1310 in the guidance direction toward the target region. The controller 1360 provides signals to the ablation component 1350 to modify the tissue in the target region. In some embodiments, the functions of the various components of the ablation catheter 1300 may otherwise be distributed among the components. In further embodiments, the ablation catheter 1300 includes more or fewer components than those enumerated herein.

[0184] Spade 1310 is positioned to contact tissue to treat electrical rhythm disorders. Typical spade dimensions are a width (W) of 5mm to 50mm, a length (L) of 5mm to 50mm, and a thickness (H) of 1mm to 5mm. In one common embodiment for cardiac applications, the dimensions are a width of 25mm, a length of 30mm, and a thickness of 2mm. In some embodiments, the spade body 1302 is formed from a thin, flexible polymer such as silicone, polydimethylsiloxane (PDMS), or a similar biocompatible polymer to allow the spade to be easily crushed or folded within the internal volume of the sheath 1330. An exemplary manufacturing process would involve introducing a liquid polymer into a mold with the wire 1305, electrodes 1340, 1350, and other components pre-positioned within the mold. Next, the mold and liquid are exposed to appropriate conditions (e.g., heat, atmosphere, light, etc.) to cure the polymer material to the desired finish and flexibility. Wires may be pre-formed for layout within the mold, while in some embodiments, the wire 1305 may be an interconnect formed using a conductive paste or thin film printed or deposited within channels in the polymer material using known patterning techniques such as thick film, thin film, inkjet, or other printing methods, to electrically connect to the electrode. With further details of the electrode and ablation component positioned within the mold provided, then a first layer of polymer would be added to the mold, leaving the connectors on the back of the components exposed and partially or fully cured. Next, the interconnect would be patterned on top of the first polymer layer. Then, a standard joining method can be used to connect the printed wire to a wire extending the length of the shaft 1320 to connect to the controller 1360. Then, a second polymer layer would be formed to complete and seal the spade structure. The material must be strong enough to withstand multiple transitions between folding and deployment.The polymer material is preferably required to be suitable for heating by high-frequency energy supply. In other embodiments, the spade 1310 can be manufactured from a combination or composite of different materials, or from the same material that has been treated in different ways to impart different properties, such as a varying degree of flexibility. In some embodiments, the thickness of the spade 1310 can vary along its length or in different areas.

[0185] The sensing electrodes 1340 are configured as an array within the body 1302 so as to be coplanar with or slightly protruding from the contact surface 1315. The electrodes in the array can be configured in any number of patterns. For example, the sensing electrodes 1340 may be evenly arranged as a simple rectangular grid as shown in Figure 13, or other configurations (e.g., dispersion and / or density) may be used. The number of sensing electrodes 1340 can range from 4 to 256 electrodes.

[0186] The size and spacing of the sensing electrodes determine the resolution of the catheter. Electrode sizes may range from 0.1 mm to 4.0 mm, and in this case, the choice of electrode size depends on the application. For example, small sensing electrodes spaced closely together will achieve high resolution, but there is a trade-off between small and large electrodes. The smallest practical electrode size can integrate small areas but may be prone to artifacts (e.g., motion artifacts), while the largest electrode size can integrate over a relatively wide area but may lose signals of small amplitude. For complex rhythms such as atrial fibrillation, to provide good signal fidelity and to detect complex signal types that may be therapeutic targets, typical sensing electrodes can range in size from 0.5 to 1.0 mm. For ventricular tachycardia, typical sensing electrodes can range in size from 1 to 2 mm. For simple rhythms such as tachycardia mediated by accessory pathways, the typical electrode size range may be 0.5 to 1 mm to identify accessory pathway potentials. The selection of an appropriate sensing electrode size will be part of the level of skill in this art. In some embodiments, the spade may include sensing electrodes sized in several different ways, either as a group or at specific locations on the spade. The sensing electrodes 1340 may also have various different shapes, such as elliptical, circular, square, rectangular, etc.

[0187] The spacing between the sensing electrodes 1340 (measured edge to edge) can vary in the range of 0.5 to 5.0 mm. In the case of atrial fibrillation, the usual spacing of the sensing electrodes is 1 to 2 mm. In the case of ventricular tachycardia, the usual electrode spacing is 2 to 4 mm. When very fine details must be elucidated, the usual sensing electrode spacing is 0.5 to 0.75 mm. In some embodiments, different spacings can be used for different groups of electrodes or at different locations on the space body.

[0188] In the embodiment shown in Figure 13, the ablation component 1350 is an array of ablation electrodes arranged within the spade body 1302 so as to be coplanar with or slightly protruding from the contact surface 1315. The number of ablation electrodes 1350 may be in the range of 4 to 36, but will depend on the size of the spade. The size of each ablation electrode ranges from 0.5 to 4.0 mm. For complex rhythms such as atrial fibrillation, the typical ablation electrode size ranges from 0.5 to 2.0 mm. For ventricular tachycardia, the typical ablation electrode size ranges from 2.0 to 3.0 mm. For simple rhythms such as tachycardia mediated by an accessory pathway, the typical ablation electrode size range is 0.5 to 1.0 mm. In some embodiments, the spade may include several different types of sized ablation electrodes, arranged as groups or at specific locations on the spade. The ablation electrodes 1350 may also have different shapes, such as elliptical, round, square, rectangular, etc.

[0189] The spacing between ablation electrodes 1350 (measured edge to edge) can vary in the range of 0.5 to 10.0 mm. For atrial fibrillation, the typical ablation electrode spacing is 2.0 to 3.0 mm. For ventricular tachycardia, the typical ablation electrode spacing is 3.0 to 6.0 mm. Typically, the density of ablation electrodes is less than the density of sensing electrodes. However, in some applications, the density may be relatively higher with respect to the ablation electrodes, depending on the target rhythm and chamber.

[0190] When the ablation catheter 1300 is ready for positioning within the individual, the spade 1310 assumes its folded configuration, held within the internal volume of the sheath 1330, to facilitate insertion into the individual's body, for example, via vascular access. Once the ablation catheter 1300 has been guided to the appropriate location, the spade 1310 will be released from the sheath 1330 to initiate the detection and / or treatment process as described above. The flexibility of the spade body 1302 allows the spade 1310 to substantially conform to the tissue surface topography. Optionally, during manufacturing, a spring wire of a further finer gauge may be placed near the spade to facilitate the unfolding of the spade once released from the sheath. These spring wires do not have enough rigidity to harden the spade, but only improve its elasticity to the unfolded, i.e., relaxed configuration. In some implementations, the spade 1310 may further include such materials, for example, by embedding the thermoelectric resistive material within a polymer material used to form the body or to form a composite between the body material. The embedding of the electroresistive material within the spade body may be beneficial for maintaining high fidelity in the electrical signal detected by the sensing electrode 1340. The embedding of one or more thermally conductive materials may be useful when modifying tissue, i.e., electromagnetic ablation, cryoablation, etc., to help diffuse or disperse energy / liquid to avoid direct concentration supply at the site of the ablation component. Note that the ablation catheter 1300 is described herein as a “spade,” but may also be referred to as a paddle, grid, array, matrix, or mesh.

[0191] As described above, the spade 1310 is formed from a thin, flexible, conformable material so that the contact surface 1315 contacts and conforms to the tissue. The shape of the spade 1310 is depicted as a rectangle in Figure 13, but this is for illustrative purposes only, and other shapes can be used. Potential modifications may include cutting or rounding the corners of the spade to facilitate drawing (or pushing) the flexible material into the sheath volume. Several examples of other shapes are provided in Figure 14A, which will be discussed in more detail later. The selection of other shapes that may facilitate the folding / unfolding or other functions of the catheter will be readily apparent to those skilled in the art. The size of the spade 1310 also affects its function. For example, a relatively large spade 1310 will represent a relatively bulky volume through the advantage of covering a relatively large surface area, and therefore deployment and / or extraction may be relatively difficult. The relatively bulky volume corresponds to an increased cross-sectional surface area of ​​the ablation catheter at its maximum width, thereby increasing the opening and conduction path dimensions required for insertion into the individual's body. Spade size can be individualized by PDP based on the predicted size of the target area for an individual with a known clinical profile. An exemplary dimensional range for cardiac arrhythmia applications is in the 1.5cm × 1.5cm to 3cm × 3cm (W × L) range. Spade thickness must be sufficiently flexible to collapse and fold within the sheath while still supporting the sensor array to the tissue's contour. An exemplary thickness range is in the 0.10mm to 4.0mm range, but may vary depending on the components and features embedded within the device. In one embodiment, a range of 0.75mm to 1.0mm would be sufficiently flexible to conform to the cardiac chamber while providing adequate support for the electrode material. In another embodiment, a range of 2 to 3mm would provide relatively greater structural stability for use outside the heart, such as in cardiac surgery applications or in ventricles with a relatively large range of contractile motion.

[0192] The shaft 1320 is an elongated hollow cylinder or tube formed from a semi-rigid material, in which case the distal end is coupled to the spade 1310 and the proximal end is coupled to the controller 1360. The hollow center of the shaft 1320 surrounds one or more bundles 1322 of wires providing electrical connections to the catheter electrodes, steering wires for manipulating the catheter, and, where applicable, fluid tubing, all of which must extend along the entire length of the shaft to provide communication between the spade 1310 and the controller 1360. The shaft 1320 is formed and coated to provide a low-friction biocompatible outer surface to facilitate catheter insertion and removal and to reduce the risk of damage to the insertion pathway. The inner surface of the hollow catheter may also have a low-friction coating to allow free movement of the bundles 1322 of wires and tubing within the center of the shaft. Suitable shafts and shaft materials are known in the art and are widely available commercially. The shaft 1320 is steerable by the controller 1360 to operate the spade 1310. The shaft 1320 must be long enough to extend from the entry point within the individual to the tissue to be evaluated / treated. For example, in the treatment of cardiac arrhythmias, the ablation catheter 1300 will be inserted through an opening in the individual's leg or groin and guided through the femoral artery to the heart. Therefore, the shaft 1320 must be long enough to extend over the entire length from the entry point to the individual's heart. The partial rigidity of the shaft protects the components housed within the center of the shaft and helps prevent deflection within the shaft that could affect the movement of the spade 1310.

[0193] To allow detection of proper contact between the spade 1310 and the tissue surface, the contact sensor 1325 can be positioned on the far end of the shaft. In other embodiments, the contact sensor 1325 can be positioned elsewhere on the ablation catheter 1300, for example, at some point on the contact surface 1315. Several different sensor types can be used for the contact sensor 1325. For example, the contact sensor 1325 may be a sensor configured to measure the force applied to a force sensor. Sufficient contact with the tissue surface can be found when the force sensor detects a force exceeding a threshold, for example, 0.25 Pascals. Another type of sensor that can be used is a proximity sensor that has the ability to measure the distance to another surface via capacitive sensing. A change in capacitance is used to calculate the distance between the tissue surface and a capacitor in the proximity sensor. Sufficient contact with the tissue surface can be found when the measured distance is within a threshold distance, for example, 0.1 millimeters.

[0194] The sheath 1330 is a rigid, hollow, substantially cylindrical body configured to hold the spade 1310 in a collapsed or folded configuration. For simplicity, the sheath 1330 is shown as a standard cylinder with its base positioned perpendicular to the sides; however, in practical applications, the sheath body can be chamfered, rounded, or tapered so that its outer surface is smooth and edge-free, thereby minimizing components with corners that could be trapped on features along the insertion / extraction pathway. For example, an oval or partially oval shape could be used. The sheath 1330 is held concentrically with respect to the shaft 1320 and is configured to slide longitudinally along the shaft 1320 to pull it back from the far end of the shaft to release the spade 1310. Specifically, in its initial position, the sheath 1330 holds and covers the spade 1310 in a folded configuration to allow the catheter to be guided into the tissue. Once the catheter is positioned in the tissue, the sheath 1330 is moved away from the distal end of the shaft to a second position, thereby releasing the spade 1310 and allowing it to return to its deployed configuration. The sheath 1330 can be connected to the controller 1360 via one or more wires that guide the sheath between the first and second positions. In one embodiment, these wires can be activated by a micromotor 1326 in the controller. In some embodiments, the sheath 1330 can be displaced using a sheath that is electronically connected to the controller 1360 by a micromotor built into the sheath and / or wires extending through the shaft to the controller.

[0195] Different organs may have different size recommendations, which may be relatively small, for example, in the brain or in the case of neural mapping, and generally will vary depending on a given application and / or biological chamber. For example, in cardiac ablation, it may be important that the detection surrounds an area such as a scar, while the detection may be a relatively uniform grid. As another example, in brain ablation, it may be important that the detection is relatively dense and focused to allow for deep, but narrow, penetration, while the detection may cover a relatively wide area. Another constraint on the number of detection electrodes in the catheter is the size of the bundle of wires 1322 that can fit within the shaft. In the example shown in Figure 13, 24 detection electrodes 1340 are configured as a rectangular grid.

[0196] The sensing electrode 1340 has the ability to detect electrical signals from tissue when in contact with it. Various types of electrodes can be implemented for the sensing electrode 1340. The electrodes can be constructed from semiconductor or conductive materials that have the ability to detect electrical signals from the tissue surface. The sensing electrode 1340 may be individually addressable by the controller and / or configured in a custom pattern adapted to the expected characteristics of the target arrhythmia region. Multiple electrodes can be formed in one or more continuous sheets of conductive material, such as one or more sensor chips, each having multiple sensors, or separate sensors can be individually arranged within the spade body 1302. Other types of sensors may be included in the catheter to measure, for example, heat (infrared), mechanical motion (piezoelectric or other sensors), chemical composition, or other indices that may have diagnostic value. The circuit supporting the sensing electrode 1340 is preferably robust and shock-resistant to withstand the energy and pressure that may occur during ablation. The electrical signals may be in the form of an electrocardiogram measuring the potential of the tissue. The electrical signals may relate, at least in part, to the electrical rhythms of the tissue. The arrangement of the sensing electrodes 1340 is used to generate a mapping of electrical signals useful for analyzing the location of the source of an electrical rhythm disturbance or other target area or guidance direction thereto, according to the principles described, for example, with reference to Figures 11A and 11B.

[0197] Referring further to Figure 13, one or more ablation components 1350 supply ablation energy to or assist in the supply of ablation energy to the tissue. In one or more embodiments, the ablation components 1350 are arranged within the contact surface 1315 of the spade 1310. As shown, one possible configuration involves the arrangement of five ablation components, with one component centrally located and two components located distally and proximal to the contact surface, respectively. This pattern is merely an example, and other variations with different spacing or patterns and different numbers of elements can be used, for example, to adapt an ablation treatment corresponding to the determined size and / or shape of the source or other target area. The activation of one or more of the ablation components 1350 is controlled by the controller 1360, thereby allowing each ablation component to be selectively controlled to provide an amount of ablation energy selected from a range of ablation energy.

[0198] In some embodiments, the ablation component 1350 is an ablation electrode that contacts the tissue to supply electromagnetic energy as ablation energy. The ablation electrode typically has a larger contact area than that of the sensing electrode 1340 to supply sufficient electromagnetic energy to the surface. The electromagnetic energy may include high-frequency electromagnetic waves or other frequencies of electromagnetic waves. Further features and variations of the catheter of the present invention will be described later with reference to Figures 14A to 19.

[0199] The controller 1360 analyzes the electrical signals received from the sensing electrodes 1340 to determine the location of the source or other target area or the direction of guidance to it. Knowledge of the physical position of each sensing electrode 1340 on the spade 1310 allows the controller to determine the location of the electrical signals corresponding to each sensing electrode 1340 in relation to others. The controller 1360 uses the steps shown in Figures 11A, 11B, and 12 to determine the location of the source or other target area and / or one of the directions of guidance to it, and then generates a signal to cause the shaft 1320 to move the spade 1310 toward the target. Upon confirmation of arrival at the target (e.g., steps 1185, 1230), the controller 1360 commands the supply of ablation energy to the tissue (steps 1190, 1240). The controller 1360 may further interact with other components of the device, namely the contact sensor 1325, to verify proper tissue contact, and if contact is deemed insufficient, it can command the movement of the shaft 1320 to properly position the spade 1310 for sufficient contact with the tissue. By executing the sequences shown in Figures 11A, 11B, and 12, the controller 1360 can continue to detect and treat any other detected target areas, if present, until the consideration for treatment of all is complete.

[0200] In one or more embodiments, the ablation catheter 1300 can operate semi-autonomously. In these embodiments, the controller 1360 performs actions such as locating the source or other target area, moving the spade 1310 to the target, treating the target area by modifying the tissue with the ablation component 1350, and terminating the procedure upon confirmation of successful treatment. In these embodiments, minimal intervention by a physician would be required to operate the ablation catheter 1300.

[0201] In other embodiments, the ablation catheter 1300 is operated by a physician. The controller 1360 detects and determines the location and / or guidance direction to the source or other target area, and then generates one or more indicators (e.g., visual and / or audio) of the location and / or guidance direction for the physician. The controller 1360 connects the catheter to an energy source, an input system, and a visual display system. The controller may also be connected to a motor 1326 that can directly move the catheter or assist in its movement. In one embodiment, the physician can physically manipulate the shaft 1320 to move the spade 1310 in the notified direction. In other embodiments, the physician can control the movement of the shaft 1320 via a user interface 1365 that is in communication with the controller (and the motor 1326), thereby allowing the physician to control the movement of the spade 1310. Examples of user interfaces that may be included in embodiments include a handle, joystick, mouse, or trackball on a computer. In one alternative embodiment, an operator, who may be a treating physician, may use a virtual, augmented, or mixed reality headset or virtual, augmented, or mixed reality goggles to guide the movement of a detection or treatment tool.

[0202] Figure 14A shows three examples of alternative spade configurations. Spade 1410 is a rectangle with rounded corners, which allows for the implementation of a rectangular grid of sensing electrodes. This configuration can provide stability within large planar structures such as the posterior wall of the left atrium. Spade 1420 is elliptical in shape, which can facilitate the positioning of the device near extreme bends, such as near pulmonary veins, but contains a relatively small number of electrodes in the vicinity of the periphery. Spade 1430 is a ring-shaped sensing electrode with sensing electrodes configured as a ring. This configuration may be most effective for “isolating” the area of ​​interest without necessarily completely abrading, for example, to minimize energy delivery and to avoid damage to highly sensitive structures at the center of the area. These illustrated spade configurations are provided as examples only. As will be apparent to those skilled in the art, other shapes and aspect ratios can be used to adapt the catheter to the specific needs of an individual.

[0203] Figure 14B provides an example of adapting a spade configuration to a source or other target region in an electrical rhythm disorder. The illustrated example is for atrial fibrillation, which can be supported not by a point, but rather by a target region sized to cover a “large domain” 1440. Alternatively, area 1450 represents a target region that may be a localized region with centrifugal propagation of activation, but may also be a rotational, partial rotational, repetitive site, and other patterns of the type shown in Figure 9. Area 1460 (within the dashed line) depicts a schematic “domain size,” or the region of tissue that must be targeted to treat the biological rhythm disorder. As illustrated, this is the domain size of the desired target region, which may be an anatomical region of target that may exhibit low voltage, or a localized source for atrial fibrillation (localized, rotational, rotor, partial reentry, repetitive site in Figure 9). Therapy for biological rhythms is made possible by matching spades of various shapes and configurations to target notified domain sizes, such as 1440, 1450, or 1460.

[0204] Figures 15A and 15B show one embodiment of the ablation catheter of the present invention configured to apply electromagnetic energy to modify tissue in a target region. The ablation catheter 1500 includes a plurality of ablation electrodes 1520 having a larger area than a small high-resolution sensing electrode 1340 to allow for the supply of electromagnetic energy for electroporation. The spacing between the ablation electrodes 1520 may be small enough to ensure substantially continuous tissue lesions. The ablation electrodes 1520 may be scattered among the sensing electrodes 1340 as shown, or other patterns may be used. The ablation electrodes 1520 may be activated together, or they may be activated within one or more sub-regions by, for example, dividing the electrodes into quadrants. The ablation electrodes 1520 may be further configured to supply energy in various energy signal shapes (waveforms).

[0205] As is known in the art, a waveform describes an electrical energy signal having a combination of variable parameters such as voltage, current, frequency, waveform shape, duration, phase, or other waveform properties generated by an ablation electrode. For example, one waveform may be a sine wave with a defined amplitude and frequency applied over a defined duration, such as a few milliseconds. An example of ablation with a changing waveform may require a controller that causes a first ablation electrode to emit a sine wave having a first defined amplitude, shape, frequency, and duration, and a second ablation electrode to emit a sawtooth wave having a second defined amplitude, shape, frequency, and duration. Various combinations and sequences of waveforms can be utilized.

[0206] The cross-sectional view of the spade 1510 in the width direction shown in Figure 15B shows a single layer of spade body 1302 having a sensing electrode 1340 and an ablation electrode 1520 positioned such that its outer surface is coplanar with the contact surface 1515. Each electrode (sensing electrode 1340 and ablation electrode 1520) is connected via corresponding wires 1530 that extend outward from the spade body and continue through the shaft 1320 to the controller 1360.

[0207] Figures 16A and 16B show one embodiment of an ablation catheter configured to deliver an irrigation fluid, such as saline or a chemical buffer, to tissue through one or more irrigation holes 1620. The irrigation fluid cools the tissue surface to avoid overheating of the ablation electrode tip and possible power shutdown, thereby allowing for a relatively deep energy delivery. The irrigation holes 1620 are shown evenly distributed throughout the spade 1610, but different configurations may be used. For example, the irrigation holes 1620 may be concentrated near the ablation electrode 1520. One or more irrigation fluid channels extend through the shaft 1320 (as part of a bundle 1322) to connect the irrigation holes to an irrigation fluid reservoir 1650 associated with and / or controlled by a controller 1360 that controls the supply of irrigation fluid to the holes.

[0208] The widthwise (transverse) cross-sectional view of the spade 1610 in Figure 16B shows a single flexible layer having a sensing electrode 1340, an ablation electrode 1520, and an irrigation hole 1620 configured on the contact surface 1615. In some implementations, each irrigation hole can be further controlled by a hole gate 1625 to mechanically gate the flow from the irrigation fluid channel to the irrigation hole 1620. One or more hole gates 1625 may be further controlled by a controller 1360. In some embodiments, multiple irrigation holes 1620 can be controlled by a single hole gate 1625. The hole gate 1625 can be further controlled to release the irrigation fluid at different flow rates, for example, by periodically releasing 5 mL of irrigation fluid every few minutes.

[0209] Figures 17A–17C show a variation of one embodiment of the ablation catheter of the present invention configured to supply freezing energy to alter tissue in a source or other target region. The ablation catheter 1700 includes a sealed coolant layer 1720 contained within the body of a spade 1705. The coolant layer 1720 may be a single coolant chamber 1740 (Figure 17B), one or more coolant splines 1780 (Figure 17C), other configurations of the coolant chamber, etc. The coolant layer 1720 is configured to hold coolant to rapidly cool part or all of the spade 1705. The cooled spade 1705 is useful for providing freezing energy to the tissue surface. The coolant chamber 1720 is coupled to a controller 1360 via one or more coolant channels extending through a shaft 1320.

[0210] The cross-sectional view of the spade 1705 shown in Figure 17B illustrates the two-layer structure of this embodiment, in which the first layer 1710 holds the sensing electrode 1340 positioned within the contact surface 1715. The second layer 1720 is defined by the coolant chamber 1740 sealed within the chamber wall 1745. The material forming the chamber wall 1745 must not only be durable enough to retain its seal after multiple exposures to the coolant, but also thin and flexible enough to expand with the coolant. When the chamber 1740 deflates, the combined first and second layers must be flexible enough to allow the spade 1705 to collapse / fold into the sheath 1330. The coolant chamber 1740 is connected to a coolant channel extending through the shaft 1320. One or more coolant channels extend through the shaft 1320 (as part of the bundle 1322) to connect the coolant chamber 1740 to a coolant reservoir 1750 associated with and / or controlled by a controller 1360 that controls the supply of coolant to the chamber (via a pump not shown). To improve the heat transfer of freezing energy from the coolant chamber 1740 to the contact surface 1715, a thermally conductive material may be embedded within the first layer 1710 or otherwise incorporated therein.

[0211] Figure 17C shows an alternative cross-sectional view of the spade 1705, in which the second layer 1720 includes one or more coolant splines 1780 instead of the single coolant chamber 1740 in Figure 17B. The sensing electrode 1340 is located within the contact surface 1715 of the first layer 1710. The coolant splines 1780 are configured to be filled with coolant to rapidly cool the portion of the contact surface 1715 corresponding to the splines. In some embodiments, the coolant splines 1780 may have similar size and shape. As shown in Figure 17C, the coolant splines 1780 extend longitudinally within the spade 1705, but different configurations, such as size and shape, can be used. The controller 1360 can separately supply freezing energy to selected splines to target areas of tissue to be treated. For example, two intermediate coolant splines can be filled with coolant to cool only a third intermediate portion of the spade. As described above, the first layer 1710 can incorporate a thermally conductive material.

[0212] Figures 18A and 18B show another embodiment of the ablation catheter of the present invention for performing cryoablation. The ablation catheter 1800 has a spade 1810 having a sensing electrode 1340 positioned on a contact surface 1815 to supply freezing energy to the tissue surface and one or more cryoablation locus 1820.

[0213] The cross-sectional view in the width direction shown in Figure 18B illustrates the two-layer structure of Spade 1810. The first layer 1830 supports the sensing electrode 1340 and the cryoablation locus 1820 within the contact surface 1815. The second layer 1840 encloses the coolant chamber 1850, which in structure is similar to the coolant chamber 1740 of Spade 1705. The cryoablation locus 1820 is a channel that partially extends through the first layer 1830 and connects to the coolant chamber 1850. These channels are sealed at the contact surface 1815 to prevent coolant from being released from the device. The cryoablation locus 1820 is configured to be filled with coolant from the coolant chamber 1850. When filled with coolant, the area on the contact surface 1815 corresponding to the cryoablation locus 1820 is rapidly cooled to provide freezing energy to the adjacent tissue. In some embodiments, the cryoablation locus 1820 has a locus gate 1825 to selectively allow coolant to flow from the coolant chamber 1850 to the cryoablation locus 1820. The locus gate 1825 can be controlled by a controller 1360. In other embodiments, the cryoablation locus 1820 can be directly connected to a coolant channel without the coolant chamber 1850. The trade-off is that having the coolant chamber 1850 increases the maximum amount of freezing energy, i.e., affects the rate of cooling, but sacrifices the thickness of the spade 1810. One or more coolant channels extend through the shaft 1320 (as part of a bundle 1322) to connect the coolant locus 1820 to a coolant reservoir 1850 associated with and / or controlled by a controller 1360, which controls the supply of coolant to the locus (via a pump not shown).

[0214] Figure 19 shows yet another embodiment of the ablation catheter of the present invention, in which targeting fiducials are included to facilitate treatment via an external energy source. The ablation catheter 1900 includes a plurality of targeting fiducials 1920 positioned within a spade 1910 to guide the delivery of ablation energy from one or more external ablation components. The targeting fiducials 1920 can be visualized using fluoroscopy as is known in the art, or can be detected by other techniques. In this embodiment, once a therapeutic target for cardiac rhythm is detected, energy can be supplied from an external source of X-rays, or other electromagnetic radiation, or a proton beam. Such energy sources may be similar to those used for radiotherapy for tumors. The spacing between the targeting fiducials 1920 is small enough to ensure continuous tissue lesions. The fiducials can be targeted together, or they can be targeted within subregions corresponding to quadrants of the sensor.

[0215] Figure 20 schematically illustrates a computer system that may be used to implement the method of the present invention, which may be embedded in a variety of devices, such as a personal computer (PC), tablet PC, personal digital assistant (PDA), mobile device, palmtop computer, laptop computer, desktop computer, communication device, control system, web appliance, or any other machine capable of executing (sequentially or otherwise) a set of instructions that define the actions to be performed by that machine. Furthermore, a single computer system 2300 is shown, but the term “system” should also be interpreted to include any collection of systems or subsystems that can individually or collaboratively execute one or more sets of instructions to perform one or more arithmetic functions.

[0216] As shown in Figure 20, the computer system 2300 may include a computer processor 2302, such as a central processing unit (CPU), a graphics-processing unit (GPU), or both. The computer system may include main memory 2304 and static memory 2306, which can communicate with each other via a bus 2326. As shown, the computer system 2300 may further include a video display unit 2310, such as a liquid crystal display (LCD), organic light-emitting diode (OLED), flat panel display, semiconductor display, or cathode ray tube (CRT). In addition, the computer system 2300 may include an input device 2312, such as a keyboard, and a cursor control device 2314, such as a mouse. Furthermore, the computer system 2300 may include a drive unit 2316, a signal generating device 2322 such as a speaker or remote control device, and a network interface device 2308.

[0217] In some embodiments, the drive unit 2316 may include a computer-readable medium 2318 on which one or more sets of instructions 2320, such as software, are stored. The drive unit 2316 may be a disk drive, a thumb drive (USB flash drive), or other storage device. Furthermore, the instructions 2320 may also implement one or more of the methods or logic described herein. In a particular embodiment, the instructions 2320 may reside entirely or at least partially in the main memory 2304, static memory 2306, and / or processor 2302 when executed by the computer system 2300. The main memory 2304 and processor 2302 may also include a computer-readable medium.

[0218] In an alternative embodiment, a dedicated hardware implementation can be constructed, such as an application-specific integrated circuit (ASIC), a programmable logic array (PLA), and other hardware devices, to implement one or more of the methods described herein. Applications, which may include devices or systems of various embodiments, can broadly encompass a wide range of electronic and computer systems. One or more embodiments described herein can implement functionality by using two or more specific interconnected hardware modules or devices having relevant control and data signals that can be transmitted between and through modules, or as part of an application-specific integrated circuit. Thus, the system encompasses software, firmware, and hardware implementations.

[0219] Depending on the various embodiments, the methods described herein can be implemented by software programs tangibly implemented in a processor-readable medium and can be executed by a processor. Furthermore, in exemplary, non-limiting embodiments, implementations may include distributed processing, component / object distributed processing, and parallel processing. Alternatively, virtual computer system processing can be constructed to implement one or more of the methods or functions described herein.

[0220] Furthermore, the computer-readable medium is assumed to include, receive, and execute commands 2320 so that a device connected to network 2324 can transmit voice, video, or data over network 2324. Moreover, commands 2320 can also be transmitted or received over network 2324 via network interface device 2308.

[0221] The above describes embodiments of systems and methods for generating personalized digital phenotypes of diseases, which are compared to digital taxonomies for the purpose of personalizing therapy. While specific exemplary embodiments are described, it will be apparent that various modifications and variations can be made to these embodiments without departing from the relatively broad scope of the invention. Accordingly, the detailed description should not be interpreted as limiting, and the scope of the various embodiments is defined solely by the appended claims, which include the entire range of equivalents of those granted thereto.

Claims

1. A system for treating cardiac rhythm disorders, wherein the system is A catheter configured to be positioned in contact with a tissue surface, wherein the catheter is A flexible body having a contact surface, An array of sensor electrodes arranged within the flexible body, wherein each sensor electrode has a conductive surface located substantially coplanar with the contact surface, and each sensor electrode is configured to detect an electrical signal at one location within the contacted area of ​​the tissue surface, and One or more therapeutic elements disposed within the flexible body configured to supply energy to the tissue surface, A catheter having, A plurality of conductors, each conductor having a far end connected to a sensor electrode and one of the one or more therapeutic elements, A controller in communication with the nearest end of the aforementioned plurality of conductors, It has, The controller has a processor, The aforementioned processor, Upon receiving the detected electrical signal, Based on the spatial gradient of the detected electrical signal, the direction of the electrical flow toward the location of the target region related to cardiac rhythm disorder in the patient is determined. From the detected electrical signal, it is determined whether the one or more therapeutic elements overlay the treatment target, and if they do not overlay the treatment target, the direction toward the treatment target is calculated, and a command is generated to guide the catheter toward the treatment target, and After determining that one or more of the therapeutic elements are at least partially overlaid on the target of treatment, a therapeutic signal is generated to activate the one or more therapeutic elements to modify the tissue related to the target of treatment within the target region. A system configured in such a way.

2. The system according to claim 1, wherein the flexible layer is substantially planar and has a shape selected from the group consisting of rectangles, ellipses, and rings.

3. An elongated, hollow shaft having a distal end, a proximal end, and a length, wherein the catheter is disposed at the distal end, the controller is disposed at the proximal end, and the plurality of conductors are held within the shaft and extend to the length thereof, and the distal end of the shaft is operable from the proximal end; A shaft motor configured to steer the far end of the shaft in response to motion commands generated by the controller; and The system according to claim 1 or 2, further comprising one or more sheaths slidably disposed on the shaft, the sheaths having an internal volume configured to hold the catheter in a folded state until the catheter is deployed by sliding the sheath away from the distal end of the shaft.

4. The system according to any one of claims 1 to 3, further comprising an irrigation fluid pore formed within the flexible body, wherein the irrigation fluid pore is in fluid communication with the controller and associated irrigation fluid source, and the irrigation fluid source is configured to supply irrigation fluid to tissue through the irrigation fluid pore in the target region in connection with the activation of the array of therapeutic elements.

5. The system according to any one of claims 1 to 4, wherein the one or more therapeutic elements have an array of ablation electrodes, and a subset of the plurality of conductors connected to the one or more therapeutic elements are conductors configured to supply electromagnetic energy to each ablation electrode.

6. One or more of the sensor electrode arrays and the ablation electrode arrays are uniformly dispersed around the contact surface; and The system according to claim 5, wherein the ablation electrodes of the array of ablation electrodes are evenly distributed within the array of sensor electrodes.

7. The aforementioned processor, Based on the detected electrical signal, the size of the treatment target is determined. Identifying one or more ablation electrodes in the array of ablation electrodes based at least on the size and location of the target to be treated, To activate the identified ablation electrode, The system according to claim 5, further configured as follows.

8. The aforementioned processor, A directional map of the heart rhythm is generated based on the average direction of the electrical flow of the detected electrical signals, and this directional map describes the conduction path of the heart rhythm. To generate guidance directions for moving the flexible body toward the treatment target, To determine the location of the target to be treated, the directional map is integrated. The system according to any one of claims 1 to 7, wherein the direction of the target to be treated is calculated accordingly.

9. The system according to claim 8, wherein the directional map is generated by applying a trained machine learning model to the electrical signal, the machine learning model being trained on training examples having electrical signals of a human heart and known target regions of cardiac rhythm disorders.

10. The system according to any one of claims 5 to 7, wherein each ablation electrode is configured to emit a separate waveform.

11. The system according to claim 10, wherein the controller is configured to address one or more subsets of the ablation electrodes of the array separately, and the treatment signal comprises a first signal to a first subset of ablation electrodes for emitting a first waveform, and a second signal to a second subset of ablation electrodes for emitting a second waveform.

12. The system according to any one of claims 1 to 11, wherein the sensor electrodes are configured to supply ablation energy such that the one or more therapeutic elements have an array of the sensor electrodes.

13. The one or more therapeutic elements are one or more coolant chambers formed within the flexible body and configured to hold coolant, and the plurality of conductors are coolant chambers having a subset of conductors configured to guide coolant fluid from a coolant source to the one or more coolant chambers in order to supply freezing energy to tissue in the therapeutic object; An array of cryoablation locus formed within the flexible body, wherein the plurality of conductors are a subset of conductors configured to guide coolant fluid from a coolant source to the cryoablation locus in response to a therapeutic signal from the controller for supplying freezing energy to the tissue in the target region; and The system according to any one of claims 1 to 12, comprising one or more arrays of targeting fiducials dispersed within the flexible body, wherein the targeting fiducials are configured to guide the supply of ablation energy from one or more external ablation energy sources.

14. The system according to claim 13, wherein the flexible body has a thermally conductive material built inside it to improve the conduction of freezing energy to the tissue in contact with the contact surface.

15. The system according to any one of claims 1 to 14, further comprising a contact sensor configured to determine whether the contact surface and the tissue surface are in sufficient contact, and, if not, to provide a signal to the controller to guide the movement of the flexible body to provide improved contact.

16. The system according to claim 4, wherein each of the irrigation fluid holes is controlled by a single hole gate to release the irrigation fluid at a controlled flow rate.

Images