Drug dosage management method and system

Abstract

A heart failure management system for a patient includes a plurality of sensors including at least one implantable pressure sensor positioned within the heart to capture synchronized left and right heart pressure waveforms, transmitted via one or more transmitters to a receiver. Control circuitry receives these waveforms and processes them using a machine learning medication dosage management model, trained on historical left and right heart pressure logs, volume status metrics (e.g., weight, fluid intake, symptoms), and / or ground truth dosage logs for certain heart failure treatment medications, such as a blood pressure reducing medication (e.g., ARNI) and an intravascular volume reducing medication (e.g., Lasix). The model can generate real-time dosage recommendations, systemic blood pressure, and volume load predictions in real-time, enabling precise, data-driven treatment adjustments aligned with clinical guidelines, supporting timely clinical decisions or automated medication dispensing to optimize patient outcomes and reduce hospitalization risks.

Description

DRUG DOSAGE MANAGEMENT METHOD AND SYSTEMFIELD OF THE DISCLOSURE

[0001] The present disclosed technology generally relates to a method and system of administering drug dosages for the treatment of heart failure.BACKGROUND

[0002] Chronic heart failure is a progressive condition that affects the heart’s ability to pump blood effectively. It often involves multiple underlying issues like high blood pressure, coronary artery disease, and diabetes. Current clinical guidelines prescribe the appropriate use of up to four drugs for treating heart failure patients. To manage these medications, a typical heart failure patient requires frequent clinic visits and regular adjustments to their drug dosages. Even with these frequent adjustments, heart failure patients are still likely to require multiple hospital admissions per year, placing an enormous burden on the healthcare system.SUMMARY OF THE DISCLOSURE

[0003] In one aspect of the disclosure, a drug dosage system includes a right heart pressure sensor which provides right heart pressure data and a left heart pressure sensor which provides left heart pressure data. The drug dosage system also includes a receiver which is electronically connected to the right heart pressure sensor and the left heart pressure sensor and receives the right heart pressure data and the left heart pressure data. The drug dosage system also includes a computing device with a processor and memory. The memory includes programming that is executable by the processor to: compute a dosage of a first drug based on the left heart pressure data and right heart pressure data; and compute a dosage of a second drug based on the left heart pressure data and right heart pressure data.

[0004] In another aspect of the disclosure, a method is disclosed of administering two medications. The method includes providing an initial dosage of a first drug and an initial dose of a second drug. The method further includes measuring a left heart blood pressure of a patient and measuring a right heart blood pressure of the patient. The dosage of the first drug is adjusted based on the left heart blood pressure and the right heart blood pressure. The dosage of the second drug is adjusted based on the left heart blood pressure and the right heart blood pressure. The method further includes providing an adjusted dosage of the first drug and an adjusted dosage of the second drug.

[0005] In yet another aspect of the disclosure, a method is disclosed for administering two medications. The method includes providing a first dosage of a first drug to a patient and providing a second dosage of a second drug to the patient. A left heart blood pressure of the patient is measured by a drug dosage system. A right heart blood pressure of the patient is measured by the drug dosage system. The drug dosage system adjusts the first dosage of the first drug based on input features from the left heart blood pressure and the right heart blood pressure. The drug dosage system adjusts the second dosage of the second drug based on the input features from the left heart blood pressure and the right heart blood pressure. The drug dosage system provides to the patient a first adjusted dosage of the first drug and a second adjusted dosage of the second drug.

[0006] In another aspect of the disclosure, a method is disclosed for administering two medications. The method includes providing a first dosage of a first drug to a patient and providing a second dosage of a second drug to the patient. A left heart blood pressure waveform of the patient is measured by a drug dosage system. A right heart blood pressure waveform of the patient is also measured by the drug dosage system. A processor of the drug dosage system performs waveform analysis of both the left heart blood pressure waveform of the patient and the right heart blood pressure waveform of the patient to determine input features of the patient. The input features of the patient are inputted into an artificial neural network stored in memory of the drug dosage system to determine a systemic blood pressure of the patient and a systemic blood volume load of the patient. The drug dosage system adjusts the first dosage of the first drug based on the systemic blood pressure of the patient determined by the artificial neural network. The drug dosage system also adjusts the second dosage of the second drug based on the systemic blood pressure of the patient determined by the artificial neural network. The drug dosage system provides to the patient a first adjusted dosage of the first drug and a second adjusted dosage of the second drug.BRIEF DESCRIPTION OF THE DRAWINGS

[0007] The description will be more fully understood with reference to the following figures and data graphs, which are presented as various embodiments of the disclosure and should not be construed as a complete recitation of the scope of the disclosure, wherein:

[0008] FIG. 1 is a graphic illustration of an operation of a system for dosing a patient in accordance with one or more examples.

[0009] FIG. 2 is a diagram of a drug dosage system, in accordance with one or more examples.

[0010] FIG. 3 is a flowchart of an example method for administering two medications, in accordance with one or more examples.

[0011] FIG. 4 is a diagram of a clinical dataset used for data mining and machine training of the drug dosage system, in accordance with one or more examples.

[0012] FIG. 5 is a flow diagram for extracting a set of input features derived from waveform features from both a right heart pressure waveform and a left heart pressure waveform of a clinical patient for training a machine learning model of the drug dosage system, in accordance with one or more examples.

[0013] FIG. 6 is a graph illustrating an example trace of a blood pressure waveform including example indicia corresponding to signal measures used to extract the input features for training the machine learning model of the drug dosage system, in accordance with one or more examples.

[0014] FIG. 7 is another flow diagram for using a neural network to extract a set of input features derived from the waveform features from both the right heart pressure waveform and the left heart pressure waveform of the clinical patient for training the machine learning model of the drug dosage system, in accordance with one or more examples.

[0015] FIG. 8 illustrates a medication dosage management machine learning architecture in accordance with one or more examples.DET AILED DESCRIPTION

[0016] Proper management of heart failure using drugs is a complex undertaking. There are at many types of drugs used to manage heart failure patients and recent studies suggest that effective management of heart failure may require patients to be taking drugs from at least four families at appropriate doses, principally: diuretics, beta-blockers, Angiotensin-Converting Enzyme (ACE) inhibitors, and Angiotensin II Receptor Blockers (ARBs). Heart failure medications prescribed to heart failure patients may be used to control multiple symptoms simultaneously such as: elevated blood pressure (e.g., arterial systemic hypertension, pulmonary hypertension); increased heart rate or workload; elevated fluid volume in the body, which may lead to pulmonary congestion and / or edema of the extremities; renal function; and / or electrolyte levels in the blood (e.g., sodium, potassium, iron).

[0017] Drug choice and dosages can be varied based on a patient’s underlying heart function and on their body’s responses to the different drugs. Current guidelines prescribe the appropriate use of these drugs, but the guidelines may also be complex, and many physicians and patients find them hard to follow. Guideline Directed Medical Therapy (GDMT) is one cornerstone of an effective treatment protocol. The multiple prescribed heart failure medications may also interact with each other. Patients can also vary their lifestyles, diets, and other behaviors. In addition, patients’ hemodynamic conditions change over time and may benefit from frequent adjustments to their drug dosage regimes. Thus, it can be challenging to tailor treatments to individual patients.

[0018] Heart failure monitoring systems have been developed to help physicians and patients optimize medical therapy. Some monitoring systems monitor blood pressure in the pulmonary artery using an implantable sensor. In large, randomized studies, physicians optimized medication based on data from a pulmonary artery pressure (PAP) sensor. Some treatment protocols associated with such studies involved getting a patient to GDMT, and then titrating diuretics up and down to keep PAP in a target value. These treatment protocols seem to be effective at keeping patients out of hospital for longer and improving the quality of life of patients. Many other monitoring technologies, both invasive and non-invasive, may help patients achieve similar results. However, a reduction in the burden on the healthcare system has not yet been demonstrated for these other approaches.

[0019] Certain monitoring systems gather readings and report their measurements to a physician or directly to patient through a screen, such as on a computer screen or mobile device screen. The application may be a mobile application on a smart device (e.g., a smart phone). If the patient follows a single drug protocol, the patient may have written instructions on how to make any necessary adjustments to their dosage based on the monitor measurements The written instructions may be in the form of a look-up table or a sliding scale.

[0020] When two or more drugs are being titrated every day, the probability of errors increases relative to single-drug protocols. Written instructions are not practical, and the probability of error may be high for both physicians and patients. Monitoring multiple symptoms with multiple sensors and try to optimize multiple medications may not be practical. It has been discovered that automation of two or more drug dosage protocols is advantageous. Further, it may be advantageous to have an application that clearly communicates the proper drug dosages to the patient. Various disclosed implementation do not follow one sensor-onedrug protocols that rely on the minimum number sensors and drugs to manage. Described herein is a dual sensor dual drug titration system for managing heart failure.

[0021] The present technology relates to a system and method for administering two different heart drugs based on patient’s physiology as measured by implantable cardiac blood pressure sensors. Various implementations include monitoring a patient by one or more sensors. For example, two implantable pressure sensors may be placed in a patient. The first sensor can measure pressure corresponding to or closely associated with blood pressure in the right side of the heart. The second sensor can measure pressure corresponding to or closely associated with blood pressure in the left side of the heart. The first sensor can be placed in or in fluid communication with the right atria, the right ventricle, the pulmonary artery (trunk or branches), the coronary sinus, the superior vena cave, the inferior vena cave or the like. The second sensor can be placed in or in fluid communication with the left atria, the left ventricle, any of the pulmonary veins, the aorta (ascending or descending) or the like. The first and second sensors can measure simultaneous blood pressure measurements or asynchronous blood pressure measurements. This monitoring may provide insight into the patient’s hemodynamics. Various implementations include a system that uses dual sensors to facilitate an enhanced treatment protocol.

[0022] Multiple sensors and computer algorithms may be utilized to monitor heart failure patients and dynamically manage titrating more than one medication at a time to optimize medical management. S y stems disclosed herein can advantageously be configured to interpret two important drivers of symptoms: blood volume and blood pressure. In heart failure, blood pressure and body fluid volume are generally not independent of each other but they may not be completely correlated. Changes in blood volume and blood pressure independently cause other symptoms and can each alone, or in combination, lead to decompensation requiring hospitalization.

[0023] While heart failure drugs affect both blood volume and blood pressure, a first group of drugs with a vasodilatation / vasoconstriction mechanism of action may have a greater and / or faster impact on pressure. Such drugs may act mostly downstream from the heart and therefore have a larger impact on the left side of the heart. Angiotensin receptor blocker (ARB), angiotensin-converting enzyme inhibitor (ACEi), and angiotensin receptor neprilysin inhibitor (ARNI) medications have a primary antihypertension mechanism of action. A second group of drugs with a primarily renal mechanism of action have a larger and / or faster impact on blood volume. They may act mostly upstream from the heart and therefore have a larger impact onthe right side of the heart. Diuretics, MRA, SGLT2i are examples of HF medications with a renal mechanism of action. A third group of drugs act on the heart muscle itself modulating heart rate and cardiac output. Beta Blockers, for example, may act primarily in the heart itself.

[0024] In some implementations, the disclosed systems and methods use two pressure sensors. For example, one sensor can be in or associated with the right side of the heart and one in or associated with the left side of the heart. The first and second sensors can measure and calculate changes in pressure separately. In some examples, the pressure sensor that senses the right side of the heart may be implanted in the right atrium of the heart. In some examples, the pressure sensor that senses the right side of the heart may sense central venous pressure, right ventricle pressure, or be a fluid volume sensor. In some examples, the pressure sensor that senses the left side of the heart may be implanted in the left atrium. In some examples, the pressure sensor that senses the left side of the heart may sense the left ventricle pressure or arterial pressure.

[0025] The first and second pressure sensors can be separate, having separate anchors, such as the anchors disclosed in PCT Pub. Nos. WO2022246161, WO2022246163, WO2022246166, WO2022246168, WO2022246169, W02022240603, WO2022197454, or WO2022197455, the entireties of each of which are hereby incorporated by reference. Alternatively, the first and second pressure sensors can share an anchor. The first and second pressure sensors can span one or more portions or chambers of the heart. The pressure sensors can span the septum or the left atrial wall with the coronary sinus, such as the implanted pressure sensors described in US Pat. Pub. No. 2021 / 0361238 or US Pat. Pub. No. 2023 / 0022499, the entireties of which are hereby incorporated by reference.

[0026] The system may use these two parts of the information to simultaneously titrate two or more drugs up or down, one from a first group (e.g., ARNI), and one from a second group (e.g., a diuretic such as Lasix). It is important to note that the pressure sensors of systems disclosed herein can be configured to sample pressure 100 times or more per second , and continuous waveforms may be recorded from both the right and left sides of the heart simultaneously. The signals can therefore be synchronized and timing differential in peak pressure detected by the system and hemodynamic attributes can be derived from the signals. In some examples, the system may include one implantable sensor and one wearable sensor. For example, the left heart sensor can be implanted in the left atrium of the patient’s heart and the right heart sensor can be a non-invasive central venous pressure sensor worn on the patient’ s neck. The central venous pressure sensor may monitor jugular vein pressure.

[0027] Further, more than two drugs may be managed using the first and second pressure sensors. The treatment of heart failure may include 5-6 drugs, and the dosing and drugs taken can be varied from patient to patient, from day to night, day to day, week to week. A major issue is that one drug can affect systemic blood pressure, but also indirectly affects systemic blood volume, and vice versa, and also affects renal function and all other related health complications and vice versa. The interaction between pressure, volume, and / or renal function may be patient specific. And response to drug may also be patient specific.

[0028] Error! Reference source not found, is a flow diagram illustrating a process for managing medication dosing based on left- and right-side pressure readings in accordance with aspects of the present disclosure. The various operations associated with FIG. 1 can be implemented at least in part by one or more users / individuals (e.g., medical practitioners) and / or control circuitry embodied one or more components of any of the systems disclosed herein. The term “control circuitry” is used herein according to its broad and ordinary meaning, and may refer to any collection of processors, processing circuitry, processing modules / units, chips, dies (e.g., semiconductor dies including one or more active and / or passive devices and / or connectivity circuitry), microprocessors, micro-controllers, application-specific integrated circuits (ASIC), digital signal processors, microcomputers, central processing units, field- programmable gate arrays, programmable logic devices, state machines (e.g., hardware state machines), logic circuitry, analog circuitry, digital circuitry, and / or any device that manipulates signals (analog and / or digital) based on hard coding of the circuitry and / or operational instructions. Control circuitry referenced herein may further include one or more circuit substrates (e.g., printed circuit boards), conductive traces and vias, and / or mounting pads, connectors, and / or components. Control circuitry referenced herein may further comprise one or more storage devices, which may be embodied in a single memory device, a plurality of memory devices, and / or embedded circuitry of a device. Such data storage may comprise readonly memory, random access memory, volatile memory, non-volatile memory, static memory, dynamic memory, flash memory, cache memory, data storage registers, and / or any device that stores digital information. It should be noted that in embodiments in which control circuitry comprises a hardware and / or software state machine, analog circuitry, digital circuitry, and / or logic circuitry, data storage device(s) / register(s) storing any associated operational instructions may be embedded within, or external to, the circuitry comprising the state machine, analog circuitry, digital circuitry, and / or logic circuitry. Furthermore, control circuitry described or referenced herein can comprise, and / or be configured to execute instructions stored on, anytype of computer-readable media / medium, wherein execution of such instructions may cause implementation of any of the process operations described herein. The term “computer- readable medium” is used herein according to its broad and ordinary meaning, and may refer to any non-transitory storage medium that electronically registers or stores information in a form readable and accessible by one or more computing devices or other control circuitry. For example, a computer-readable medium may include, but is not limited to, magnetic storage devices (e.g., hard disk, floppy disk, magnetic strips), optical disks (e.g., compact disk (CD), digital versatile disk (DVD), Blu-ray disk (BD)), smart cards, flash memory devices (e.g., card, stick, key drive), random access memory (RAM), read only memory (ROM), programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), registers, removable disks, solid state drives, or any other suitable medium for storing software and / or instructions that may be accessed and read by a machine or processor. Computer-readable medium used in connection with examples disclosed herein may also include, for example, one or more carrier waves, transmission lines, and / or any other suitable medium for transmitting software and / or instructions that may be accessed and read by computing devices or other control circuitry. In some implementations, computer-readable medium may be resident in one or more processors, external to a processor, or distributed across multiple entities including a one or more processors. Computer-readable media associated with systems and processes disclosed herein may be embodied in a computer program product, for example.

[0029] In connection with FIG. 1 or any process or method disclosed herein, the term “block” is used to refer to a functional component or segment within a flow diagram or process flow, representing one or more processes, actions, steps, operations, or computations associated with a portion of the described method or process, without necessarily implying a specific temporal sequence or order relative to other blocks. Flow diagram blocks described herein may include, or have associated therewith, any number of operations, data transformations, or interactions performed concurrently or in varying orders, and may be implemented in hardware, software, or a combination thereof.

[0030] At block 102, a patient is provided initial baseline doses of a first drug and a second drug. The initial baseline doses can include physician-preprogrammed baseline doses of a systemic blood pressure drug (e.g., angiotensin receptor-neprilysin inhibitor (ARNI)) and a systemic volume load drug (e.g., Lasix), and may be aligned with Guideline Directed Medical Therapy (GDMT) protocols, which may be encoded as parameters in a non-transitory computer-readable medium of dosage management system as described herein. These baselinedoses can be determined and / or administered by control circuitry within a networked dualsensor system, as described in detail herein, which can advantageously integrate high- frequency pressure waveform data from synchronized left and right heart sensors to enable precise, automated titration.

[0031] After the initial dosing, in block 104, the left heart blood pressure is measured. The left heart blood pressure may be measured by measuring the blood pressure at the left atrial of the patient’s heart. Further, after the initial dosing and in parallel with block 104, at block 106, the right heart blood pressure is measured. The right heart blood pressure may be measured by measuring the blood pressure at the right atrial pressure. The method described in connection with FIG. 1 may be performed using the system described below in connection with FIG. 2.

[0032] At block 108, a combination of the left heart blood pressure and the right heart blood pressure may be used to calculate the patient’s systemic blood pressure. It has been discovered that the left heart blood pressure is more indicative of a patient’s systemic blood pressure though a combination of left heart blood pressure and right heart blood pressure may be used to determine the patient’s systemic blood pressure. An algorithm may be applied to determine the patient’s systemic blood pressure which utilizes the left heart blood pressure and the right heart blood pressure as inputs. The algorithm to determine the patient’s systemic blood pressure may be a machine learning algorithm which utilizes previous correlations between various left heart blood pressure and right heart blood pressure and system blood pressure as training data. Further, at block 109, a combination of the right heart blood pressure and the left heart blood pressure may be used to calculate the patient’s systemic volume load. It has been discovered that the right heart blood pressure is more indicative of a patient’s systemic volume load thought a combination of right heart blood pressure and left heart blood pressure may be used to determine the patient’s system blood pressure. An algorithm may be applied to determine the patient’s systemic volume load which utilizes the left heart blood pressure and the right heart blood pressure as inputs. The algorithm to determine the patient’s systemic volume load may be a machine learning algorithm which utilizes previous correlations between various left heart blood pressure and right heart blood pressure and systemic volume load as training data.

[0033] In some examples, monitoring can be periodic. In some examples, monitoring can be continuous for a period of time), but dosing can be periodic. Dosing does not necessarily need to be performed each monitoring period. For example, a sensor may monitor 10 minsevery hour and all that data captured could be used for a single dosing determination the next morning.

[0034] At block 110, the system determines the patient’s systemic blood pressure as low, normal, or high. If the patient’s systemic blood pressure is low, at block 110a, then the first drug dosage is lowered. If the patient’s systemic blood pressure is normal, at block 110b, then the first drug dosage is maintained. If the patient’s systemic blood pressure is high, at block 110c, then the first drug dosage is raised. The first drug may be ARB, ACEi, and ARNI.

[0035] At block 112, the system determines the patient’s systemic blood volume load as low, normal, or high. If the patient’s systemic blood volume load is low, at block 112a, then the second drug dosage is lowered. If the patient’s systemic blood volume load is normal, at block 112b, then the second drug dosage is maintained. If the patient’s systemic blood pressure is high, at block 112c, then the second drug dosage is raised. The second drug may be a diuretic. The second drug may be LASIX.

[0036] At block 114, for the first drug dosage determined at block 110 is provided to the patient or physician. Further, at block 114, for the second drug dosage determined at block 112 is provided to the patient or physician. Optionally, the first and second drugs are administered to the patient according to the recommendations.

[0037] At block 116, the operations associated with blocks 104-114 may be repeated at a specific frequency (e.g., once a day, week, etc.). Other frequencies may be utilized such as once an hour. The operations / steps associated with the various blocks may be repeated intermittently through the day, once daily, or less often. The operations / steps associated with the various blocks may be utilized to determine alterations in dosing with periodicity (e.g., daily). The operations / steps associated with the various blocks may automate dosing determinations that are generally very difficult to determine due to the highly interactive and dynamic nature of heart failure.

[0038] The patient’ s physician may supervise the protocol by preprograming the algorithm with the recommended baseline doses for medication and how much to change them in response for changes in volume and pressure. The physician can also set the normal expected pressure and volume for the patients and the maximum dosages for each drug. When the maximum dosage is set for each drug, if it is determined to raise the dosage of the drug such as in blocks 110c and / or 112c, these operations may not occur if the maximum dosage for each of the drugs would be reached or exceeded. Instead, the dosage of the first drug and / or the second drug may be set at the maximum dosage.

[0039] Error! Reference source not found, is a diagram of a drug dosage system 200. The drug dosage system 200 includes a dual-sensor implant device or subsystem implanted, at least in part, in a heart of a patient, with electronic circuitry configured to record and / or transmit pressure waveforms from each sensor. The dual-sensor subsystem can include a right atrium pressure sensor 202 and a left atrium pressure sensor 204 which each electronically communicate with a transmitter 206. The dual-sensor system is implanted at least partly in the patient’s heart, with electronic circuitry configured to record and transmit the pressures waveforms from both sensors. The benefit of two sensors is that they may provide two locations and / or timepoints in the cardiovascular system, which can better interpret the hemodynamics (or more generally, the fluid dynamics). Better capture of the dynamics can better determine optimization of two different drugs that would have a primary effect, and secondary effects.

[0040] The clinical setting may not be useful for heart failure management. Thus, catheterization systems and clinical monitoring systems may not be useful. It is advantageous for the sensors 202, 204 to be implants and / or wearable devices. The pressure sensor implant system, as part of the heart failure and drug dosage management system, comprises one or more implantable devices configured to measure synchronized pressure waveforms from the left and right sides of the heart, respectively, such as pressure conditions of the left atrium and right atrium. Such sensor measurements can be used to provide inputs for a machine learning model as described in detail herein, and may include additional physiological parameter sensing capabilities. In some implementations, the sensor system includes a single implant device anchored to the atrial septum at a suitable location, such as the fossa ovalis, incorporating at least two pressure sensors (e.g., MEMS transducers) positioned on opposite sides of the septum to measure left and right atrial pressures simultaneously, with the implant secured using expandable anchoring mechanisms such as wire mesh, arms, cloth, sutures, or stents. In some examples, the implant comprises an expandable nitinol mesh that sandwiches the septum for stable fixation.

[0041] In some implementations, the sensor system may comprise two or more separate implants, each with at least one pressure sensor, deployed in distinct locations (e.g., one in the left atrium or ventricle and one in the right atrium or ventricle), communicating either via wire tethers for direct signal transmission or wirelessly using low-power protocols (e.g., Bluetooth Low Energy or near-field communication) to synchronize data collection.

[0042] Sensor systems disclosed herein can comprise any suitable or desirable device types and configurations for capturing left and right heart pressure waveforms, providing flexibilityin achieving accurate hemodynamic data as inputs for a machine learning model. For example, the system 200 may employ a single dual-sensor implant, as described above, wherein such device may be anchored to the atrial septum (e.g., fossa ovalis) with pressure sensors (e.g., MEMS transducers) positioned on both sides to measure left atrial pressure (LAP) and right atrial pressure (RAP) simultaneously, secured using expandable mechanisms like wire mesh, stents, or sutures. Alternatively, the system may use two (or more) separate implantable sensors, one placed in the left heart (e.g., left atrium or left ventricle via transcatheter delivery) and another in the right heart (e.g., right atrium or right ventricle). In some implementations, one sensor may be implantable (e.g., a left atrial sensor) while the other is a non- invasive wearable device, such as a central venous pressure sensor (e.g., worn on the patient’s neck to monitor jugular vein pressure), deducing right heart pressure non-invasively through optical, ultrasound, or impedance-based sensing techniques. Additional options include wearable arterial pressure sensors (e.g., wrist-based cuffs or patches) used to estimate left heart pressure, or fluid volume sensors (e.g., bioimpedance sensors on the chest) used to infer right heart volume status, each integrated into the system to provide supplementary physiological inputs. In terms of intracardiac / intravascular implantation, the various sensors disclosed herein can be implanted in the right atrium, right ventricle, superior vena cava, inferior vena cava, left atrium, left ventricle, pulmonary artery, aorta, and / or other anatomy.

[0043] The sensor implant system can include signal conditioning circuitry, including circuitry configured to function as an analog-to-digital converter (ADC) for digitizing pressure waveforms into digital data / value arrays, a low-noise amplifier to boost sensor signals, a digital signal processing (DSP) module for noise filtering (e.g., low-pass filtering) and / or synchronization of left and right waveforms, and / or data compression circuitry configured to optimize transmission bandwidth. The transmitter 206 may be integrated into one of the sensor devices, serving both sensors, or implemented as a separate subcutaneous or external relay unit that communicates bidirectionally with the sensors 202, 204 to collect and transmit data to the receiver 210, potentially providing inductive power to the implants via wireless charging protocols. In some implementations, at least one sensor may be external to the body, such as a wearable central venous pressure sensor (e.g., monitoring jugular vein pressure on the neck) or a non-invasive arterial pressure sensor, configured to deduce left or right heart pressure or other physiological parameters (e.g., heart rate, oxygen saturation) as supplementary inputs to the machine learning model, providing flexibility across various clinical settings while maintainingsynchronized, high-fidelity data for systemic blood pressure, volume load, and medication dosage determinations / predictions.

[0044] In some examples, the sensor implant system includes one or more antennas and / or coils for receiving power. In some examples, the antenna is configured to receive power from radio frequency (RF) signals, ultrasound power, inductive power, etc. In some examples, the sensor implant system can include a battery or capacitor for storing power. In some examples, the battery or the capacitor is stored within a housing of a sensor or stored externally from the housing of the sensor. In some embodiments, the antenna can be dual purposed to receive power and transmit data.

[0045] The transmitter 206 is configured to communicate with an external receiver 210. The receiver 210 is in communication with, or integrated with, a smart device 212 which may be in communication with a physician’s device / system 216 through a network connection 214. The receiver 210 reads the measurements from the sensors 202, 204 and sends them to the smart device 212. An application on the smart device 212 that receives the measurements from the receiver 210 may send the measurements to a cloud server. The cloud server may calculate systemic blood pressure and systemic blood volume levels and recommended drugs doses and provide them to the patient via the smart device 212 for display thereon. In some examples, the smart device 212, a separate remote monitor / server, and / or the physician’s device 216 may calculate systemic blood pressure and systemic blood volume levels and recommended drugs dosages. Each of these operations may be spread out on different devices. For example, the smart device 212 may calculate systemic blood pressure and systemic blood volume levels and the physician’s device 216 or other remote server may recommend drug dosages.

[0046] In some examples, the remote monitoring server (e.g., cloud server) may analyze the data uploaded from the smart device 212 and return calculated pressures and volumes. The remote monitoring server may also calculate the recommended drug doses. The remote monitoring server may also relay the patient’s data to the physician’s device 216. In cases of extreme changes in pressure or volume an alert can be sent to the physician’s device 216. In some examples, the system 200 may be a closed loop system where the system automatically doses drugs to the patient without input from the physician’s device 216 or the smart device 212. In closed loop systems, the automatic dosing of one or more drugs can be more reactive to changes in patient conditions than a physician reviewing data could, which can result in outcomes better than a physician making a decision based on received data.

[0047] A trained machine learning model may be utilized to calculate the systemic blood pressure changes and systemic blood volume changes from the dual sensor pressure waveform readings, and derive the correct doses for the first drug and the second drug. In some examples, a machine learning model may be trained using data from a large cohort of patients. In some examples, the training data may include a large sample size of patients that includes information about their systemic blood volume level (e.g., weight, fluid intake, urine output, symptoms of edema or pulmonary congestion), their systemic blood pressure (e.g., daily arterial pressure from an arm cuff) and records of all their medications can be used to train an algorithm to find correlations between different drug dose combinations given to patients across a spectrum of systemic blood volume levels and systemic blood pressures.

[0048] In some examples, a large sample size of patients may be implanted or otherwise fitted with dual-sensor implants / systems and tracked as their medication dosages change from time to time until treatment is optimized to minimal symptoms. The data collected from the implanted sensors in correlation with the given medications may be the training data for the machine learning algorithm.

[0049] In some examples, the machine learning algorithm may be trained using the patient’s own data. For example, a patient may be implanted with a dual sensor implant and may be followed for a time period (such as a few weeks) recording sensor data, medication doses, fluid intake, urine output, arterial pressure, and / or any symptoms of edema, congestion, or other heart failure symptoms. During this time period, the patient’s doctor may titrate a first drug (e.g., ARNI) and a second drug (e.g., diuretic) up, down, or maintained at a current level in order to optimize the dosage blinded to the implant sensor data. After an initial training period, data may be continually collected from the sensors and the patients record may be used to continually train the algorithm to deduce the patient’s status and optimal medication doses from the patient’s dual sensors reading.

[0050] A combination of the above-mentioned training models may be utilized. For example, previous patient data may be used along with the patient’ s own data in order to train the machine learning model. A database of different patient data may be accumulated on the cloud server which may be utilized to improve the machine learning model.

[0051] In various examples, the receiver 210 and / or the smart device 212 may form part of a local monitoring system, which provides a subsystem configured to handle initial data acquisition, preliminary processing, and / or user interaction in proximity to the patient, thereby enabling real-time or near-real-time feedback without sole reliance on remote / cloud resources.The term “real-time” is used herein according to its broad and ordinary meaning and may refer to the processing, analysis, and / or generation of outputs by a system, including but not limited to medication dosage recommendations, physiological parameter predictions (e.g., systemic blood pressure and volume load), or related data, with minimal latency following the acquisition of input data, such as pressure waveforms or other physiological signals from sensors. For example, this processing may occur within a timeframe sufficient to support timely clinical decision-making, automated interventions, or user feedback. Real-time data acquisition or generation in connection with examples disclosed herein has a latency of less than 10 minutes from the time of manifestation of a measured physiological signal / event or from a data collection or provision time for physiological parameter or medication dosage calculation based upon such data collection or provision.

[0052] The monitor / monitoring circuitry, which may be responsible for generating medication dosage values or recommendations, may be embodied at least in part within the control circuitry of the local monitoring subsystem, utilizing one or more processors (e.g., ARM-based or multi-core CPUs), non-transitory computer-readable media storing executable instructions, and communication modules to process synchronized pressure waveforms and volume status metrics, applying a machine learning dosage generation and / or systemic pressure and volume data generation engine to compute predicted systemic blood pressure, systemic volume load, and optimal dosages for ARNI and Lasix, or similar medication(s).

[0053] In some implementations, the receiver 210 may be integrated with the smart device 212, such as a smartphone, tablet, wearable smartwatch, or dedicated handheld monitor, leveraging the device’s built-in antennas (e.g., Bluetooth, NFC, or Wi-Fi) to receive the transmitted pressure waveforms from the dual-sensor implant / system, thereby streamlining the system architecture and reducing the need for separate hardware components. Alternatively, the receiver 210 may be configured as a standalone wand, puck, or other antenna-based communication device (e.g., wearable strap or article) that wirelessly interfaces with the sensor implant / system. In addition to receiving sensor signals, the receiver 210 may be configured to also transmit commands (e.g., to initiate measurements or adjust sensor parameters) to the sensor system or provide inductive power to the sensor system via wireless charging protocols, such as Qi or resonant coupling, for example to recharge an implant’ s battery or enable passive operation in power-constrained designs.

[0054] The smart device 212, as part of the local monitoring system, may receive (e.g., from a remote monitoring subsystem over the network connection 214) or generate userinterface data that renders pressure and / or medication dosage information, facilitating intuitive visualization for patients or caregivers through graphical displays, numerical readouts, or trend charts generated using outputs from the machine learning dosage generation engine (e.g., recommended ARNI and Lasix doses in mg) and / or systemic pressure (e.g., predicted values in mmHg with high / normal / low labels) and volume load data generation engine (e.g., composite scores incorporating volume status metrics). The smart device 212, which may be a smartphone, tablet, or wearable device, can be configured to present any type of user interface on its display that renders physiological and / or medication dosage information derived from the machine learning model, facilitating intuitive patient and caregiver interaction with realtime data. For example, the user interface can display elements including left atrial pressure (LAP) and right atrial pressure (RAP) as numerical values or graphical trends (e.g., time-series plots), systemic blood pressure as a numerical value with a categorical label (e.g., high, moderate, low), and systemic volume load as a composite score similarly labeled, providing clear visualization of hemodynamic status. Dosage recommendations for the intravascular volume reducing medication (e.g., Lasix) and the blood pressure reducing medication (e.g., ARNI) can be presented, and such presentations can include directional arrows (e.g., upward or downward) or color-coded icons to indicate dose adjustments relative to prior values. Other intuitive visual cues, such as progress bars or alert symbols, can be included to highlight significant changes in physiological parameters or dosages. In some implementations, the user interface may incorporate tactile feedback (e.g., haptic vibrations) or audible signaling (e.g., beeps or voice prompts through the device’s speaker to indicate dosage updates or warnings).

[0055] In embodiments where computational resources are distributed, the local monitoring system’s control circuitry may perform preliminary calculations (e.g., basic waveform feature extraction like peak detection) on the smart device 212 before uploading data to the cloud server for advanced machine learning model (e.g., CNN) inference. In fully local implementations, the control circuitry of the computing device 212 may execute the entire machine learning model on-device using optimized frameworks (e.g., TensorFlow Lite), which can promote privacy and / or offline functionality. This flexible configuration allows the system to be adapted for various clinical scenarios, such as home-based monitoring with minimal external dependencies or integrated hospital setups where the receiver 210 communicates bidirectionally with the implant to synchronize data collection with patient activities, while adhering to security protocols like encryption to protect sensitive health information.

[0056] The drug dosage system 200 can include various non-volatile data stores configured to persistently store various types of patient-related and operational data, providing data integrity, accessibility, and scalability across different embodiments of the system. Nonvolatile data stores disclosed or contemplated herein can comprise one or more solid-state drives (SSDs), hard disk drives (HDDs), flash memory, or distributed cloud storage solutions, and may be implemented as relational databases (e.g., SQL-based), non-relational databases (e.g., NoSQL like MongoDB), or file-based systems (e.g., HDF5 for structured waveform data). These data stores can accumulate and organize patient-specific data, such as personal identifiers (e.g., age, gender, medical history), demographic information (e.g., ethnicity, geographic location), historical logs of sensor readings (e.g., synchronized left and right heart pressure waveforms stored as floating-point values with timestamps), volume status metrics (e.g., a composite data structure including weight, fluid intake and urine output, and / or symptom flags), and GDMT-related parameters (e.g., baseline and maximum dosages for medications, along with normal ranges for systemic blood pressure and volume load scores). Population-level data, including anonymized datasets from cohorts of heart failure patients comprising aggregated pressure waveforms, systemic blood pressure values, volume load metrics, and medication records, may also be stored to support machine learning model training and refinement, enabling the system to learn correlations across diverse patient profiles.

[0057] In various embodiments, the non-volatile data store(s) may be embodied at least partially locally on the smart device 212 (e.g., using embedded flash memory for real-time caching of daily sensor logs), remotely on a cloud server or other remote monitor (e.g., via distributed storage services for scalable access to large training datasets), or in a hybrid configuration where patient-sensitive data is stored locally for privacy while aggregated population data resides in a remote server for collaborative model updates. For example, a cloud server may host a central repository that securely stores encrypted patient data relayed from the physician’s device, allowing for federated learning where models are trained on decentralized data without compromising individual privacy. The non-volatile data stores of the system can facilitate continuous system operation by enabling retrieval of historical trends for patient-specific algorithm fine-tuning, integration with electronic health records (EHRs) for GDMT compliance checks, and generation of alerts based on deviations from stored normal ranges. Security measures, such as encryption (e.g., AES-256) and access controls (e.g., rolebased authentication), can be used to facilitate HIPAA-compliant handling, while redundancy (e.g., RAID configurations or cloud replication) can be implemented to prevent data loss,allowing the system to be implemented across any combination of devices, including wearable sensors, edge computing nodes, or centralized servers, to accommodate varying clinical environments and resource constraints.

[0058] The machine learning model, which may advantageously serve as both a systemic blood pressure data and systemic volume load data generator and a medication dosage engine, is embodied in control circuitry comprising one or more processors (e.g., multi-core CPUs, GPUs, or microcontrollers) and non-transitory computer-readable media storing executable instructions, and can be implemented across various system components, including the local monitoring system (e.g., smart device 212), the remote (e.g., cloud) monitoring system / server, or, in specialized configurations, partially on the dual-sensor implant, though the implant’s limited computational capacity may restrict the sensor control circuitry to preprocessing tasks. The model(s) (e.g., convolutional neural network (CNN)) can be configured to process synchronized pressure waveform inputs (e.g., arrays of values from left and right heart sensors) and volume status metrics (e.g., a composite data structure with weight, fluid intake, urine output, and / or symptom flags), along with certain constraint parameter(s), which may be physician-preprogrammed parameter(s) such as parameter(s) defining baseline and maximum medication (e.g., ARNI and Lasix) dosages and normal ranges for systemic blood pressure and volume load vales. The model(s) can generate outputs including predicted systemic blood pressure, predicted systemic volume load, and recommended medication dosages (e.g., ARNI and Lasix), optionally along with high / normal / low categorical labels and / or confidence scores. Such functionality can advantageously enable dynamic GDMT-compliant dose titration.

[0059] The machine learning model’s parameters, such as weights and biases in a neural network’s convolutional, pooling, and fully-connected layers, can be trained using a clinical dataset comprising historical patient logs (e.g., pressure waveforms, systemic blood pressure, volume load, and medication records) from individual patient histories or population- level demographic cohorts, stored in nonvolatile data stores (e.g., HDF5, SQL, or cloud-based NoSQL databases). Training may involve supervised learning with backpropagation to optimize correlations between inputs and outputs, supporting both population-based generalization and patient-specific fine-tuning, and can be distributed across local (e.g., smart device using TensorFlow Lite) and / or remote (e.g., cloud server with GPU acceleration) subsystems to balance computational efficiency, privacy, and real-time performance, with secure data handling providing compliance with medical standards.

[0060] The heart failure management and drug dosage system 200 may include an automatic medication dispenser with one or more reservoirs (e.g., cartridges or vials) for dispensing one or more of the categories of medication described herein (e.g., ARNI and Lasix) based on dosage values generated by the machine learning model. For example, the automatic medication dispenser can be configured as a portable unit, standalone device, wearable patch pump, or implantable system, wherein the dispenser uses control circuitry (e.g., microcontroller / ASIC) to actuate dispensing mechanisms (e.g., motorized pill counters or microfluidic pumps) in response to model outputs received via wired or wireless communication from the smart device or cloud server. The dispenser can promote accurate dosing with safety features (e.g., lockout mechanisms based on physician-set maximums), log dispensing events in local or cloud-based data stores, and / or send alerts for errors or low reservoirs to the smart device 212 or physician’s device 216, enabling automated, precise medication delivery tailored to predicted systemic blood pressure and volume load.

[0061] FIG. 3 is a flowchart of an example method for administering two medications. At block 302, an initial dosage of a first drug and an initial dose of a second drug is provided to patient. At block 304, the left heart blood pressure of the patient is measured. Measuring the left heart pressure may include measuring the pressure at the left atrium or ventricle of a heart. Measuring the right heart pressure may include measuring arterial pressure.

[0062] At block 306, the right heart blood pressure of the patient is measured. Measuring the right heart pressure may include measuring the pressure at the right atrium or ventricle of a heart. Measuring the right heart pressure may include utilizing a non-invasive central venous pressure sensor worn on a neck.

[0063] At block 308, the dosage of the first drug is adjusted based on the left heart pressure data and right heart pressure data. In some examples, the systemic blood pressure may be computed from the left heart pressure data and the right heart pressure data. The systemic blood pressure may depend more heavily on the left heart pressure data than the right heart pressure data. Adjusting the first drug dosage may be based on the systemic blood pressure. Adjusting the first drug dosage may be further based on a systemic blood volume load. Computing the systemic blood pressure of the patient from the left heart pressure data and the right heart pressure data may be performed using a machine learning algorithm. The machine learning algorithm may be trained with data including previous correlations between a plurality of left heart blood pressure and right heart blood pressure and system blood pressure.

[0064] At block 310, the dosage of the second drug is adjusted based on the left heart pressure data and right heart pressure data. The systemic blood volume load may be computed from the left heart pressure data and the right heart pressure data. The systemic blood pressure may depend more heavily on the right heart pressure data than the left heart pressure data. Adjusting the second drug dosage may be based on the systemic blood volume load. Adjusting the first drug dosage may be further based on a systemic blood pressure. Computing the systemic blood volume load of the patient from the left heart pressure data and the right heart pressure data may be performed using a machine learning algorithm. The machine learning algorithm may be trained with data including previous correlations between a plurality of left heart blood pressure and right heart blood pressure and system blood volume load.

[0065] In some examples, if the systemic blood pressure is low, adjusting the first drug dosage lower, if the systemic blood pressure is normal, keeping the same dosage of the first drug, or if the systemic blood pressure is high, adjusting the first drug dosage higher.

[0066] In some examples, if the systemic blood volume load is low, adjusting the second drug dosage lower, if the systemic blood volume load is normal, keeping the same dosage of the second drug, or if the systemic blood volume load is high, adjusting the second drug dosage higher.

[0067] In some examples, adjusting the dosage of the first drug and adjusting the dosage of the second drug are performed utilizing a machine learning algorithm. The machine learning algorithm may be trained using previous data including systemic blood pressures correlated with usage of dosages of the first drug and systemic blood volume levels correlated with usage of a dosages of the second drug. The machine learning algorithm may be trained by data produced by: monitoring a patient for a period of time while recording the systemic blood pressures and systemic blood volume levels while recording medication doses, fluid intake, urine output, arterial pressure, and / or symptoms of disease; and titrating a dosage of the first drug and a dosage of the second drug over a period of time at different levels while continuing the record the systemic blood pressures and systemic blood volume levels.

[0068] At block 312, an adjusted dosage of the first drug and an adjusted dosage of the second drug is provided to the patient.

[0069] The blocks 304-312 may be repeated in a specific interval such a once a day. The method described in connection with FIG. 3 may be performed using the system described in connection with FIG. 2.

[0070] While the system is described in terms of a first drug which regulates systemic blood pressure and a second drug which regulates systemic blood volume load, many HF drugs affect both blood pressure and blood volume. Diuretics, for example, work in the kidneys by increasing urine output. This primarily reduces volume by evacuating water, but inevitably results also in reduction in blood pressure. Vasodilator drugs work primarily to relax the smooth muscles in the arteries walls. The vasodilators may reduce blood pressure, but can lead to reduction in volume in some patients and increase in volume in others. Volume may not be 1:1 correlated to right heart pressures to diuretic dose, and systemic pressure may not be 1 :1 correlated to left heart pressure or ARNI dose. Some heart failure drugs work on the interfere with the cardio-renal hormonal chemical communication, and can create more or less coupling between the heart and the kidneys. With some heart failure drugs may utilize both inputs of pressure and volume to optimize. A system with both inputs of pressure and volume from two sensors and an algorithm that uses both inputs to calculate multiple drug doses may be superior to two independent systems side by side (e.g., a first independent system which merely senses pressure and calculates dosage of a first drug and a second independent system which merely senses volume and calculates dosage of a second drug). The algorithm may take into account the inter-drug influence of each drug.

[0071] The algorithm may include a control loop. The control loop may be a negative feedback loop with inputs from both sensors that keep both blood pressure and blood volume in a target range. The algorithm may be an empiric algorithm. The algorithm may collect data on a large cohort of patients and feed this into a machine learning algorithm to produce an optimal drug dose combinations for specific histories of the two inputs of blood pressure and blood volume.

[0072] The algorithm may be adaptive. The algorithm may track specific patient responses to the applied drugs and adjust the drug dosage based on these responses. Patient drug response may change over time. Patients may develop resistance to drugs and may need higher dosages over time to get the same effect.

[0073] As discussed below with reference to FIG. 4, a machine learning model of the drug dosage system can be trained to determine systemic blood pressure of the patient from both the right heart pressure and the left heart pressure of the patient, and trained to determine systemic blood volume load of the patient from both the right heart pressure and the left heart pressure of the patient.

[0074] FIG. 4 is a diagram of a clinical dataset 400 used for data mining and machine training of the drug dosage system. The clinical dataset 400 includes one or more data structures representing a plurality of samples gathered from a set of clinical patients. Each sample of the plurality of samples can include a first pressure waveform 402 from a right side of a heart of a clinical patient, a second pressure waveform 404 from a left side of the heart of the clinical patient, a systemic blood pressure 406 of the clinical patient, a systemic blood volume load 408 of the clinical patient, a first dosage 410 of a first drug, and a second dosage 412 of a second drug.

[0075] The clinical dataset 400 can be implemented as a relational database or structured data frame (e.g., in HDF5 or SQL format) stored on a non-transitory computer-readable medium of control circuitry of any system component described herein. Each sample of the dataset 400 can comprise a record with a plurality of variable- or fixed-length fields. In some implementations, the pressure waveforms (402, 404) can be stored as floating-point time-series arrays (e.g., 32-bit floats), whereas systemic blood pressure (406) and / or volume load (408) data structures may be labeled floating-point scalars. The dosages (410, 412) can be stored as integer or floating-point values representing, for example, milligrams of medications. The dataset 400 can advantageously be indexed by patient ID and / or timestamp to enable efficient querying and batch processing during machine learning model training.

[0076] For each sample of the dataset 400, the first pressure waveform 402 and the second pressure waveform 404 in each sample can be gathered from the clinical patient simultaneously such that the first pressure waveform 402 and the second pressure waveform 404 share a common time period. The first pressure waveform 402 from the right side of the heart of the clinical patient was gathered from the clinical patient using a first sensor similar to the sensor 202 described above. For example, the first sensor can be implanted into the clinical patient using a transcatheter implantation process and configured to measure the first pressure waveform 402 of the clinical patient. In some examples, the first sensor can be a non-invasive central venous pressure sensor worn on a neck of the clinical patient and configured to measure the first pressure waveform 402 of the clinical patient. The first pressure waveform 402 of the clinical patient can represent the pressure at the right atrium or right ventricle of the heart of the clinical patient. In some examples, the first pressure waveform 402 of the clinical patient can represent the arterial pressure of the clinical patient.

[0077] For each sample of the clinical dataset 400, the second pressure waveform 404 from the left side of the heart of the clinical patient may be gathered from a clinical patient using asecond sensor similar to the sensor 204 described above. For example, the second sensor can be implanted into the clinical patient using a transcatheter implantation process and configured to measure the left heart blood pressure of the clinical patient. The second pressure waveform 404 of the clinical patient can represent the pressure at the left atrium or left ventricle of the heart of the clinical patient.

[0078] In each sample of the clinical dataset 400, the systemic blood pressure 406 of the clinical patient can be gathered using a third sensor, or can be derived using a machine learning model as described herein. In some examples, the third sensor can be an arm-cuff hemodynamic pressure monitor that measures arterial pressure from an arm of the clinical patient. For each sample, the third sensor can measure the arterial pressure of the clinical patient within the same common time period of the first sensor and the second sensor, and the measured arterial pressure can be saved to the sample as the systemic blood pressure 406 of the clinical patient. The systemic blood pressure 406 in each sample can be labeled relative to a high-pressure threshold and relative to a low-pressure threshold.

[0079] In each sample of the clinical dataset 400, the systemic blood volume load 408 of the clinical patient can be based upon weight data of the clinical patient, fluid intake data of the clinical patient, urine output data of the clinical patient, symptoms of edema in the clinical patient, and / or symptoms of pulmonary congestion in the clinical patient during the time period of the respective sample (i.e., during the time period of the first pressure waveform 402 and the second pressure waveform 404). The systemic blood volume load 408 in each sample can be labeled relative to a high-volume threshold and relative to a low-volume threshold.

[0080] The first dosage 410 of the first drug in each sample of the clinical dataset 400 can be a dosage of a drug for managing systemic blood pressure in the clinical patient, such as an ARNI, that is active in the clinical patient at the time the respective sample is taken (i.e., during the time period of the first pressure waveform 402 and the second pressure waveform 404). The second dosage 412 of the second drug in each sample of the clinical dataset 400 can be a dosage of a drug for managing systemic blood volume load in the clinical patient, such as a diuretic, that is active in the clinical patient at the time the respective sample is taken (i.e., during the time period of the first pressure waveform 402 and the second pressure waveform 404).

[0081] FIG. 5 is a flow diagram of method 500 for data mining the clinical dataset 400 of FIG. 4 for machine training a machine learning model of a drug dosage system. For example, the method 500 can be applied to the first pressure waveform 402 and the second pressurewaveform 404 in each of the samples in clinical dataset 400 to train a drug dosage system to find input features for the machine learning model of the drug dosage system that can accurately determine the systemic blood pressure and the systemic blood volume load of a patient.

[0082] To machine train the drug dosage system to generate / determine the systemic blood pressure from the first pressure waveform 402 and the second pressure waveform 404 of the clinical patient, and to generate / determine the systemic volume load of the clinical patient from the first pressure waveform 402 and the second pressure waveform 404, the method 500 can be applied to each sample of the clinical dataset 400. At block 502, the method 500 involves performing waveform analysis of the first pressure waveform 402 and the second pressure waveform 404 in each sample collected in the clinical dataset 400 to calculate a plurality of signal measures. Performing waveform analysis of the first pressure waveform 402 and the second pressure waveform 404 in each sample can include identifying individual cardiac cycles in each of the pressure waveforms 402 / 404.

[0083] FIG. 6 provides a graph illustrating an example trace of a pressure waveform with an individual cardiac cycle identified and enlarged. Performing waveform analysis of the first pressure waveform 402 and the second pressure waveform 404 can include identifying a dicrotic notch in each of the individual cardiac cycles of each of the first pressure waveform 402 and the second pressure waveform 404 in each sample, similar to the example shown in FIG. 6. Waveform analysis on each sample can further include identifying a systolic rise phase, a systolic decay phase, and a diastolic phase in each of the individual cardiac cycles of each of the first pressure waveform 402 and the second pressure waveform 404, similar to the example shown in FIG. 6.

[0084] For each sample, signal measures are extracted from each of the systolic rise phase, the systolic decay phase, and the diastolic phase from each of the individual cardiac cycles of each of the first pressure waveform 402 and the second pressure waveform 404. The signal measures can correspond to hemodynamic effects from each of the systolic rise phase, the systolic decay phase, and the diastolic phase from each of the individual cardiac cycles. Those hemodynamic effects can include contractility, aortic compliance, stroke volume, vascular tone, afterload, and full cardiac cycle. The signal measures calculated or extracted by the waveform analysis operation(s) associated with block 502 of method 500 can include a mean, a maximum, a minimum, a duration, an area, a standard deviation, derivatives, and / or morphological measures from each of the systolic rise phase, the systolic decay phase, and thediastolic phase from each of the individual cardiac cycles. The signal measures can also include heart rate, respiratory rate, stroke volume, pulse pressure, pulse pressure variation, stroke volume variation, mean arterial pressure (MAP), systolic pressure (SYS), diastolic pressure (DIA), heart rate variability, cardiac output, peripheral resistance, vascular compliance, and / or left- ventricular contractility extracted from each of the individual cardiac cycles of each of the first pressure waveform 402 and the second pressure waveform 404 of each sample of clinical dataset 400.

[0085] After the signal measures are determined for each sample of clinical dataset 400, operation(s) associated with block 504 of method 500 can be performed on the signal measures of each sample. Block 504 of method 500 involves computing combinatorial measures between the signal measures of the first pressure waveform 402 and the second pressure waveform 404 of each sample. Computing the combinatorial measures between the signal measures can include performing operations associated with blocks 506, 508, 510, and 512 shown in FIG. 5 on all the signal measures of the first pressure waveform 402 and the second pressure waveform 404 of each sample. Block 506 is performed by arbitrarily selecting three signal measures from the signal measures. Next, different orders of power are calculated for each of the three signal measures to generate powers of the three signal measures, as shown at block 508 of FIG. 5. At block 510 of FIG. 5, the powers of the three signal measures are then multiplied together to generate the product of the powers of the three signal measures.

[0086] Block 512 involves performing receiver operating characteristic (ROC) analysis of the product to arrive at a combinatorial measure for the three signal measures. Blocks 506, 508, 510, and 512 are repeated until all of the combinatorial measures have been computed between all of the signal measures of the first pressure waveform 402 and the second pressure waveform 404 in each sample. The signal measures with most predictive top combinatorial measures (i.e., combinatorial measures satisfying a threshold prediction criteria) are selected as top signal measures for determining the systemic blood pressure and the systemic blood volume load of the clinical patient, and are labeled as the input features for the machine learning model of the drug dosage system. With the input features determined, the drug dosage system is trained or programmed to perform waveform analysis on both a pressure waveform of the right heart blood pressure of the patient and a pressure waveform of the left heart blood pressure of the patient and extract the input features from the right heart blood pressure and the left heart blood pressure of the patient, and use those input features to determine both a systemic blood pressure of the patient and a systemic blood volume load of the patient. The drug dosage system canthen utilize the systemic blood pressure of the patient and the systemic blood volume load of the patient to determine a dosage adjustment of the first drug for the patient, and to determine a dosage adjustment of the second drug for the patient.

[0087] In some examples, the machine learning model of the drug dosage system can be a deep learning artificial neural network, as shown in the method 700 of FIG. 7. To train the artificial neural network of the machine learning model of the drug dosage system, each sample of the dataset 400 can first be annotated with the systemic blood pressure 406 of the clinical patient and the systemic blood volume load 408 of the clinical patient, as shown at block 702 of the method 700 of FIG. 7. At block 703 of method 700, waveform analysis is performed on the first pressure waveform 402 and the second pressure waveform 404 of each sample of the clinical dataset 400 in a manner similar to the operation(s) associated with block 502 of method 500 of FIG. 5 as described above to generate the plurality of signal measures. At block 704 of the method 700, the plurality of signal measures of each of the samples are inputted into an input layer of the artificial neural network. At block 706 of the method 700, the plurality of signal measures of each of the samples is forward propagated through at least one hidden layer of the artificial neural network to compute at least one output of the respective sample. Each node of the at least one hidden layer of the artificial neural network receives inputs from the previous layer, performs a weighted sum of these inputs, applies an activation function to the weighted sum, and passes a result of the activation function to a next layer of the at least one hidden layer of the artificial neural network until the at least one output has been computed. The at least one output of the artificial neural network (or other machine learning model) can be an estimation of the systemic blood pressure of the clinical patient and an estimation of the systemic blood volume load of the clinical patient.

[0088] At block 708, the at least one output is backward propagated through the hidden layer of the artificial neural network to determine a cost function for each weight of every node of the at least one hidden layer, computing the gradient layer by layer, and iterating backward from the last layer to avoid redundant computation of intermediate terms in the chain rule. The systemic blood pressure 406 of the clinical patient and the systemic blood volume load 408 of the clinical patient can be used as desired outputs during backward propagation to determine the error in the at least one output. The artificial neural network can then update the weights of each node in the at least one hidden layer and rerun operation(s) associated with blocks 704- 708 of the method 700 until the error between the at least one output and the desired outputs has been minimized to an acceptable degree where the at least one output is substantially equalto the desired outputs. In this training of the artificial neural network, some signal measures of the plurality of signal measures will be minimized and weighted to zero, indicating that those signal measures are not predictive of systemic blood pressure of the clinical patient or not predictive of systemic blood volume load of the clinical patient. The signal measures that are predictive of systemic blood pressure or systemic blood volume load will eventually be selected as most predictive and given a great weight. At block 710 of the method 700, the most predictive signal measures are selected as inputs for the machine learning model of the drug dosage system for determining the systemic blood pressure and the systemic blood volume load of the patient.

[0089] Control circuitry of systems of the present disclosure can be configured to implement machine learning functionality for systemic blood pressure and volume load determination, as well as deriving optimal dosages for drug administration based thereon. For example, machine-leaming-derived functionality can be implemented to enable a dual-sensor, dual-drug treatment protocol for heart failure management. FIG. 8 illustrates a framework 800 for determining systemic blood pressure and systemic volume load, and for generating medication dosage guidance based thereon in accordance with one or more examples.

[0090] The framework 800 can process synchronized, high-frequency (e.g., 100 Hz or greater) pressure waveform data from dual implanted sensors (e.g., one in the right heart, such as the right atrium, and one in the left heart, such as the left atrium) to calculate systemic blood pressure and systemic volume load. Pressure waveform features, such as peak pressure differentials and timing differences, can serve as bases for the determinations made by the framework 820. Beyond calculating pressure and volume, the framework 820 can be configured / trained to generate output indicating appropriate dosages for a first medication (e.g., ARNI or other medication to manage systemic blood pressure) and a second medication (e.g., Lasix or other medication to manage systemic volume load).

[0091] The machine learning model / system 800 includes a transform / classification framework 820 configured through training to use the left and right pressure data, possibly along with certain other patient- or demographic-associated metrics, to generate medication dosage outputs, which can be displayed via a mobile app, relayed to healthcare practitioners, or otherwise used to cause or direct medication dosing. For example, the framework 820 can advantageously map hemodynamic data to specific dosage values based on particularized training datasets, wherein the framework / model output can be constrained by physician- preprogrammed dosage baselines and / or limits.

[0092] The machine learning framework / model 820 can be trained using population-based training, wherein a relatively large cohort dataset including pressure waveforms, systemic blood pressure (e.g., from arm-cuff), volume status (e.g., weight, fluid intake, urine output, edema / congestion symptoms), and / or medication records may be used to generate correlations between different drug dose combinations given to patients across a spectrum of pressure and volume statuses. The framework / model 820 can be trained to generate both predicted hemodynamic metrics and optimal medication dosages. Additionally or alternatively, patientspecific training data may be used to fine-tune the framework / model 920 for individual patients using a patient’s own data (e.g., pressure waveforms, doses, fluid metrics, symptoms, etc.) to deduce the patient status and / or optimal medication doses from dual sensors readings of the patient’s left and right heart pressure.

[0093] The training process, represented in training workflow 810, can involve both population-based and patient-specific methods to develop, for example, a convolutional neural network (CNN) that generates systemic blood pressure and volume load outputs, as well as optimal dosages for pressure and volume control medications, such as to align with Guideline Directed Medical Therapy (GDMT) parameters.

[0094] The training workflow 810 for training the framework / model 820 can operate using various training inputs derived from a clinical dataset, which may comprise samples from a large cohort of heart failure patients (e.g., for population-based training) and / or an individual patient’s data collected over a period of time (e.g., weeks or other period, for patient-specific training). Such inputs can be structured to enable the framework 820 to learn correlations between pressure waveforms, hemodynamic metrics, and / or medication dosages.

[0095] In some implementations, the training inputs include known left heart pressure waveforms 811 associated with the patient or patient demographic population, which may comprise high-frequency (e.g., 100 Hz or greater) time-series of pressure measurements from the left heart, such as the left atrium. The pressure data may be captured by an implantable sensor (e.g., MEMS transducer), as described in detail herein. The known left heart pressure data 811 can comprise any type of data structure(s), such as an array of floating-point values (e.g., 32-bit), representing continuous pressure waveform data with features like amplitude, peak pressure, and / or phase shifts. The left heart pressure data 811 may be collected from a clinical patient using an implantable sensor or equivalent experimental setup during training data collection. The left heart pressure data 811 may inform systemic blood pressure calculations, as the left heart pressure is generally more indicative of systemic arterial pressure.

[0096] The training inputs can further include right heart pressure waveforms 812, which can comprise synchronized, high-frequency time-series of pressure measurements from the right heart, such as the right atrium. The source of the right heart pressure data 812 can be an implantable or non-invasive sensor, such as a dual-sensor implant device or a wearable central venous pressure sensor. The known right heart pressure data 812 can comprise any type of data structure(s), such as an array of floating-point values, which may advantageously be synchronized with the left heart waveform data 811 via a common timestamp to capture timing differentials (e.g., peak pressure delays). The right heart pressure data 812 can inform systemic volume load conditions, as right heart pressure is generally more indicative of fluid status compared to left heart pressure.

[0097] In addition to pressure waveform data, the training input may comprise certain other parameters, such as volume status metrics 813, which may include parameters indicating patient / subject weight, fluid intake, urine output, symptoms or the like, which may be used to determine / generate circulatory volume load value(s). The known volume status metric data 813 can generally correspond to discrete, measurable data (e.g., weight in lbs or kg, fluid in mL, boolean symptom flags, etc.). Weight, fluid balance, and the presence or absence of certain symptoms can provide indicators of fluid status in heart failure management.

[0098] Further training input for the framework / model 820 can include certain constraint parameters 814, such as physician-preprogrammed / inputted parameters, which may comprise metadata stored in non-transitory computer-readable media representing baseline medication doses, maximum dose limits, and / or expected normal ranges for systemic blood pressure and volume load. Such data 814 may be defined as floats (e.g., for medication baselines, maximum dosages, pressure values, volume values, etc.). These parameters may be set by a physician / technician for each patient or cohort, and may constrain the machine learning framework / model 820 during training to ensure that generated dosages align with Guideline Directed Medical Therapy (GDMT) safety and efficacy standards. The parameters 814 can advantageously guide the model 820 to learn correlations within clinically acceptable boundaries, enhancing the accuracy and safety of dose recommendations derived from dualsensor pressure waveforms and volume status metrics.

[0099] The inputs, individually and / or collectively, may be mapped to known target / label ground truths that represent categories of generated output that the framework / model 820 is trained to produce in response to varying operational inputs. The training outputs can include known systemic blood pressure data 831, which comprises reference measurements of systemicblood pressure of the patient and / or cohort / population, as obtained from an external device like an arm-cuff hemodynamic monitor or other pressure-measurement device. The data 831 can comprise any type of data structures, such as float scalars (e.g., mmHg), which may be labeled with an enumerated category (e.g., high, normal, low) based on predefined thresholds. The pressure data 831 may advantageously be measured during the same time period as the waveform data 811, 812 to serve as ground truth for training. The pressure data 831 is used as a target output to train the machine learning framework / model 820 (e.g., CNN) to predict systemic blood pressure from waveform inputs, enabling correlation with medication dosing, such as vasodilator medication dosing.

[0100] The training ground truth data can further comprise systemic volume load data 832, which may provide a composite metric of an associated patient’s fluid status and may be derived from clinical measurements such as weight, fluid intake, urine output, and / or symptoms (e.g., edema, pulmonary congestion, or other heart-failure-related symptoms). The volume load data 832 may be maintained as a set of data structures (e.g., 32-bit floats or integers) representative of weight, fluid intake, urine output, or the like. In some implementations, the data 823 includes one or more symptom flags, which may comprise boolean data structures (e.g., true / false for edema, pulmonary congestion, etc.). In some implementations, the volume load data 832 can be represented as a composite volume score of the various parameters, aggregated via a weighted formula, and may be labeled as high, normal, or low based on thresholds. The volume load data 832 may be collected from clinical records or patient reports during the same time period as associated pressure waveform data. The known volume load data 832 can be used as a target output to train the framework / model 820 to predict systemic volume load, enabling correlation with diuretic medication (e.g., Lasix) dosing.

[0101] The machine learning framework / model 820 can further be trained to correlate the input(s) with known medication dosages, such as pressure 833 and volume 834 control medication dosages. Dosage of an Angiotensin Receptor-Neprilysin Inhibitor (ARNI) medication is described as an example pressure-management medication herein, wherein such medication can be used to treat heart failure by reducing fluid buildup and relaxing blood vessels, improving the heart’s pumping ability. The pressure medication dosage data 833 can represent dosages administered to a relevant patient at the time that correlates to input data sampling, reflecting the drug’s effect on systemic blood pressure. The ARNI dosage 833 can be maintained in non-volatile storage as any type of data structure(s), such as 32-bit floats, and may be constrained by physician-set baseline and / or maximum doses. The ARNI dosage data833 serves as a target output to train the framework 820 to recommend ARNI doses based on predicted blood pressure, learning optimal dosing correlations.

[0102] In some implementations, known dosage data 834 for a second medication, such as a volume-management medication (e.g., a loop diuretic like Lasix / furosemide), is further used as a ground truth output data type. The volume-management medication dosage data 834 can represent example dosages of medication for treat heart failure by increasing urine production to reduce fluid overload, alleviating symptoms like edema and pulmonary congestion. The volume medication dosage data 834 can represent dosages administered to a relevant patient at the time that correlates to input data sampling, reflecting the drug’s effect on systemic volume load. The volume-management medication dosage 834 can be maintained in non-volatile storage as any type of data structure(s), such as 32-bit floats, and may be constrained by physician-set baseline and / or maximum doses. The volume-management medication dosage data 834 serves as a target output to train the framework 820 to recommend volume- management medication doses based on predicted volume load, learning optimal dosing correlations.

[0103] The operational, real-time use of the machine learning framework 820 for heart failure management and medication dosage systems as described herein involves processing live / real-time inputs through the trained machine learning frame work / model 820 (e.g., convolutional neural network) to generate predictions and recommendations, enabling dynamic GDMT-compliant dosing adjustments. The real-time inputs can include synchronized left heart pressure waveforms 815 (e.g., arrays of 32-bit floats from a left atrium sensor of a dual-sensor implant device or system) and right heart pressure waveforms 816 (e.g., arrays of 32-bit floats from a right atrium sensor of a dual-sensor implant device or system or body-worn pressure sensor). The real-time pressure waveform inputs can optionally be supplemented by volume status metrics 817, which may comprise data structure(s) indicating patient weight, fluid intake, urine output, and / or symptom flag(s) for edema, pulmonary congestion, and / or other symptom(s). Additionally, physician-preprogrammed parameters or other constraint parameters 818 relating to mediation dosage baselines, dosage maximums, and / or pressure / volume thresholds may be input to the model 820.

[0104] The real-time pressure waveform inputs 815, 816 can be preprocessed by the dualsensor system’s signal conditioning circuitry (e.g., analog-to-digital converter, filtering, amplification, etc.), for noise reduction and / or feature extraction (e.g., peak differentials, phase shifts), and transmitted via wireless transmission to a receiver, aggregated in a monitorapplication / device, and uploaded to a remote / cloud monitoring system / server for model inference. In real-time operation, the framework 820 outputs can provide real-time / predicted systemic blood pressure 835, real-time / predicted systemic volume load 836, real-time recommended pressure-management medication dosage 837, and / or real-time recommended volume-management medication dosage 838, with optional alert flags for extreme values triggering physician notifications. In some implementations, the model outputs are transmitted to a mobile device for display as numerical values, graphical trends, or alerts, facilitating dose titration based on real-time hemodynamic data. The machine learning framework’s functionality in generating pressure / volume determinations and deriving medication dosages provides improvements in heart failure management technologies by transforming raw waveform data into actionable medical outputs, reducing manual titration errors, and improving patient outcomes, wherein such machine learning model operation is tied to physical sensors and a networked system, providing a complex technical improvement.Additional Discussion of Examples

[0105] The following are non-exclusive descriptions of possible examples of the present invention.

[0106] In one example of the disclosure, a drug dosage system includes a right heart pressure sensor which provides right heart pressure data and a left heart pressure sensor which provides left heart pressure data. The drug dosage system also includes a receiver which is electronically connected to the right heart pressure sensor and the left heart pressure sensor and receives the right heart pressure data and the left heart pressure data. The drug dosage system also includes a computing device with a processor and memory. The memory includes programming that is executable by the processor to: compute a dosage of a first drug based on the left heart pressure data and right heart pressure data; and compute a dosage of a second drug based on the left heart pressure data and right heart pressure data.

[0107] The drug dosage system of the preceding paragraph can optionally include, additionally and / or alternatively, any one or more of the following features, configurations and / or additional components in the paragraphs below.

[0108] In an example of the drug dosage system, the memory further includes programming executable by the processor to compute systemic blood pressure from the left heart pressure data and the right heart pressure data, wherein the systemic blood pressure depends more heavily on the left heart pressure data than the right heart pressure data.

[0109] In an example of the drug dosage system, computing the dosage of the first drug is based on the systemic blood pressure.

[0110] In an example of the drug dosage system, computing the dosage of the first drug is further based on a systemic blood volume load.

[0111] In an example of the drug dosage system, the memory further includes programming executable by the processor to compute systemic blood volume load from the left heart pressure data and the right heart pressure data, wherein the systemic blood pressure depends more heavily on the right heart pressure data than the left heart pressure data.

[0112] In an example of the drug dosage system, computing the dosage of the second drug is based on the systemic blood volume load.

[0113] In an example of the drug dosage system, computing the dosage of the second drug is further based on the systemic blood pressure.

[0114] In an example of the drug dosage system, the right heart pressure sensor and the left heart pressure sensor are connected to a transmitter which is in wireless communication with the receiver.

[0115] In an example of the drug dosage system, the receiver is in wireless communication with the computing device.

[0116] In an example of the drug dosage system, the computing device is a smart device.

[0117] In an example of the drug dosage system, the receiver is in wireless communication with a smart device which is in wireless communication with the computing device.

[0118] In an example of the drug dosage system, the computing device is a server.

[0119] In an example of the drug dosage system, the right heart pressure sensor is implanted in the right atrium or ventricle of a heart.

[0120] In an example of the drug dosage system, the left heart pressure sensor is implanted in the left atrium or ventricle of a heart.

[0121] In an example of the drug dosage system, the right heart pressure sensor is a non- invasive central venous pressure sensor worn on a neck.

[0122] In an example of the drug dosage system, the left heart pressure sensor is an arterial pressure sensor.

[0123] In another example of the disclosure, a method is disclosed of administering two medications. The method includes providing an initial dosage of a first drug and an initial dose of a second drug. The method further includes measuring a left heart blood pressure of a patient and measuring a right heart blood pressure of the patient. The dosage of the first drug is adjustedbased on the left heart blood pressure and the right heart blood pressure. The dosage of the second drug is adjusted based on the left heart blood pressure and the right heart blood pressure. The method further includes providing an adjusted dosage of the first drug and an adjusted dosage of the second drug.

[0124] The method of the preceding paragraph can optionally include, additionally and / or alternatively, any one or more of the following features, configurations and / or additional components in the paragraphs below.

[0125] In an example of the method, the method further comprises computing a systemic blood pressure of the patient from the left heart pressure data and the right heart pressure data, wherein the systemic blood pressure depends more heavily on the left heart pressure data than the right heart pressure data.

[0126] In an example of the method, computing the systemic blood pressure of the patient from the left heart pressure data and the right heart pressure data is performed using a machine learning algorithm.

[0127] In an example of the method, the machine learning algorithm is trained with data including previous correlations between a plurality of left heart blood pressure and right heart blood pressure and system blood pressure.

[0128] In an example of the method, adjusting the dosage of the first drug is based on the systemic blood pressure of the patient.

[0129] In an example of the method, adjusting the dosage of the first drug is further based on a systemic blood volume load of the patient.

[0130] In an example of the method, adjusting the dosage of the first drug based on the systemic blood pressure comprises: adjusting the first drug dosage lower if the systemic blood pressure is low, keeping the same dosage of the first drug if the systemic blood pressure is normal, or adjusting the first drug dosage higher if the systemic blood pressure is high.

[0131] In an example of the method, the dosage adjustment increment is preselected.

[0132] In an example of the method, the dosage adjustment increment is based on at least one previous dosage adjustment and the systemic blood pressure and / or the system blood volume load.

[0133] In an example of the method, the method further comprises computing a systemic blood volume load of the patient from the left heart pressure data and the right heart pressure data, wherein the systemic blood volume load depends more heavily on the right heart pressure data than the left heart pressure data.

[0134] In an example of the method, computing the systemic blood volume load of the patient from the left heart pressure data and the right heart pressure data is performed using a machine learning algorithm.

[0135] In an example of the method, the machine learning algorithm is trained with data including previous correlations between a plurality of left heart blood pressure and right heart blood pressure and system blood volume load.

[0136] In an example of the method, adjusting the dosage of the second drug is based on the systemic blood volume load of the patient.

[0137] In an example of the method, adjusting the dosage of the second drug is further based on a systemic blood pressure of the patient.

[0138] In an example of the method, adjusting the dosage of the second drug based on the systemic blood volume load comprises: adjusting the second drug dosage lower if the systemic blood volume load is low, keeping the same dosage of the second drug if the systemic blood volume load is normal, or adjusting the second drug dosage higher if the systemic blood volume load is high.

[0139] In an example of the method, the dosage adjustment increment is preselected.

[0140] In an example of the method, the dosage adjustment increment is based on at least one previous dosage adjustment and the systemic blood pressure and / or the system blood volume load.

[0141] In an example of the method, adjusting the dosage of the first drug is performed utilizing a machine learning algorithm.

[0142] In an example of the method, adjusting the dosage of the second drug is performed utilizing a machine learning algorithm.

[0143] In an example of the method, the machine learning algorithm is trained using previous data including systemic blood pressures correlated with usage of dosages of the first drug and systemic blood volume levels correlated with usage of a dosages of the second drug.

[0144] In an example of the method, the machine learning algorithm is trained by data produced by: monitoring a patient for a period of time while recording the systemic blood pressures and systemic blood volume levels while recording medication doses, fluid intake, urine output, arterial pressure, and / or symptoms of disease; and titrating a dosage of the first drug and a dosage of the second drug over a period of time at different levels while continuing the record the systemic blood pressures and systemic blood volume levels.

[0145] In an example of the method, the machine learning algorithm is trained using previous data including systemic blood pressures correlated with usage of dosages of the first drug and systemic blood volume levels correlated with usage of a dosages of the second drug.

[0146] In an example of the method, measuring the right heart pressure comprises measuring the pressure at the right atrium or ventricle of a heart.

[0147] In an example of the method, measuring the left heart pressure comprises measuring the pressure at the left atrium or ventricle of a heart.

[0148] In an example of the method, measuring the right heart pressure comprises utilizing a non-invasive central venous pressure sensor worn on a neck of the patient.

[0149] In an example of the method, measuring the right heart pressure comprises measuring arterial pressure.

[0150] In an example of the method, the method further comprises implanting a left heart blood pressure sensor into the patient via a transcatheter implantation process.

[0151] In an example of the method, the method further comprises implanting a right heart blood pressure sensor into the patient via a transcatheter implantation process.

[0152] In another example of the disclosure, a method is disclosed for administering two medications. The method includes providing a first dosage of a first drug to a patient and providing a second dosage of a second drug to the patient. A left heart blood pressure of the patient is measured by a drug dosage system. A right heart blood pressure of the patient is measured by the drug dosage system. The drug dosage system adjusts the first dosage of the first drug based on input features from the left heart blood pressure and the right heart blood pressure. The drug dosage system adjusts the second dosage of the second drug based on the input features from the left heart blood pressure and the right heart blood pressure. The drug dosage system provides to the patient a first adjusted dosage of the first drug and a second adjusted dosage of the second drug.

[0153] The method of the preceding paragraph can optionally include, additionally and / or alternatively, any one or more of the following features, configurations and / or additional components in the paragraphs below.

[0154] In an example of the method, the method further comprises: training the drug dosage system for determining the input features, wherein the training the drug dosage system comprises: collecting a clinical dataset of samples, wherein each sample of the clinical dataset comprises: a first pressure waveform from a right side of a heart of a clinical patient sensed by a first sensor on a date of the sample; a second pressure waveform from a left side of the heartof the clinical patient sensed by a second sensor on the date of the sample; a first dose level of the first drug administered to the clinical patient on the date of the sample; a second dose level of the second drug administered to the clinical patient on the date of the sample; a system pressure score of the clinical patient relative to a low-pressure threshold and a high-pressure threshold, wherein the system pressure score is based on the first pressure waveform and the second pressure waveform of the sample; and a system volume score of the clinical patient relative to a low-volume threshold and a high- volume threshold, wherein the system volume score is based on the first pressure waveform and the second pressure waveform of the sample; performing waveform analysis of the first pressure waveform and the second pressure waveform in each of the samples to calculate a plurality of signal measures; and determining the input features by computing combinatorial measures between the plurality of signal measures and selecting signal measures from the plurality of signal measures with most predictive combinatorial measures as belonging to the input features.

[0155] In an example of the method, performing waveform analysis of the first pressure waveform and the second pressure waveform in each of the samples to calculate a plurality of signal measures comprises: identifying individual cardiac cycles in both the first pressure waveform and the second pressure waveform of each sample; identifying a dicrotic notch in each of the individual cardiac cycles; identifying a systolic rise phase, a systolic decay phase, and a diastolic phase in each of the individual cardiac cycles; and extracting signal measures from each of the systolic rise phase, the systolic decay phase, and the diastolic phase from each of the individual cardiac cycles.

[0156] In an example of the method, the signal measures correspond to hemodynamic effects from each of the systolic rise phase, the systolic decay phase, and the diastolic phase from each of the individual cardiac cycles, and wherein the hemodynamic effects comprise contractility, aortic compliance, stroke volume, vascular tone, afterload, and full cardiac cycle.

[0157] In an example of the method, the signal measures comprise a mean, a maximum, a minimum, a duration, an area, a standard deviation, derivatives, and / or morphological measures from each of the systolic rise phase, the systolic decay phase, and the diastolic phase from each of the individual cardiac cycles.

[0158] In an example of the method, the signal measures comprise heart rate, respiratory rate, stroke volume, pulse pressure, pulse pressure variation, stroke volume variation, mean arterial pressure (MAP), systolic pressure (SYS), diastolic pressure (DIA), heart ratevariability, cardiac output, peripheral resistance, vascular compliance, and / or left- ventricular contractility extracted from each of the individual cardiac cycles.

[0159] In an example of the method, computing the combinatorial measures between the plurality of signal measures comprises: performing step one by arbitrarily selecting three signal measures from the plurality of signal measures; performing step two by calculating different orders of power for each of the three signal measures to generate powers of the three signal measures; performing step three by multiplying the powers of the three signal measures together to generate the product of the powers of the three signal measures; performing step four by performing receiver operating characteristic (ROC) analysis of the product to arrive at a combinatorial measure for the three signal measures; and repeating steps one, two, three, and four until all of the combinatorial measures have been computed between all of the plurality of signal measures.

[0160] In an example of the method, measuring the right heart blood pressure comprises measuring the pressure at the right atrium or ventricle of a heart of the patient by the drug dosage system.

[0161] In an example of the method, measuring the right heart pressure comprises measuring arterial pressure of the patient by the drug dosage system.

[0162] In an example of the method, measuring the left heart blood pressure comprises measuring the pressure at the left atrium or ventricle of the heart of the patient by the drug dosage system.

[0163] In an example of the method, the method further comprises implanting a left heart blood pressure sensor into the patient via a transcatheter implantation process.

[0164] In an example of the method, the method further comprises implanting a right heart blood pressure sensor into the patient via a transcatheter implantation process.

[0165] In an example of the method, the drug dosage system comprises a non-invasive central venous pressure sensor worn on a neck of the patient and configured to measure the right heart blood pressure.

[0166] In another example of the disclosure, a method is disclosed for administering two medications. The method includes providing a first dosage of a first drug to a patient and providing a second dosage of a second drug to the patient. A left heart blood pressure waveform of the patient is measured by a drug dosage system. A right heart blood pressure waveform of the patient is also measured by the drug dosage system. A processor of the drug dosage system performs waveform analysis of both the left heart blood pressure waveform of the patient andthe right heart blood pressure waveform of the patient to determine input features of the patient. The input features of the patient are inputted into an artificial neural network stored in memory of the drug dosage system to determine a systemic blood pressure of the patient and a systemic blood volume load of the patient. The drug dosage system adjusts the first dosage of the first drug based on the systemic blood pressure of the patient determined by the artificial neural network. The drug dosage system also adjusts the second dosage of the second drug based on the systemic blood pressure of the patient determined by the artificial neural network. The drug dosage system provides to the patient a first adjusted dosage of the first drug and a second adjusted dosage of the second drug.

[0167] The method of the preceding paragraph can optionally include, additionally and / or alternatively, any one or more of the following features, configurations and / or additional components in the paragraphs below.

[0168] In an example of the method, the method further comprises: training the drug dosage system for determining the input features, wherein the training the drug dosage system for determining the input features comprises: collecting a clinical dataset of samples from a plurality of clinical patients, wherein each of the samples of the clinical dataset comprises: a first pressure waveform from a right side of a heart of a clinical patient of the plurality of patients sensed by a first sensor during a time period of the respective sample; and a second pressure waveform from a left side of the heart of the clinical patient sensed by a second sensor during the time period of the respective sample; annotating each of the samples of the clinical dataset with a systemic blood pressure of the clinical patient measured by a third sensor during the time period of the respective sample; annotating each of the samples of the clinical dataset with a systemic blood volume load of the clinical patient based upon a weight, fluid intake, urine output, symptoms of edema, and / or symptoms of pulmonary congestion during the time period of the respective sample; performing waveform analysis of the first pressure waveform and the second pressure waveform in each of the samples to calculate a plurality of signal measures for the respective sample; inputting the plurality of signal measures of each of the samples into an input layer of the artificial neural network; forward propagating the plurality of signal measures of each of the samples through a hidden layer of the artificial neural network to compute at least one output of the respective sample; backward propagating the at least one output through the hidden layer of the artificial neural network to determine a cost function; minimizing, based on the cost function, the plurality of signal measures to signal measures most predictive of both the systemic blood pressure of the clinical patient and the systemicblood volume load of the clinical patient; and selecting the signal measures most predictive of both the systemic blood pressure of the clinical patient and the systemic blood volume load of the clinical patient as input features of the patient to the drug dosage system for determining the systemic blood pressure and the systemic blood volume load of the patient.

[0169] In an example of the method, performing waveform analysis of the first pressure waveform and the second pressure waveform in each of the samples to calculate a plurality of signal measures comprises: identifying individual cardiac cycles in both the first pressure waveform and the second pressure waveform of each of the samples; identifying a dicrotic notch in each of the individual cardiac cycles; identifying a systolic rise phase, a systolic decay phase, and a diastolic phase in each of the individual cardiac cycles; and extracting signal measures from each of the systolic rise phase, the systolic decay phase, and the diastolic phase from each of the individual cardiac cycles.

[0170] In an example of the method, the signal measures correspond to hemodynamic effects from each of the systolic rise phase, the systolic decay phase, and the diastolic phase from each of the individual cardiac cycles, and wherein the hemodynamic effects comprise contractility, aortic compliance, stroke volume, vascular tone, afterload, and full cardiac cycle.

[0171] In an example of the method, the signal measures comprise a mean, a maximum, a minimum, a duration, an area, a standard deviation, derivatives, and / or morphological measures from each of the systolic rise phase, the systolic decay phase, and the diastolic phase from each of the individual cardiac cycles.

[0172] In an example of the method, the signal measures comprise heart rate, respiratory rate, stroke volume, pulse pressure, pulse pressure variation, stroke volume variation, mean arterial pressure (MAP), systolic pressure (SYS), diastolic pressure (DIA), heart rate variability, cardiac output, peripheral resistance, vascular compliance, and / or left- ventricular contractility extracted from each of the individual cardiac cycles.

[0173] In an example of the method, measuring the right heart blood pressure waveform comprises measuring the pressure at the right atrium or ventricle of a heart of the patient by the drug dosage system.

[0174] In an example of the method, measuring the right heart pressure waveform comprises measuring arterial pressure of the patient by the drug dosage system.

[0175] In an example of the method, measuring the left heart blood pressure waveform comprises measuring the pressure at the left atrium or ventricle of the heart of the patient by the drug dosage system.

[0176] In an example of the method, the method further comprises implanting a left heart blood pressure sensor into the patient via a transcatheter implantation process.

[0177] In an example of the method, the method further comprises implanting a right heart blood pressure sensor into the patient via a transcatheter implantation process.

[0178] In an example of the method, the drug dosage system comprises a non-invasive central venous pressure sensor worn on a neck of the patient and configured to measure the right heart blood pressure.

[0179] Provided below is a list of numbered examples, each of which may include aspects of any of the other examples disclosed herein. Furthermore, aspects of any example described above may be implemented in any of the numbered examples provided below.

[0180] Example 1 : A system for managing heart failure in a patient, the system comprising: a first pressure sensor and a second pressure sensor, wherein at least one of the first pressure sensor or the second pressure sensor is configured to be implanted within a heart of a patient, wherein the first pressure sensor and the second pressure sensor are configured to provide right heart pressure data and left heart pressure data of the heart, respectively; one or more transmitters coupled to the first pressure sensor and the second pressure sensor and configured to transmit the right heart pressure data and the left heart pressure data; a receiver configured to receive the right heart pressure data and the left heart pressure data from the one or more transmitters; and control circuitry configured to: receive the right heart pressure data and the left heart pressure data from the receiver; access a machine learning medication dosage management model trained at least on historical left and right heart pressure logs and ground truth medication dosage logs associated with a blood pressure reducing medication and an intravascular volume reducing medication; and use the machine learning medication dosage management model to generate, based at least on the right heart pressure data and the left heart pressure data, real-time dosages of the blood pressure reducing medication and the intravascular volume reducing medication, respectively.

[0181] Example 2: The system of any example herein, in particular example 1, wherein the first pressure sensor and the second pressure sensor are both configured for implantation within the heart of the patient.

[0182] Example 3: The system of any example herein, in particular example 2, wherein: the first pressure sensor and the second pressure sensor are part of a dual-sensor implant; and the dual-sensor implant is configured to span two chambers of the heart when implanted in the heart.

[0183] Example 4: The system of any example herein, in particular example 1, wherein at least one of the first pressure sensor or the second pressure sensor is coupled to a medical implant device configured to alter flow of blood within a circulatory system when implanted.

[0184] Example 5: The system of any example herein, in particular example 1, wherein the left heart pressure data is synchronized with the right heart pressure data.

[0185] Example 6: The system of any example herein, in particular example 1, wherein at least one of the one or more transmitters is a wireless transmitter configured to transmit at least one of the right heart pressure data or the left heart pressure data to the receiver.

[0186] Example 7 : The system of any example herein, in particular example 1, wherein the control circuitry is further configured to generate systemic pressure and systemic volume load values based on the right heart pressure data and the left heart pressure data.

[0187] Example 8: The system of any example herein, in particular example 7, wherein the machine learning medication dosage management model is further trained on historical systemic volume logs and historical systemic pressure logs.

[0188] Example 9: The system of any example herein, in particular example 1, wherein the ground truth medication dosage logs are further associated with cardiac output controlling medication.

[0189] Example 10: The system of any example herein, in particular example 1, wherein one of the first pressure sensor or the second pressure sensor is integrated with a wearable device, the wearable device being configured to be worn on a skin surface of the patient.

[0190] Example 11: The system of any example herein, in particular example 10, wherein the wearable device includes the receiver and the control circuitry.

[0191] Example 12: The system of any example herein, in particular example 10, wherein the wearable device is configured to be worn near a major vein of the patient, such that pressure data of the major vein can be recorded by the wearable device.

[0192] Example 13: The system of any example herein, in particular example 1, wherein the machine learning medication dosage management model is further trained to generate the real-time dosages based at least on volume status metrics that are based on at least one of patient weight data, fluid intake data, urine output data, or symptom flags.

[0193] Example 14: The system of any example herein, in particular example 1, wherein: the control circuitry is further configured to access constraint parameters stored in a non- transitory computer-readable medium, the constraint parameters comprising at least one of baseline dosages for the blood pressure reducing medication and the intravascular volumereducing medication, maximum dosages for the blood pressure reducing medication and the intravascular volume reducing medication, or normal ranges for systemic blood pressure and systemic volume load; and the real-time dosages of the blood pressure reducing medication and the intravascular volume reducing medication are constrained by the constraint parameters.

[0194] Example 15: A system for managing heart failure in a patient, the system comprising: a right heart pressure sensor and a left heart pressure sensor, wherein at least one of the right heart pressure sensor or the left heart pressure sensor is configured to be implanted within a heart of a patient, wherein the right heart pressure sensor and the left heart pressure sensor are configured to provide right heart pressure data and left heart pressure data of the heart, respectively; one or more transmitters coupled to the right heart pressure sensor and the left heart pressure sensor and configured to transmit the right heart pressure data and the left heart pressure data; and a local computing system including control circuitry configured to: receive the right heart pressure data and the left heart pressure data from the one or more transmitters; and transmit the right heart pressure data and the left heart pressure data to a remote computing system over a network; wherein the remote computing system includes control circuitry configured to: access a machine learning medication dosage management model trained at least on historical left and right heart pressure logs and ground truth medication dosage logs associated with a blood pressure reducing medication and an intravascular volume reducing medication; receive the right heart pressure data and the left heart pressure data from the local computing system over the network; use the machine learning medication dosage management model to generate, based at least on the right heart pressure data and the left heart pressure data, real-time dosages of the blood pressure reducing medication and the intravascular volume reducing medication, respectively; and transmit the real-time dosages to the local computing system over the network. The control circuitry of the local computing system is further configured to: receive the real-time dosages; and render a user interface on a graphical display of the local computing system that indicates the real-time dosages.

[0195] Example 16: The system of any example herein, in particular example 15, wherein the local computing system includes: a receiver device configured to wirelessly couple to the one or more transmitters and receive the right heart pressure data and the left heart pressure data therefrom; and a mobile computing device that is separate from the receiver device includes the graphical display and is configured to wirelessly couple to the receiver device and receive the right heart pressure data and the left heart pressure data from the receiver device.

[0196] Example 17 : The system of any example herein, in particular example 15, wherein: the remote computing system is further configured to: generate systemic volume and pressure data using the machine learning medication dosage management model; and transmit the systemic volume and pressure data to the local computing system over the network; and the user interface rendered by the local computing system further indicates the systemic volume and pressure data.

[0197] Example 18: The system of any example herein, in particular example 15, wherein: the left heart pressure sensor is configured to be implanted in a left atrium of the heart; and the right heart pressure sensor is a non-invasive central venous pressure sensor configured to be worn on a body of the patient.

[0198] Example 19: A system for managing heart failure in a patient, comprising: a dualsensor implant adapted to be positioned within a heart of a patient, the dual-sensor implant comprising: a right heart pressure sensor positioned to measure a right heart pressure waveform when the dual-sensor implant is implanted within the heart; a left heart pressure sensor positioned to measure a left heart pressure waveform when the dual-sensor implant is implanted within the heart, the left heart pressure waveform being synchronized with the right heart pressure waveform; signal conditioning circuitry configured to generate digitized right heart pressure data and left heart pressure data based on the right heart pressure waveform and the left heart pressure waveform, respectively; and a wireless transmitter communicatively coupled to the signal conditioning circuitry and configured to wirelessly transmit the right heart pressure data and the left heart pressure data; a receiver configured to wirelessly receive the right heart pressure data and the left heart pressure data from the wireless transmitter; and monitor control circuitry configured to: communicatively coupled to the receiver; access the right heart pressure data and the left heart pressure data; access a machine learning medication dosage management model trained on historical left and right heart pressure logs and ground truth medication dosage logs associated with a blood pressure reducing medication and an intravascular volume reducing medication; and use the machine learning medication dosage management model to generate, based on the right heart pressure data and the left heart pressure data, real-time dosages of the blood pressure reducing medication and the intravascular volume reducing medication, respectively.

[0199] Example 20: The system of any example herein, in particular example 19, wherein the dual-sensor implant is configured to be implantable in an interatrial septum wall with theright heart pressure sensor exposed in a right atrium and the left heart pressure sensor exposed in a left atrium.DOCTRINE OF EQUIVALENTS

[0200] While the above description contains many specific implementations of the invention, these should not be construed as limitations on the scope of the disclosure, but rather as an example of one implementation thereof. It is therefore to be understood that the present disclosure may be practiced in ways other than specifically described, without departing from the scope and spirit of the present disclosure. Thus, implementations of the present disclosure should be considered in all respects as illustrative and not restrictive. Accordingly, the scope of the disclosure should be determined not by the implementations illustrated, but by the appended claims and their equivalents.

Claims

WHAT IS CLAIMED IS:

1. A system for managing heart failure in a patient, the system comprising: a first pressure sensor and a second pressure sensor, wherein at least one of the first pressure sensor or the second pressure sensor is configured to be implanted within a heart of a patient, wherein the first pressure sensor and the second pressure sensor are configured to provide right heart pressure data and left heart pressure data of the heart, respectively; one or more transmitters coupled to the first pressure sensor and the second pressure sensor and configured to transmit the right heart pressure data and the left heart pressure data; a receiver configured to receive the right heart pressure data and the left heart pressure data from the one or more transmitters; and control circuitry configured to: receive the right heart pressure data and the left heart pressure data from the receiver; access a machine learning medication dosage management model trained at least on historical left and right heart pressure logs and ground truth medication dosage logs associated with a blood pressure reducing medication and an intravascular volume reducing medication; and use the machine learning medication dosage management model to generate, based at least on the right heart pressure data and the left heart pressure data, real-time dosages of the blood pressure reducing medication and the intravascular volume reducing medication, respectively.

2. The system of claim 1, wherein the first pressure sensor and the second pressure sensor are both configured for implantation within the heart of the patient.

3. The system of claim 2, wherein: the first pressure sensor and the second pressure sensor are part of a dualsensor implant; and the dual-sensor implant is configured to span two chambers of the heart when implanted in the heart.

4. The system of claim 1, wherein at least one of the first pressure sensor or the second pressure sensor is coupled to a medical implant device configured to alter flow of blood within a circulatory system when implanted.

5. The system of claim 1, wherein the left heart pressure data is synchronized with the right heart pressure data.

6. The system of claim 1, wherein at least one of the one or more transmitters is a wireless transmitter configured to transmit at least one of the right heart pressure data or the left heart pressure data to the receiver.

7. The system of claim 1, wherein the control circuitry is further configured to generate systemic pressure and systemic volume load values based on the right heart pressure data and the left heart pressure data.

8. The system of claim 7, wherein the machine learning medication dosage management model is further trained on historical systemic volume logs and historical systemic pressure logs.

9. The system of claim 1, wherein the ground truth medication dosage logs are further associated with cardiac output controlling medication.

10. The system of claim 1, wherein one of the first pressure sensor or the second pressure sensor is integrated with a wearable device, the wearable device being configured to be worn on a skin surface of the patient.

11. The system of claim 10, wherein the wearable device includes the receiver and the control circuitry.

12. The system of claim 10, wherein the wearable device is configured to be worn near a major vein of the patient, such that pressure data of the major vein can be recorded by the wearable device.

13. The system of claim 1, wherein the machine learning medication dosage management model is further trained to generate the real-time dosages based at least on volume status metrics that are based on at least one of patient weight data, fluid intake data, urine output data, or symptom flags.

14. The system of claim 1, wherein: the control circuitry is further configured to access constraint parameters stored in a non-transitory computer-readable medium, the constraint parameters comprising at least one of baseline dosages for the blood pressure reducing medication and the intravascular volume reducing medication, maximum dosages for the blood pressure reducing medication and the intravascular volume reducing medication, or normal ranges for systemic blood pressure and systemic volume load; and the real-time dosages of the blood pressure reducing medication and the intravascular volume reducing medication are constrained by the constraint parameters.

15. A system for managing heart failure in a patient, the system comprising: a right heart pressure sensor and a left heart pressure sensor, wherein at least one of the right heart pressure sensor or the left heart pressure sensor is configured to be implanted within a heart of a patient, wherein the right heart pressure sensor and the left heart pressure sensor are configured to provide right heart pressure data and left heart pressure data of the heart, respectively; one or more transmitters coupled to the right heart pressure sensor and the left heart pressure sensor and configured to transmit the right heart pressure data and the left heart pressure data; and a local computing system including control circuitry configured to: receive the right heart pressure data and the left heart pressure data from the one or more transmitters; and transmit the right heart pressure data and the left heart pressure data to a remote computing system over a network; wherein the remote computing system includes control circuitry configured to: access a machine learning medication dosage management model trained at least on historical left and right heart pressure logs and ground truth medication dosage logs associated with a blood pressure reducing medication and an intravascular volume reducing medication; receive the right heart pressure data and the left heart pressure data from the local computing system over the network;use the machine learning medication dosage management model to generate, based at least on the right heart pressure data and the left heart pressure data, real-time dosages of the blood pressure reducing medication and the intravascular volume reducing medication, respectively; and transmit the real-time dosages to the local computing system over the network; and wherein the control circuitry of the local computing system is further configured to: receive the real-time dosages; and render a user interface on a graphical display of the local computing system that indicates the real-time dosages.

16. The system of claim 15, wherein the local computing system includes: a receiver device configured to wirelessly couple to the one or more transmitters and receive the right heart pressure data and the left heart pressure data therefrom; and a mobile computing device that is separate from the receiver device includes the graphical display and is configured to wirelessly couple to the receiver device and receive the right heart pressure data and the left heart pressure data from the receiver device.

17. The system of claim 15, wherein: the remote computing system is further configured to: generate systemic volume and pressure data using the machine learning medication dosage management model; and transmit the systemic volume and pressure data to the local computing system over the network; and the user interface rendered by the local computing system further indicates the systemic volume and pressure data.

18. The system of claim 15, wherein: the left heart pressure sensor is configured to be implanted in a left atrium of the heart; and the right heart pressure sensor is a non-invasive central venous pressure sensor configured to be worn on a body of the patient.

19. A system for managing heart failure in a patient, comprising: a dual-sensor implant adapted to be positioned within a heart of a patient, the dual-sensor implant comprising: a right heart pressure sensor positioned to measure a right heart pressure waveform when the dual-sensor implant is implanted within the heart; a left heart pressure sensor positioned to measure a left heart pressure waveform when the dual-sensor implant is implanted within the heart, the left heart pressure waveform being synchronized with the right heart pressure waveform; signal conditioning circuitry configured to generate digitized right heart pressure data and left heart pressure data based on the right heart pressure waveform and the left heart pressure waveform, respectively; and a wireless transmitter communicatively coupled to the signal conditioning circuitry and configured to wirelessly transmit the right heart pressure data and the left heart pressure data; a receiver configured to wirelessly receive the right heart pressure data and the left heart pressure data from the wireless transmitter; and monitor control circuitry configured to: communicatively coupled to the receiver; access the right heart pressure data and the left heart pressure data; access a machine learning medication dosage management model trained on historical left and right heart pressure logs and ground truth medication dosage logs associated with a blood pressure reducing medication and an intravascular volume reducing medication; and use the machine learning medication dosage management model to generate, based on the right heart pressure data and the left heart pressure data, real-time dosages of the blood pressure reducing medication and the intravascular volume reducing medication, respectively.

20. The system of claim 19, wherein the dual-sensor implant is configured to be implantable in an interatrial septum wall with the right heart pressure sensor exposed in a right atrium and the left heart pressure sensor exposed in a left atrium.

Images