Methods of determining a heart-failure congestion index for improved speed and accuracy of heart-failure diagnostic and treatment decisions, and related devices and systems

Abstract

Devices, systems, and methods for monitoring of heart failure patient fluid volume with improved accuracy and reduced lag time to identify changes in volume status using one or more of passive, implantable sensors and non-invasive receiving equipment that measure the temporal variation in IVC area as a means to remotely assess and manage heart failure disease status and support self-management optionally employing a new, normalized index of heart failure (HF) congestion. The new normalized HF congestion index further contributes to improved speed and accuracy of heart failure diagnostic and treatment decisions and thus can help to improve patient outcomes and reduce hospitalization occurrences and costs.

Description

METHODS OF DETERMINING A HEART-FAILURE CONGESTION INDEX FOR IMPROVED SPEED AND ACCURACY OF HEART-FAILURE DIAGNOSTIC AND TREATMENT DECISIONS, AND RELATED DEVICES AND SYSTEMS RELATED APPLICATION DATA

[0001] This application claims the benefit of priority of U.S. Provisional Patent Application SerialNo. 63 / 694,460, filed September 13, 2024, and titled “Heart Failure Diagnostic Tools and Methods Using Signal Trace Analysis to Determine Normalized Congestion Score”; U.S. Provisional Patent Application Serial No. 63 / 694,522, filed September 13, 2024, and titled “Heart Failure Diagnostic Tools and Methods using Hazard Analysis”; and U.S. Provisional Patent Application Serial No. 63 / 784,528, filed April 7, 2025, and titled “Heart Failure Diagnostic Tools and Methods Using Signal Trace Analysis”, each of which is incorporated by reference herein in its entirety. FIELD OF THE DISCLOSURE

[0002] The present disclosure relates to heart failure diagnostics and, more specifically, to toolsand methods to facilitate use of data produced by heart-failure-related sensors to improve patient outcomes. BACKGROUND

[0003] Hemodynamic congestion is generally measured using catheter-based filling pressure ofright atrial pressure (RAP) and pulmonary capillary wedge pressure (PCWP). However, due to the small ranges of these pressures relative to the potential measurement error, in the range of 2 mmHg to 6 mmHg (RAP), and 4 mmHg to 12 mmHg (PCWP), it is challenging to capture relevant pressure changes using common pressure-sensing detectors such as a Swan-Ganz catheter or other implantable pressure sensors, which generally may have accuracy limitations in the range of about ±5 mmHg.

[0004] The present Applicant has previously developed and disclosed a number of differentsensors for determining patient fluid status based on direct measurement of a vascular dimension, which indicates geometry, such as diameter or cross-sectional area, and distension or collapse of the vessel. This measurement of vessels, particularly of the inferior vena cava (IVC), may relate more directly to a patient’s circulating blood volume and congestion status. Such measurements using Applicant’s disclosed sensors can be used to estimate a patient’s circulating blood volume and congestion status. In particular, these sensors can be used to determine whether circulating blood volume is too high or too low, whether circulating blood volume is increasing or decreasing, and potentially what treatment should be prescribed, such as diuretics or vaso-dilators. 1 Attorney Docket No.15653-022WOU2

[0005] New devices developed and disclosed by the present Applicant include external ultrasounddevices as well as implantable sensors capable of long-term placement suitable for monitoring patients with chronic conditions. Examples of such implantable, wireless sensors and external ultrasound devices are disclosed, for example, in U.S. Patent Application No.15 / 549,042, filed August 4, 2017 (U.S. Patent No. 10,905,393, granted February 2, 2021), and entitled “Implantable Devices and Related Methods for Heart Failure Monitoring” and U.S. Patent Application No. 16 / 177,183, filed October 31, 2018 (U.S. Patent No. 10,806,352, granted October 20, 2020) and entitled “Wireless Vascular Monitoring Implants,” each of which is incorporated herein in its entirety. In other clinical situations, such as shorter term acute condition monitoring and in-hospital treatments, vascular- dimension sensors for direct fluid state determination and monitoring may be catheter-based. Examples of such catheter-based sensors are disclosed, for example, in U.S. Patent Application No. 15 / 750,100, filed February 2, 2018 (U.S. Patent No.11,039,813, granted June 22, 2021) and entitled “Devices and Methods for Measurement of Vena Cava Dimensions, Pressure and Oxygen Saturation,” which is incorporated herein in its entirety.

[0006] The present Applicant has also developed and disclosed novel diagnostic and treatmentsystems and methods based on the use of the aforementioned sensor devices, some of which are disclosed in the above-mentioned patent applications, and further of which are disclosed in U.S. Patent No. 11,564,596, granted January 31, 2023, entitled “Systems and Methods for Patient Fluid Management”; U.S. Patent Pub. No. US 2021 / 0244381 A1, published August 12, 2021, entitled “Patient Fluid Management Systems and Methods Employing Integrated Fluid Status Sensing”; and International Patent Pub. No. WO 2023 / 170632, published September 23, 2023, entitled “Heart Failure Diagnostic Tools and Methods Using Signal Trace Analysis”, each of which is incorporated by reference herein in its entirety.

[0007] Notwithstanding the foregoing advances, challenges remain with respect to clinicalinterpretation of the signals produced by various vascular area sensing technologies and, in some cases, novel datasets produced thereby.

[0008] Among the challenges is calibration of measurements for vessel volume-based parametersarising from intra-individual and inter-individual variations due to heterogeneity of the vessel shape in order to generate a comparable quantitative output metric indicative of RAP or congestion status. For example, the area of the IVC depends on the location observed along the IVC and on the individual, making it difficult to draw comparisons of absolute area measures between patients as a 2 Attorney Docket No.15653-022WOU2group, which may inhibit clinical use of the data generated. While generally available data normalization algorithms might be applied to address this limitation, when arbitrarily normalizing any feature the normal range will very likely be enclosed by thresholds of unusual and non-intuitive value, i.e., the normalized variable will be in units (possibly dimensionless), with associated thresholds, that are unfamiliar to the clinician and not a part of standard practice and guidelines. Such unusual numbering is a challenge for physicians, and patients, especially as methods evolve and diagnostic algorithms are refined with higher degrees of complexity and the patients increasingly become relevant consumers of the data.

[0009] General clinical acceptance of new data types can also present challenges. Long-standingand well-known physiological parameters have their well-understood ranges (such as blood pressure, body temperature, respiration rate, etc.). These are developed over time, become domain specific knowledge, and become included in clinical guidelines, etc. To help facilitate acceptance and encourage use of new, advantageous systems and data sets, there is thus a need to develop correlations to known physiological parameters for the new signals and datasets to facilitate use by clinicians. Moreover, because the datasets generated from Applicant’s disclosed sensors are so large (far too large for a human mind to manage effectively or efficiently), processing those datasets is a challenge. There is also a need for faster response times for observing changes in fluid volume status, as a challenge presented by heart failure patients is bringing down excess fluid volume as quickly as possible. The challenges with reducing response times can be exacerbated by large datasets that are difficult to process quickly.

[0010] With multiple discrete inputs to diagnostic and treatment algorithms, it is challenging toset actionable thresholds without the instructions becoming overly complex. There is thus a need to reduce the data review burden on the user (whether patient or healthcare provider) as new datasets and parameters are introduced. Challenges facing the user include definition of action thresholds and also determining which drug from their arsenal to deploy and at what dosage because they often do not have solid inputs to inform drug titration decisions, nor good methods to monitor patient drug adherence, nor good measures of individual responses to different doses of drugs.

[0011] Various embodiments disclosed herein address these challenges and present solutionsdesigned to improve speed and accuracy of heart failure diagnostic and treatment decisions and thus improve patient outcomes and reduce hospitalization occurrences and costs. Improving patient outcomes is critical, as heart failure (HF) is among the most costly and frequent causes of 3 Attorney Docket No.15653-022WOU2hospitalization worldwide, estimated to cost $50 billion a year in the United States alone. Diagnosis is associated with increased mortality and morbidity with sufferers reporting very poor quality of life and negative impact on their families and carers. Frequent readmission to hospital is a recurrent feature of HF with increased hospitalization linked to incident decline. Despite progress in recent years with improved monitoring and medical therapy, rates of hospitalization remain stubbornly high with one-year readmission rates topping 55% in the USA. SUMMARY

[0012] In one implementation, the present disclosure is directed to a method for determining aheart-failure congestion index. The method includes receiving a first vessel area trace containing at least two characteristic features representing vessel area parameters for a patient; identifying the characteristic features representing at least a first maximum area (Amax), a minimum area (Amin), and a second maximum area; setting an upper area boundary (UB) as the second maximum area; setting a lower area boundary (LB) as an initial predetermined value; generating a normalized IVC area (^^^^^^^^^^) based on the upper area boundary (UB), the lower area boundary (LB), and the minimum area (Amin); generating an adjusted collapsibility index (^^^^^^^^^^) based on the first maximum area (Amax), the minimum area (Amin), and the lower area(LB); and generating a normalized congestion index for the patient based on the normalized IVC area (^^^^^^^^^^) and the adjusted collapsibility index (^^^^^^^^^^).

[0013] In another implementation, the present disclosure is directed to a heart failure diagnosticsystem. The system includes a trace feature detector configured to identify characteristic features within wirelessly received periodic vessel area traces representing changes in fluid state of a patient over time, wherein the identified characteristic features include features representative of maximum and minimum vessel areas; a metrics generator configured to generate heart function-related parameters for each area trace based on the identified characteristic features, the heart function related parameters including a maximum IVC area corresponding to a first identified maximum vessel area (Amax), and a minimum IVC area corresponding to an identified minimum vessel area (Amin); a boundary generator configured to generate a lower area boundary (LB) for the patient andan upper area boundary (UB) for the patient, wherein the lower area boundary (LB) is set to an initialpredetermined value and the upper area boundary (UB) is set to an identified absolute maximum vessel area for the patient, wherein the metrics generator further comprises a normalized IVC area module configured to set a normalized IVC area (^^^^^^^^^^) based on the upper area boundary (UB), 4 Attorney Docket No.15653-022WOU2the lower area boundary (LB), and the identified minimum vessel area (Amin), and wherein the metrics generator further comprises an adjusted collapsibility index (^^^^^^^^^^) module configured to set an adjusted collapsibility index (^^^^^^^^^^) based on the lower area boundary (LB), the first identified maximum vessel area (Amax), and the identified minimum vessel area (Amin); and an index generator configured to set a normalized congestion index for the patient based on the normalized IVC area (^^^^^^^^^^) and the adjusted collapsibility index (^^^^^^^^^^).

[0014] In yet another implementation, the present disclosure is directed to a heart failurediagnostic system. The system includes a trace feature detector configured to identify characteristic features within wirelessly received periodic vessel area traces representing changes in fluid state of a patient over time, wherein identified characteristic features comprise maneuver types and maximum and minimum areas associated with identified maneuvers; a metrics generator to generate heart function-related parameters for each area trace based on the identified characteristic features, wherein the metrics generator comprises a maximum IVC area module to set a maximum vessel area (Amax) as a maximum area associated with a first maneuver for each area trace, and a minimum IVC area module to set a minimum vessel area (Amin) as a minimum area associated with the first maneuver for each area trace; a boundary generator to generate a lower area boundary (LB) for thepatient and an upper area boundary (UB) for the patient, wherein the upper area boundary (UB) is setto a maximum area associated with a second maneuver out of all area traces for the patient, wherein the metrics generator further comprises a normalized IVC area module to set a normalized IVC area (^^^^^^^^^^) based on the upper area boundary (UB), the lower area boundary (LB), and the minimum vessel area (Amin), and wherein the metrics generator further comprises an adjusted collapsibility index (^^^^^^^^^^) module to set an adjusted collapsibility index (^^^^^^^^^^) based on the lower area boundary (LB), the maximum vessel area (Amax), and the minimum vessel area (Amin); and an index generator to set a normalized congestion index for the patient based on the normalized IVC area (^^^^^^^^^^) and the adjusted collapsibility index (^^^^^^^^^^). BRIEF DESCRIPTION OF THE DRAWINGS

[0015] For the purpose of illustrating the disclosure, the drawings show aspects of one or moreembodiments of the disclosure. However, it should be understood that the present disclosure is not limited to the precise arrangements and instrumentalities shown in the drawings, wherein:FIG. 1 is a high-level schematic depiction of systems encompassed by the present disclosure;5 Attorney Docket No.15653-022WOU2FIG. 2 is an example of an IVC area trace, including a detailed view of a trace portion with specifictrace features identified;FIG. 3 is a block diagram of an embodiment of an overall system according to the present disclosure;FIG. 4 is a block diagram of an embodiment of a trace generator according to the present disclosure;FIG. 5 is a block diagram of an embodiment of a feature detector according to the present disclosure;FIG. 6 is a series of area trace plots showing area response to different patient maneuvers;FIG.6A is an enlarged, detailed view of the area trace for a supine breath-hold maneuver as shown in FIG.6;FIG. 7 is a block diagram of an embodiment of a metrics generator according to the present disclosure;FIG. 7A is an area trace illustrating another component of the metrics generator in an alternative embodiment for determination of area trace features; FIG.7B is an area trace illustrating components of the metrics generator in alternative embodiments for determination of collapsibility index; FIG. 7C is a block diagram of an alternative embodiment of a metrics generator according to the present disclosure;FIG. 8 is a block diagram of an embodiment of a boundary generator according to the presentdisclosure; FIG. 8A is a block diagram of an alternative embodiment of a boundary generator according to the present disclosure;FIG. 9 shows area trace plots for normal respiration and an inspiration breath-hold maneuver, andillustrates an example of identifying an upper boundary from using a maneuver;FIG. 10 is a block diagram of an embodiment of an index generator according to the presentdisclosure;FIG. 11 is an example pressure-volume curve of the IVC;FIG. 12 is an example of a vessel area trace from a daily reading from a sensor according to the presentdisclosure, including both a supine reading and a breath-hold reading; 6 Attorney Docket No.15653-022WOU2FIG. 13 shows normalized IVC area (^^^^^^^^^^) plotted against measured collapsibility index (^^^^^^) foreach of the 61 patients in a First in Human study (FIH) and Early Feasibility Study (EFS) conducted using systems, devices, and methods according to aspects of the present disclosure; FIG.14A shows data for each of the 63 patients from FIG.13 combined into one plot of normalized IVC area (^^^^^^^^^^) plotted against measured collapsibility index (^^^^^^); FIG.14B shows adjusted collapsibility index (^^^^^^^^^^) plotted against normalized IVC area (^^^^^^^^^^) for each of the 61 patients in the FIH and EFS studies combined into one plot, wherein adjusted collapsibility index (^^^^^^^^^^) is represented by the mean of the measured collapsibility index (^^^^^^) and estimated collapsibility index (^^^^ ) to account for breathing style and effort;FIG. 15 is a block diagram of an embodiment of a decision logic according to the present disclosure;FIG. 16 is a block diagram showing embodiments of interface devices according to the presentdisclosure;FIG. 17 is a flow diagram illustrating an embodiment of an overall system flow according to thepresent disclosure;FIG. 18 is a flow diagram illustrating another embodiment of an overall system flow according to thepresent disclosure, which was used in the FIH and EFS studies;FIG. 19 is a schematic diagram illustrating an example approach to determining a number ofconsecutive days above a threshold for use in determining rules for alerts based on consecutive days above a threshold according to the present disclosure; FIG.20A is a graph of receiver operating characteristic (ROC) curves obtained from event-detection accuracy and the mean false positive rate (FPR) showing that the Congestion Index was a significant predictor of adjudicated heart failure events (HFEs) in the FIH and EFS studies; FIG.20B is a graph of ROC curves obtained from event detection accuracy and the unexplained alert rate (UAR) showing that the Congestion Index was a significant predictor of adjudicated HFEs in the FIH and EFS studies; FIG.21A is a graph showing how the Congestion Index of patients in the FIH and EFS studies varies in relation to daily normalized IVC area (^^^^^^^^^^) and measured collapsibility index (^^^^^^), where HFEs and deaths are overlaid on top of the Congestion Index values; 7 Attorney Docket No.15653-022WOU2FIG. 21B is a graph showing the same data as FIG. 21A, plotted in a different orientation against adjusted collapsibility index (^^^^^^^^^^) and without deaths included; FIG.22A is a boxplot comparing the Congestion Index of patients in the FIH and EFS studies in a 7- day event window against the Congestion Index in non-event windows of 7 days on either side of event windows as well as against the Congestion Index of patients who have not had an HFE; FIG.22B is a boxplot comparing the Congestion Index of patients in the FIH and EFS studies in a 90- day event window against the Congestion Index in non-event windows of 90 days on either side of event windows as well as against the Congestion Index of patients who have not had an HFE; FIG.22C is a graph showing the variation in sensitivity and UAR for HFE detection across a varying number of days prior to the HFE, using a 30-day data window and an event detection threshold of 70%; FIG. 22D is a graph showing the relationship between a data window, a prediction horizon, and an HFE; FIG.23 is a graph showing variation of UAR and event detection accuracy with decision threshold for patients in the FIH and EFS studies; FIG. 24 is a graph of ROC curves obtained from event detection accuracy and UAR in the FIH and EFS studies; FIG.25A is an ensemble analysis showing Congestion Index for all patients in the FIH and EFS studies who experienced an HFE in a 7-day event window compared against a random sampling of data from outside the 7-day event window as well as against patients who have not experienced an HFE; FIG.25B is an ensemble analysis showing Congestion Index for all patients in the FIH and EFS studies who experienced an HFE in a 90-day event window compared against a random sampling of data from outside the 90-day event window; FIG.26A is a boxplot showing comparisons of the Congestion Index in patients from the FIH and EFS studies 7 days before death (during a 7-day event window) to the Congestion Index prior to this 7-day event window as well as to the Congestion Index of patients who were still alive; FIG.26B is a boxplot showing comparisons of the Congestion Index in patients from the FIH and EFS studies 90 days before death (during a 90-day event window) to the Congestion Index prior to this 90-day event window as well as to the Congestion Index of patients who were still alive; 8 Attorney Docket No.15653-022WOU2FIG. 27A is an ensemble analysis of the Congestion Index prior to death, showing data for patients with mortality events in the FIH and EFS studies in the 7-day period prior to death compared against a random sampling of data from outside this 7-day period as well as against patients who were still alive; FIG. 27B is an ensemble analysis of the Congestion Index prior to death, showing data for patients with mortality events in the FIH and EFS studies in the 90-day period prior to death compared against a random sampling of data from outside this 90-day period; FIG.28 is a boxplot and histogram illustrating the distribution of Congestion Index scores across the 43 patients in the FIH and EFS studies who successfully completed at least 180 home readings; FIG.29 is a graph illustrating the association between RAP values at implant and the Congestion Index at implant for patients in the FIH and EFS studies; FIG. 30A is a boxplot of Congestion Index between high (RAP≥10mmHg) and not-high (RAP<10 mmHg) groups in the FIH and EFS studies; FIG. 30B is a boxplot of Congestion Index between very-high (RAP≥ 15mmHg) and not-high (RAP<10 mmHg) groups in the FIH and EFS studies; FIG. 31 is a graph with an ROC curve used to assess the ability of the Congestion Index to classify high (RAP>10mmHg) / not-high (RAP<10 mmHg) RAP for patients in the FIH and EFS studies, and an ROC curve used to assess the ability of the Congestion Index to classify very-high (RAP>15mmHg) / not very-high (RAP<15 mmHg) RAP for patients in the FIH and EFS studies; FIG. 32 is a scatterplot showing variation of Congestion Index with log valued NT-proBNP across clinic visits for patients in the FIH and EFS studies; FIG.32A is a bar chart representation of the data in FIG.32 bucketed; FIG.33A is a bar graph showing the mean Congestion Index for patients from the FIH and EFS studies who have high N-terminal pro-B-type natriuretic peptide (NT-proBNP) and low NT-proBNP across clinic visits; FIG.33B is a bar graph showing the mean Congestion Index for patients from the FIH and EFS studies who have stable NT-proBNP and reduced NT-proBNP across clinic visits; FIG. 34 is a graph showing variation in Congestion Index with NT-proBNP, and in particular with high NT-proBNP (greater than or equal to 1000 ng / L) and low NT-proBNP (less than 1000 ng / L) for patients in the FIH and EFS studies; 9 Attorney Docket No.15653-022WOU2FIG. 35 is a graph of ROC curves for classifying high / low NT-proBNP for patients in the FIH and EFS studies using the Congestion Index; FIG. 36 is a bar graph showing mean Congestion Index per New York Heart Association (NYHA) class for patients in the FIH and EFS studies across clinic visits; FIG.37 is a boxplot showing change in Congestion Index with change in NYHA class for patients in the FIH and EFS studies; FIG.38 is an ensemble analysis showing how the Congestion Index behaves relative to an increase of oral diuretics for patients in the FIH and EFS studies; FIG.39 is a boxplot comparing slopes from FIG.38 before and after the increase of oral diuretics;FIG. 40 is a block diagram illustrating an embodiment of a component of the metrics generator forextraction of respiration rate from an area trace; FIG. 40A includes an example of a spectrogram generated from an area trace illustrating another component of the metrics generator in an alternative embodiment for extraction of heart rate and respiration rate from an area trace;FIG. 41 illustrates an example of derivation of IVC axes based on area and collapse parametersaccording to an embodiment disclosed herein;FIG. 42 illustrates an example of estimation of venous return or cardiac output based on change invessel areas according to a further embodiment disclosed herein;FIG. 43 shows examples of signal traces used in determination of cardiac output according toembodiments disclosed herein, wherein trace (a) is a measured area trace, trace (b) is a filtered respiration component trace, and trace (c) is a filtered cardiac component trace;FIG. 44 is a flow diagram illustrating another alternative embodiment for cardiac output estimationbased on sensor data;FIG. 45 shows further examples of signal traces used in determination of cardiac output according toembodiments disclosed herein, wherein trace (a) is a measured area trace, trace (b) is a simulated respiration component trace, and trace (c) is a simulated cardiac component trace; FIGS.46A-D are plots of different HF metrics over time, illustrating an example of use of systemic vascular resistance based on sensor data and blood pressure data showing trends as a function of days for a heart failure patient experiencing symptoms according to an embodiment disclosed 10 Attorney Docket No.15653-022WOU2herein, wherein FIG.46A shows systemic vascular resistance (SVR), FIG.46B shows IVC area mean, FIG.46C shows respiration rate, and FIG.46D shows daily diuretic dosing; and FIG.47 is a block diagram illustrating components of an exemplary computing device. DETAILED DESCRIPTION

[0016] Unless otherwise defined, all terms of art, notations, and other scientific terminology usedherein are intended to have the meanings commonly understood by those of skill in the art. In some cases, terms with commonly understood meanings are defined herein for clarity and / or for ready reference (such as in the below glossary), and the inclusion of such definitions herein should not necessarily be construed to represent a difference over what is generally understood in the art. As used herein, the following terms have the following meanings: Glossary Term Definition AmaxMaximum inferior vena cava area calculated across a Congestion Index readingAminMinimum inferior vena cava area Amean Mean inferior vena cava area ^^^^^^^^^^Normalized inferior vena cava area ^^^^^^^^^^−^^Measured normalized inferior vena cava area ^^^^^^^^^^Estimated normalized inferior vena cava area AUC Area Under CurveBOSS Bag of Symbolic Fourier Approximation Symbols^^^^^^^^^^Adjusted collapsibility index ^^^^^^Measured collapsibility index ^^^^Estimated collapsibility index CI Collapsibility indexCO Cardiac outputCongestion Index A new index of heart failure congestion status based on regularly collected data from a passive sensor implanted in the inferior vena cava CRT Cardiac Resynchronization TherapyData engine A framework that collects, processes, and routes and organizes data into structured formats Data window Time window of data considered for analysis (e.g.30 days) Decision logic An arrangement of one of more analog or digital circuits, gates or switches configured to determine an output based on stated input parameters and which may comprise or be responsive to one or more instruction sets Detector or Feature A structure comprised of analog or digital circuits, algorithms or Detector a combination thereof specifically configured to identify features of interest in an input 11 Attorney Docket No.15653-022WOU2EFS Early Feasibility Study, conducted using Applicant’s disclosed sensors Event detection accuracy The percentage of heart failure events that the Congestion Index correctly detects, where a heart failure event is considered to be correctly detected if the mean Congestion Index for a grouping of days crosses an upper thresholdEvent detection threshold An upper threshold defined to identify elevated risk of hospitalization False positive A day (or mean across several days) when a threshold is crossed outside of an event windowFFT Fast Fourier TransformationFIH study First in Human study, conducted using Applicant’s disclosed sensors FPR False positive rate, which is the average value across all patients of the percentage of days, outside an event window, triggering an alert Generator A structure comprised of analog or digital circuits, algorithms or a combination thereof specifically configured to generate stated parameters from identified inputs Geometric mean The Nth root of the product of all values, used to select an upper threshold for event detection by maximizing the distance from a line of chanceGUI Graphical User InterfaceHazard ratio The ratio between the rates at which events occurHcol Cardiac collapseHF Heart failureHFE Heart failure eventHFH Heart failure hospitalizationHFpEF Preserved ejection fractionHFrEF Reduced ejection fractionHR Heart rateIVC Inferior vena cavaLAN Local area networkLB Lower boundary of inferior vena cava areaLCD Liquid crystal displayLME Linear mixed effectsLower threshold A Congestion Index score cutoff indicating that Congestion Index scores below that cutoff are below a normal range, which could indicate that the patient has a low risk of heart failure hospitalization due to volume overload yet has a risk of hypovolemia, that patient data needs attention and review, and / or reduced diuresisMAP Mean arterial pressureModule A code structure encoding one or more algorithms or such code structure plus processing platform specifically configured to produce a stated output from identified inputsNT-proBNP N-terminal pro-B-type natriuretic peptideNYHA New York Heart Association12 Attorney Docket No.15653-022WOU2Odds ratio The probability of an event occurringPAN Personal area networkPAP Pulmonary Artery PressurePCWP Pulmonary capillary wedge pressurePLD Programmable logic devicesPLR Passive leg raisePrediction horizon The time from when a prediction is made until the event is expected to appear, e.g. the time from when a prediction is made until an HFE is expected to appear RAP Right atrial pressureRcol Respiration collapseRHC Right heart catheterizationROC Receiver operating characteristicROI Region of interestRR Respiration rateSEM Standard error of measurementSensitivity The percentage of time windows labelled as “event” (such as an HFE) that are correctly detected Specificity 100 minus the false positive rate (may be expressed as a percentage), or the percentage of time windows labelled “non- event” (such as no HFE) that are correctly identified as a non- event SQ Signal qualitySVR Systemic vascular resistanceti,card Interval time per cardiac cycle ti,respInterval time per respiration cycle UAR Unexplained alert rate, which is the rate of false positive alerts per time period UB Upper boundary of inferior vena cava areaUpper threshold A Congestion Index score cutoff indicating that Congestion Index scores above that cutoff are above a normal range, which could indicate elevated risk of impending heart failure hospitalization, that the patient data needs attention and review, and / or increased diuresis VR Venous returnWAN Wide area network

[0017] Embodiments disclosed herein include devices, systems, and methods allowing forperiodic monitoring of fluid volume more accurately than current techniques and with reduced lag time before changes in volume status are observed, thus providing earlier warning of hypervolemia or hypovolemia and enabling the modulation of patient treatments to permit more stable long term fluid management. 13 Attorney Docket No.15653-022WOU2

[0018] The present disclosure also addresses above-identified problems by developing a newindex of heart failure (HF) congestion index. In some embodiments, systems described herein include passive, implantable sensors and non-invasive receiving equipment that measure the temporal variation in IVC area as a means to remotely assess and manage heart failure disease status and support self-management. Use of the new HF congestion index—particularly when the congestion index incorporates normalizing information described herein—improves speed and accuracy of heart failure diagnostic and treatment decisions and thus improves patient outcomes and reduces hospitalization occurrences and costs. Systems and methods described herein also improve the functioning of heart- failure-diagnostic systems compared to prior systems.

[0019] FIG.1 depicts at a high level a system 10 for receiving and analyzing hemodynamic sensortraces to assist in providing more accurate diagnosis and improved treatment of cardiac-related conditions, particularly in heart failure patients. In general, systems in accordance with embodiments described herein will include a sensor subsystem that produces a signal representing patient hemodynamic function, a data analysis subsystem that receives signal data from the sensor subsystem and processes the received data to automatedly generate diagnostic and treatment recommendations, and an interface subsystem that permits patient and healthcare-provider interaction with the system. System 10 as shown in FIG. 1 exemplifies one such embodiment, including an implanted sensor 12 and a patient-worn processing device 14 that receives the raw sensor signal and provides initial signal processing. Processing device 14 communicates with further processing platform(s) 16 via wireless communications links 18. Between processing device 14 and processing platform(s) 16, system 10 reads and interprets cardiac-health-state information contained within the sensor signal to provide diagnostic and treatment recommendations based on the extracted information. Data communication may be optionally facilitated through a patient personal device 20 such as a phone or tablet, which also may function as an input and output device for the patient as further described herein. An optional healthcare provider device 22 also may be provided to facilitate healthcare provider interaction with the patient and the system as further described herein.

[0020] Sensor 12 may comprise an external sensor system or an implanted sensor system.Examples of sensor 12 include vascular dimension sensors, such as an IVC area or diameter sensor, and vascular pressure sensors. With respect to vascular dimension sensors, a number of different sensor types may be used to produce an area trace signal including, for example, implanted variable inductance coils and implanted or external ultrasound devices. In one specific implementation, sensor 12 is an implanted wireless resonant circuit sensor and processing device 14 comprises a belt antenna 14 Attorney Docket No.15653-022WOU2as described, for example, in the incorporated USP 10,806,352. Other sensor types, such as implanted or external ultrasound, and implanted resistance-based sensors, may be employed as described in the foregoing incorporated patents and published applications.

[0021] Communication links 18 may be wired, wireless, or a combination thereof based on thespecific configuration of a system in accordance with the present disclosure. Persons of ordinary skill in the art may configure an appropriate data transmission protocol for communication link 18, selected from among many available standards. Communication links 18 are preferably bi-directional communication links. For example, a personal area network (PAN) connection such as Bluetooth may connect sensor processing device 14 to patient personal device 20. In some embodiments, personal device 20 may contain one or more software applications to perform signal processing functions with respect to the sensor signal. In other embodiments, personal device 20 may in this system merely act as an edge device to facilitate communication with processing platform(s) 16 configured as cloud platforms, with communication occurring via cellular data links. Communication links 18 between different system platforms and components also may comprise internet connections to provide data transfer, for example between a cloud platform comprising processing platform(s) 16 and healthcare provider device 22. In some embodiments, all or part of the functions of processing device 14 also may be executed as a cloud-based computing device.

[0022] An example of an IVC area signal trace 24 produced by system 10 is shown in FIG. 2. Ingeneral, as used herein area trace refers to a signal that presents vessel area as a function of time over a specific discrete time period corresponding to the trace reading and contains data representing cardiac and respiratory function, among other features. Processing platform(s) 16, such as diagnostics engine 32 described below, extracts that data by interpreting the features of the area trace. Some features of the area trace are identified in the enlarged box portion 26 of FIG.2. As shown therein, feature (a) represents the interval time per respiration cycle (ti,resp). Feature (b) represents the area magnitude of respiration modulation. Feature (c) represents the interval time per cardiac cycle (ti,card). Feature (d) represents the area magnitude of cardiac modulation in the IVC. Feature (e) represents the dominant cardiac peaks. Feature (f) represents the second cardiac peak of same cardiac cycle as the preceding peak (e). Feature (g) represents the respiration-related area reduction of the IVC.

[0023] Persons of ordinary skill in the art may derive many different ways to configure andexecute system 10 of FIG. 1 based on the teachings of the present disclosure. One example embodiment is exemplified in FIG. 3 as system 28. As illustrated therein, a sensor subsystem of 15 Attorney Docket No.15653-022WOU2system 28 comprises area trace generator 30, a data analysis subsystem comprises diagnostic engine 32, including programmable machines such as feature detector 34, metrics generator 36, boundary generator 38, index generator 40, decision logic 42, and alert generator 44, and an interface subsystem comprising one or more interface devices 46 to facilitate user interaction with the system, including uploading of patient specific data and receiving reports and notifications on patient cardiac health state. Diagnostic engine 32 also may communicate with one or more databases 48 and may receive additional patient-related information as external inputs from other patient sensors 50.

[0024] Trace generator 30 is a combined hardware and processing device comprised of a dataacquisition device 52 and signal processing device 54 as depicted in FIG.4. Data acquisition device 52 may comprise one or more of the aforementioned sensors 12. Signal processing device 54 is a signal amplifier / processor configured to receive raw data signals from data acquisition device 52 and to produce readable area trace signals suitable for communication and further processing within diagnostic engine 32. One example of a signal processing device is shown in FIG. 4 of the incorporated USP 10,806,352. In other examples, signal processing device 54 may comprise a software app executed on a patient personal device or a cloud platform. Examples of area traces produced in signal processing device 54 for communication to feature detector 34 include a regular respiration trace 56 and a maneuver trace 58.

[0025] Feature detector 34 is a machine such as circuitry, software, or a combination thereof thatextracts relevant data from the area trace signal. In some embodiments, feature detector wirelessly receives the area trace signal. Extracted data typically will include magnitudes and timing for selected features identified in an area trace through the reporting period (as shown as 60s trace in FIG.2); that is, plural of interval time per respiration cycle, area magnitude of respiration modulation, interval time per cardiac cycle, area magnitude of cardiac modulation, dominant cardiac peaks, second cardiac peaks, and respiration-related area reduction. As depicted in FIG.5, functional components for this purpose, which are generally understood and configurable by persons of ordinary skill in the art, may include an input signal integrity check for trace input 60 from trace generator 30, envelope detection 62, region of interest (ROI) detection 64, peak detection 66, and frequency analyzer 68. In some embodiments, feature detector 34 also may be an artificial intelligence-driven pattern recognizer 70. Pattern recognizer 70 employs machine learning and pattern analysis techniques to identify specific patterns in the area trace, which may be relevant in normalization and in diagnostic and treatment determinations as discussed more below. 16 Attorney Docket No.15653-022WOU2

[0026] In one embodiment, pattern recognizer 70 compares incoming area trace signals withknown area trace patterns to determine whether the incoming area trace is reflective of a feature such as a signal response to a patient maneuver. “Maneuver” as used herein refers to a physical action taken by a patient, on his or her own initiative or in response to instructions, which stimulates an identifiable perturbation of IVC area. Some examples of area traces for different patient maneuvers are shown in FIG.6, including supine: quiet respiration 72; sniff 74; supine: PLR (passive leg raise) 76; supine: breath-hold 78; seated 80; and seated to standing 82. An enlarged view of the area trace pattern 84 for supine: breath-hold 78 is shown in FIG.6A. Area trace patterns representing different maneuvers may be stored, such as in database 48 (FIG.3), and accessed by pattern recognizer 70. In one embodiment, this was achieved by transforming trace signals into a bag-of-words and applying a Bag-of-SFA-Symbols (BOSS) model. The model was then trained to detect maneuver-relevant features from such bag-of-words in order to classify maneuver / no maneuver with an accuracy of better than 80%. Knowledge of the type of trace pattern may be utilized by boundary generator 38 as further described below. In another embodiment, pattern recognizer 70 can be used to identify features within the area trace signals that have been shown to be predictive of clinical occurrences, such as atrial fibrillation or tricuspid regurgitation, for example.

[0027] Feature detector 34 also may perform data integrity checks. Features of the area tracesignal may be used to confirm if the system has been used correctly. For example, quality checks could be trained on supervised data as, for instance, the type of maneuver prescribed. Models predicting maneuver type or artifact can then be used to quality check every single area trace to avoid deriving corrupted information such as excess movement during recording, insufficient quantity or quality of maneuver performed, etc. Auxiliary sensors such as accelerometers, blood pressure, weight, activity monitors, and patient input can also be used to facilitate data integrity checks.

[0028] Data extracted from the area trace signal by feature detector 34 is provided to metricsgenerator 36. Metrics generator 36 is a programmable machine configured with a number of different data engines, as shown in FIG. 7, that generate specific cardiac-related metrics on which cardiac- related diagnostic and treatment determinations can be based. Included within optional frequency- derived metrics group 86 are heart rate module 88 and respiration rate module 90. Based on information received from feature detector 34, heart rate module 88 determines heart rate (HR) based on the relationship: [1] HR = 1 / (ti,card) 17 Attorney Docket No.15653-022WOU2where ti,cardis the cardiac cycle interval time, which may use an average or median, feature (c) in FIG. 2.

[0029] Also based on information from feature detector 34, respiration rate module 90 calculatesrespiration rate (RR) based on the following relationship: [2] RR = 1 / (ti,resp) where ti,resp is the respiration cycle interval time, which may use an average or median, feature (a) in FIG.2.

[0030] An alternative embodiment for determination of respiration rate by extraction from thearea trace is shown in FIG.42 and discussed in more detail below.

[0031] Included within area derived metrics group 92 are at least maximum IVC area (Amax)module 94, minimum IVC area (Amin) module 96, normalized IVC area (Anorm) module 98, and collapsibility index (CI) module 100. In some embodiments, instead of having collapsibility index (CI) module 100, area derived metrics group 92 comprises adjusted collapsibility index (^^^^^^^^^^) module 100A (see FIG. 7C), discussed in more detail below. Additional optional data engines may include mean IVC area (Amean) module 102 and data engines for determination of collapse 104 and cardiac output 106. Maximum IVC area (Amax) module 94 may determine Amax based on the value of the largest dominant cardiac peak, feature (e) in FIG. 2, occurring during the relevant sampling period. Minimum IVC area (Amin) module 96 may determine Amin based on the value of the greatest respiration area reduction valley, feature (g) in FIG. 2, occurring during the relevant sampling period. Alternatively, the maximum and minimum area values are extracted from a trace by finding the global maximum and minimum, excluding trace sections that are unusual in the light of the overall trace / excluding area sections of the trace that may have been corrupted by artifacts. An example of this technique is illustrated in FIG.7A. In this example, a moving window 108 of a set size is applied over area trace 110 with the minimum points 112 and maximum points 114 within this window extracted. After the window has moved across the entire trace, a series of Moving Amax and Moving Amin values will have been gathered. The median of all these Moving Amaxvalues is taken as the final result for Amaxand similarly for Amin. This size of moving window 108 is appropriately chosen to minimize or remove the effect of spurious spikes or dips in area trace 110 on the final reported Amax and Amin. Optional mean IVC area (Amean) module 102 determines Amean based on the maximum and minimum areas (one half the sum of Amaxand Amin), as shown, for example, in FIG. 7B. Maximum area 116 18 Attorney Docket No.15653-022WOU2and minimum area 118 for area trace 120 are added together and then divided in half to determine mean area 122.

[0032] In some embodiments, collapsibility index (CI) module 100 of metrics generator 36 usesthe determined area parameters to determine collapsibility index for the IVC based on the relationship: [3] CI = (Amax-Amin)∕Amax *100% ∙

[0033] In a further alternative embodiment, collapse (identified on the area trace in FIG. 7B bysubtracting minimum area 118 from maximum area 116) is separately included via collapse module 104, wherein collapse (Collapse) is determined based on the following relationship: [4] Collapse = Amax – Amin

[0034] Thus, collapsibility index also may be stated in some embodiments as:[5] CI = Collapse / Amax*100%

[0035] For example, respiratory collapse can be determined from the area trace. This has thepotential advantage that it is another signal that can be used to predict volume or congestion status or pressure—low collapse at high volume / pressure, high collapse at euvolemia / normal pressure, and low collapse at hypovolemia / low pressure giving an ‘n’ shaped collapse vs. area curve. Trending of features of the raw trace or the maneuver traces may also prove useful, i.e., accelerating increase in area may necessitate more urgent or severe action than gradual increases. In another example, other frequency-based signals can also be extracted and could be used as inputs to the calculation. As a further example, it is known that the IVC area signal changes with each breath of the patient and therefore the low frequency oscillation of the IVC can be used to extract the respiration rate. Respiration rate has been shown to be predictive of heart failure status and would therefore be a strong input into the overall patient status estimation. Additionally, metrics combining frequency and area inputs 124 may be determined.

[0036] Inputs from wearable devices such as activity trackers can also be predictive of cardiacpatient outcomes and can also be integrated into the calculation with reduced activity being a predictor of worsening status. Also, weight is a factor that may be used in prediction of heart failure decompensation, for example a gain in the region of more than 2 kg in 2 days may be considered significant. Patients taking their daily weight can also be integrated into the calculation. Sleep, 19 Attorney Docket No.15653-022WOU2activity, heart rate variability, blood pressure, and other wearable outputs could also be integrated into the Congestion Index. Additionally, data from the implanted sensor and its overall system 28 can be integrated into wearable devices’ systems as a source of additional data for the prediction and / or monitoring of health status.

[0037] Returning to FIG.7, included within the other input-derived metrics group 126 are metricsderived from other external sensors (meaning external with respect to system 28, which may include both in vivo and ex vivo sensors) 128, such as pulse oximetry, temperature, blood pressure, urine output, cardiac output, and catheter pressures, etc., and patient-specific information 130, which generally comprises information about patient physiological parameters such as height, age, weight, and sex, and may also comprise current activity information, input through a user interface by a patient or care provider. Another alternative external sensor input 128 is accelerometer readings from a patient-worn component of the sensing system, such as patient-worn processing device 14, which may be embodied as an antenna belt as described in incorporated patent publications. Such accelerometer readings can be used to determine patient position and activity / motion during a trace period and thus increase accuracy of metrics derived from the trace.

[0038] As shown in FIGS. 3 and 15, metrics from metrics generator 36 also may be directlyprovided to decision logic 42 for application per specific diagnostic or treatment algorithms. Metrics generator 36 may also send metrics directly to alert generator 44 and / or to interface device 46 for display and monitoring by user. Area-based metrics are also provided to boundary generator 38 for use in boundary determination.

[0039] FIG. 7C is a block diagram of an alternative embodiment of a metrics generator accordingto the present disclosure. Like metrics generator 36, metrics generator 36A is configured to generate heart function-related parameters for each area trace based on identified characteristic features from feature detector 34, such as features representative of maximum and minimum vessel areas. Metrics generator 36A comprises area derived metrics group 92A, which includes maximum IVC area (Amax) module 94, minimum IVC area (Amin) module 96, normalized IVC area (Anorm) module 98, and adjusted collapsibility index (^^^^^^^^^^) module 100A. In addition to receiving identified characteristic features from feature detector 34, normalized IVC area (Anorm) module 98 receives the upper area boundary (UB) and lower area boundary (LB) from boundary generator 38 or boundary generator 38A, and adjusted collapsibility index (^^^^^^^^^^) module 100A receives the lower area boundary (LB) from boundary generator 38 or boundary generator 38A. The upper area boundary (UB) and lower 20 Attorney Docket No.15653-022WOU2area boundary (LB) are discussed in more detail below in connection with FIG.8 (showing boundary generator 38) and FIG.8A (showing boundary generator 38A).

[0040] Normalized IVC area (Anorm) module 98 is configured to set the normalized IVC area(^^^^^^^^^^) based on the upper area boundary (UB), the lower area boundary (LB), and the minimum IVC area (Amin). In some embodiments, normalized IVC area (Anorm) module 98 is configured to set the normalized IVC area (^^^^^^^^^^) to a measured normalized IVC area (^^^^^^^^^^−^^), which in some embodiments is set based on Equation 12 below. In other embodiments, normalized IVC area (Anorm) module 98 is configured to set the normalized IVC area (Anorm) as equal to the mean of the measured normalized IVC area (^^^^^^^^^^−^^) and an estimated normalized IVC area (^^^^^^^^^^). In these embodiments where the normalized IVC area (Anorm) is based in part on estimated normalized IVC area (^^^^^^^^^^), normalized IVC area (Anorm) module 98 is configured to generate estimated normalized IVC area (^^^^^^^^^^) by applying a mixed-effects regression model to a patient population containing plural vessel area traces for each patient of the patient population, as described in more detail below.

[0041] Adjusted collapsibility index (^^^^^^^^^^) module 100A is configured to set the adjustedcollapsibility index (^^^^^^^^^^) based on the lower area boundary (LB), the maximum IVC area (Amax), and the minimum IVC area (Amin). In some embodiments, adjusted collapsibility index (^^^^^^^^^^) module 100A is configured to set the adjusted collapsibility index (^^^^^^^^^^) as equal to a measured collapsibility index (^^^^^^), which in some embodiments is set based on Equation 13 below. In other embodiments, adjusted collapsibility index (^^^^^^^^^^) module 100A is configured to set the adjusted collapsibility index (^^^^^^^^^^) as equal to the mean of the measured collapsibility index (^^^^^^) and an estimated collapsibility ). In these embodiments where the adjusted collapsibility index (^^^^^^^^^^) is based in part on the estimated collapsibility index (^^^^ ), adjusted collapsibility index (^^^^^^^^^^) module 100A is configured to generate the estimated collapsibility index (^^^^ ) by applying a mixed-effects regression model to a patient population containing plural vessel area traces for each patient of the patient population, as described in more detail below. An embodiment of metrics generator 36A was used in the FIH and EFS, which are also described in more detail below.

[0042] In order to overcome the calibration challenges created by intra-individual and inter-individual variations arising from heterogeneity of vessel shape as mentioned above, embodiments of the present disclosure generate specific reference points against which periodic IVC area readings can be compared to assess current patient fluid state and related cardiac health. In one example, boundary 21 Attorney Docket No.15653-022WOU2generator 38, shown in FIG. 8, uses IVC area data extracted or predicted from the area trace to determine upper and lower boundaries for this purpose. Boundaries may be considered as static or dynamic. Static boundaries are maximum / absolute values that would not be expected to significantly change over longer time periods. The absolute max / min area values represent upper and lower edges of the individual IVC area range and can be considered as static boundaries. Absolute maximum area may be related to total circulating blood volume. It could be assumed that these boundaries are anatomical constants—that the IVC in a specific patient can always only get to be a certain size. In this way, the boundaries can be considered to be static over time. This allows a single maneuver to be used to understand a patient’s IVC range and for those boundaries be used over time. This essentially means that a single calibration of the feature is performed. Static boundaries may be used, for example, in relationship to a specific set of conditions to be diagnosed.

[0043] On the other hand, dynamic boundaries may change over shorter time periods. Dynamicboundaries do not necessarily indicate absolute anatomical limits, but represent the current limits arising out of changing physiological parameters such as venous tone, and / or intra-abdominal pressure. Dynamic boundaries thus may represent new and clinically relevant limits for each quiet respiration reading or occasionally when estimation is available from maneuver or other means to use the information that is in closest proximity time-wise to the readings used for volume status assessment. In other words, based on overall patient fluid state in terms of total fluid distribution between vascular and extravascular fluid, dynamic boundaries as defined herein may represent a more clinically relevant basis for assessing patient fluid state at the time of a specific reading. Furthermore, dynamic maximum area boundary may be related to RAP or vascular tone.

[0044] As shown in FIG. 8, boundary generator 38 includes two components for generation ofboundaries, lower boundary determination module 132 and upper boundary determination module 134. In various embodiments, lower boundary determination module 132 and upper boundary determination module 134 may be alternatively configured to determine the lower and upper boundary based on a maneuver or to predict the boundary based on regular respiration. In a further alternative, lower and upper boundary determination data engines 132 and 134 may comprise both maneuver determination sub-data engines 136, 138 and prediction sub-data engines 140, 142. FIG.9 illustrates the operation of maneuver determination sub-data engines 136, 138, showing an area trace for an inspiration breath-hold maneuver 144 overlaid over a regular respiration area trace 146. Lower boundary determination 132 accounts for any physiological or device-related restriction preventing the IVC from full collapse, e.g. sensor radial force, sensor positioning across a venous branch, or other 22 Attorney Docket No.15653-022WOU2physically restricting features of the anatomy or implanted device. Maneuver determination sub- module 138 may receive information on the type of maneuver in a number of ways. For embodiments employing a feature detector 34 including trace pattern recognizer 70, the system may provide the maneuver type information directly. Alternatively, a user such as a care provider, when instructing the patient to perform the maneuver, would also input information via an interface device 46 informing of the instruction to perform a maneuver. Trace pattern recognizer 70, where present, may also be used to confirm patient compliance with the maneuver instruction by comparing the received area trace pattern with stored patterns for the instructed maneuver. Regardless of the source, when informed of the performance of a maneuver, maneuver determination sub-module 138 identifies the maximum area in the maneuver area trace, point 148 in FIG.9, as the upper boundary.

[0045] In some situations it may not be practical or desirable to require the patient to perform amaneuver. Prediction sub-module 142 is configured to predict the upper boundary (UB) based on a normal respiration area trace based on the following relationship developed by the Applicant: [6] UB = (-1 ∕ slope) ∙ CI + Amean

[0046] In one alternative, the “slope” used in Eq. [6] is a constant reference slope of -0.18 % / mm2as identified in Huguet et al., Three-Dimensional Inferior Vena Cava for Assessing Central Venous Pressure in Patients with Cardiogenic Shock, J Am Soc Echocardiogr. 2018; 31: 1034-43 (https: / / doi.org / 10.1016 / j.echo.2018.04.003), which is incorporated by reference herein in its entirety. The slope used may be retrieved from data storage 150. In other alternative embodiments, this slope could be individualized based on each patient’s area and collapse data or may be trained on a collection of area and collapse or other data over some period of time.

[0047] In a preferred embodiment, upper boundary determination module 134 of boundarygenerator 38 sets the upper area boundary (UB) to a largest historically observed maximum IVC area value measured for the patient. Further alternative techniques 152 for upper boundary generation also may be derived based on the teachings of the present disclosure.

[0048] Lower boundary (LB) determination module 132 in some embodiments also employs amaneuver determination sub-module 136 to identify the lower boundary as the minimum area point (point 154 in FIG.9) in a maneuver area trace in substantially the same manner that upper boundary maneuver determination sub-module 138 identifies the upper boundary. For example, lower boundary determination may be made via sniff maneuver, or any other maneuver that reduces the IVC size, 23 Attorney Docket No.15653-022WOU2using methodology described above. Alternatively, lower boundary prediction 140 uses individual patient or population area and collapse data to predict the lower boundary (LB) using a method similar to upper boundary prediction 142 using Eq. [7]: [7] LB = (-1 ∕ slope) ∙ C + Amean

[0049] Further alternative sub-data engines 156 may also be provided to derive the lowerboundary value from a population basis as the average minimum sensor size measured in the IVC, or sensor information regarding IVC area at given equivalent pressures, such as by comparing sensor- derived area / pressure curves with standard area / pressure curves (determined from radial or flat plate force information) and be a programed constant for all patients. In an embodiment, lower boundary determination module 132 of boundary generator 38 sets the lower area boundary (LB) to the smallest historically observed minimum IVC area value measured for the patient. In a preferred embodiment, lower boundary determination module 132 of boundary generator 38 sets the lower area boundary (LB) to about 150mm2. Then, if the smallest historically observed minimum IVC area value measured for the patient falls below 150mm2, lower boundary determination module 132 resets the lower area boundary (LB) to that smallest historically observed minimum IVC area value that is less than 150mm2after checking for the presence of a signal artifact.

[0050] FIG. 8A is a block diagram of an alternative embodiment of a boundary generatoraccording to the present disclosure. Like boundary generator 38, boundary generator 38A is configured to generate a lower area boundary (LB) for the patient and an upper area boundary (UB) for the patient. Boundary generator 38A comprises lower boundary determination module 132A and upper boundary determination module 134A. Lower boundary determination module 132A sets the lower area boundary (LB) to an initial predetermined value. In some embodiments, the initial predetermined value is about 150mm2. In some embodiments, if the smallest historically observed minimum IVC area value measured for the patient (out of all area traces for the patient) falls below the initial predetermined value, then lower boundary determination module 132 resets the lower area boundary (LB) to that smallest historically observed minimum IVC area value that is less than the initial predetermined value after checking for the presence of a signal artifact. Upper boundary determination module 134A comprises maneuver determination sub-module 138A, which receives maximum areas associated with breath-hold maneuvers from feature detector 34. Upper boundary determination module 134A sets the upper area boundary (UB) to a largest historically observed maximum IVC area value measured for the patient (out of all area traces for the patient). An 24 Attorney Docket No.15653-022WOU2embodiment of boundary generator 38A was used in the FIH and EFS studies, which are described in more detail below.

[0051] Upper and lower boundaries generated by boundary generator 38 and boundary generator38A are provided to index generator 40, shown in FIG.10. Upper and lower boundaries generated by boundary generator 38 and boundary generator 38A may also be provided to metrics generator 36 and / or metrics generator 36A. These boundary values enable normalizing the area measurement of the vessel to return a number ranging from 0% to 100% or 0 to 10, with 0% being close to 0 mmHg internal filling pressure or low volume / low congestion / hypovolemia and 100% being close to the maximal possible filling pressure in this vessel (~20 mmHg) or high volume / high congestion / hypervolemia. Index generator 40 assigns the boundaries to a relative scale with the lower boundary (LB) set as the scale minimum and the upper boundary (UB) set as the scale maximum. In some embodiments, the relative scale is established as a 0–100% scale. Once the relative scale is established, current area readings are scored on the relative scale, for example, an area reading falling halfway between the upper and lower boundaries would be scored as 50%. Because each patient is capable of displaying a range of IVC areas based on a number of factors including, but not limited to, anatomical size, current fluid status, body position, vascular tone status, and cardiac and abdominal pressures, normalization as described herein accounts for these parameters to produce a more standardized Congestion Index that is comparable across time for the patient and comparable across patients to help facilitate more accurate diagnosis and treatment of heart-failure-related conditions. As described above, at the outset, the new IVC area-based metrics disclosed herein may not be well- understood and correlations to well-known physiological measures can aid in the clinical understanding of the early adopters and ultimately the clinical masses. In one embodiment, the Congestion Index may be determined based on the relationship of mean area and the upper and lower boundaries as in equation [8]: [8] Congestion Index = 100 *((Amean-LB) / (UB-LB))

[0052] In another example embodiment, index generator 40 uses the lower boundary determinedfrom boundary generator 38 or boundary generator 38A with collapsibility index (CI) to produce a vascular Congestion Index using the following equation: [9] Congestion Index = (Amax-Amin)∕(Amax -LB) *100% 25 Attorney Docket No.15653-022WOU2

[0053] An alternative of this equation might be to subtract from 100 in order to reverse thedirection of the index resulting in the equation:

[0010] Congestion Index = 100-(Amax-Amin)∕(Amax -LB) *100%

[0054] In yet another alternative, Congestion Index may be determined using the upper boundaryalone as the ratio of mean area to the upper boundary as follows:

[0011] Congestion Index = 100 * (Amean / UB)

[0055] In a further alternative embodiment, index generator 40 optionally receives and factorsmetrics from metrics generator 36 or metrics generator 36A and boundary generator 38 or boundary generator 38A into the normalized IVC Congestion Index. In such an embodiment, index generator 40 may compress multiple features into a single metric. For example, index generator 40 may be configured to output a single number within a standardized range for each patient. This single number output may be an indicator of a measurable variable such as RAP, worsening heart failure, impending hospitalization, probability of an impending event, a prompt to change medication, etc. This output will be common to all patients and contain or summarize the information from all of the inputs to provide an actionable output for clinicians. In a way, this is comparable to scales such as temperature, with 37 degrees Celsius as an accepted normal level and accepted high and low thresholds. In such a configuration, index generator 40 may output a Congestion Index ranging from 0 to 100% indicating the risk of an impending heart failure hospitalization within a selected time period, for example, the next thirty days.

[0056] In preferred embodiments, index generator 40 generates the Congestion Index based onnormalized IVC area (^^^^^^^^^^) and collapsibility index (CI) (including, for example, the adjusted collapsibility index (^^^^^^^^^^)). The IVC has been modelled as a collapsible tube, which, from a fluid- dynamics perspective, has a pressure-volume curve. Generating the Congestion Index based on normalized IVC area (^^^^^^^^^^) and collapsibility index (CI) is beneficial because IVC area can be used as a surrogate of volume and collapsibility index (CI) can be used as a surrogate of pressure, and together they provide insight into the compliance of the IVC and its utility in clinical evaluation of congestion status in heart failure (HF). Pre-clinical research using implantable sensors has demonstrated a consistent relationship between RAP and IVC area that can be used to elucidate a compliance-curve relationship in human subjects as a means to normalize an individual’s IVC area into a consistent range and to compare the individual’s IVC area against other HF patients. 26 Attorney Docket No.15653-022WOU2

[0057] FIG. 11 shows an example pressure-volume curve 158 of the IVC. In HF, volumeoverload usually precedes pressure overload, with a relatively large change in volume resulting in little corresponding change in pressure, until a critical point is reached at which pressure increases. To date, Pulmonary Artery Pressure (PAP) as an index of volume has been found to reduce hospitalization when it is used to actively monitor patients with HF. However, detection of elevated PAP may be a later sign of impending decompensation than early detection of increased volume. The IVC, on the other hand, continuously changes diameter and collapsibility along pressure-volume curve 158. Incorporating IVC area into the Congestion Index therefore allows earlier detection of changes in congestion, a reduction in hospitalization, and better patient outcomes.

[0058] Respiratory pressure variation 160 causes low collapse 162; respiratory pressure variation164 causes high collapse 166; and respiratory pressure variation 168 causes low collapse 170. The change in area of high collapse 166, which is associated with euvolemic region 172, is relatively greater than the change in area of low collapse 162, which is associated with hypovolemic region 174, and the change in area of low collapse 170, which is associate with hypervolemic region 176.

[0059] When using atrial pressure-based sensors or non-invasive systems for heart failuremanagement, clinician action is mandated by target thresholds derived from population data that must then be tuned towards the individual patient using a training period of several weeks as well as requiring period sensor recalibration. In contrast, systems and methods described herein introduce an approach that provides patient-specific thresholds which leverage the physiology of the IVC in providing a more direct measure of fluid volume. This personalized threshold is a more sensitive means of titrating HF medication and offers the possibility of patient-self-management. Systems and methods described herein also have the advantage of not requiring calibration or a training period, and instead providing patient-specific thresholds which leverage the physiology of the IVC, which is a more direct measure of fluid volume than conventional measures.

[0060] In some embodiments where the Congestion Index is based in part on normalized IVCarea (^^^^^^^^^^), metrics generator 36 or metrics generator 36A sets ^^^^^^^^^^as follows:

[12] ^^^^ −^^^ ^= 100 ∗ ^^^^^^^ ^^^^^^^^^^^−^^^^Where Aminis the minimum area obtained from the supine, normal respiration portion of the 60 second daily trace. 27 Attorney Docket No.15653-022WOU2

[0061] Normalized IVC area (^^ ) alternatively may be set as follows:^^^^^^^^[12a] ^^^^^^^^^^−^^^^ ^^^^^^^^ = 100 ∗^^^^−^^^^ −^^^^[12b] ^^ =^^^^^^^^

[0062] A limitation in conventional HF diagnostic tools and methods is the difficulty incomparing values obtained from sensors on a daily basis both within and across patients, given the wide variation in the population. Normalizing IVC area according to Equation 12 solves this problem by normalizing IVC area data onto a 0-100 range.

[0063] In some embodiments where the Congestion Index is based in part on collapsibility index(CI), adjusted collapsibility index (^^^^ ) module 100A of metrics generator 36A sets the adjusted^^^^^^collapsibility index (^^^^ ) as equal to the mean of a measured collapsibility index (^^^^ ) and an^^^^^^ ^^estimated collapsibility index (^^^^ ). Metrics generator 36A may set the measured collapsibility index (^^^^ ) as follows:^^^^ −^^^^^^^^ ^^^^^^

[0013] ^^^^ = 100 ∗^^^^ −^^^^^^^^^^generate the estimated collapsibility index (^^^^ ) by applying amixed-effects regression model, such as, but not limited to, a linear regression model, to a patient population containing plural vessel area traces for each patient of the patient population. In some embodiments, the mixed-effects regression model is applied to a plot of the measured collapsibility index (^^^^ ) versus the normalized IVC area (^^ ) for a patient population. For example, metrics^^ ^^^^^^^^generator 36A may set the estimated collapsibility index (^^^^ ) as follows: 2

[0014] (^^^^ ) = A − B ∗ A + C ∗ A ^^^^^^^^ ^^^^^^^^

[0065] are A is the intercept; B is the linear term (slope);and C is the curvature term. In an embodiment, A is in a range of 0 to 100,000, B is in a range of 0 to 100,000, and C is in a range of 0 to 100,000. In the example shown in FIG. 14B and discussed inmore detail below, A is 88.8592, B is 1.8764, and C is 0.0119, making (^^^^ ) = 88.8592 − 1.8764 ∗2A + 0.0119 ∗ A.^^^^^^^^ ^^^^^^^^28 Attorney Docket No.15653-022WOU2

[0066] With normalized IVC area (^^^^^^^^^^), measured collapsibility index (^^^^^^), and estimatedcollapsibility index (^^^^ ) all set, index generator 40 may set the Congestion Index as follows: (100−^^^^^^)+(100−^^^^) ^^

[0015] Congestion Index =^^^^^^^^+ ( 2 ) 2

[0067] index (^^^^^^^^^^) is equal to the mean of themeasured collapsibility index (^^^^^^) and the estimated collapsibility index (^^^^ ), Equation 15 may be rewritten as Congestion Index =^^^^^^^^^^+ (100−^^^^^^^^^^)2 .

[0068] of estimating the collapsibility index (CI),normalized IVC area (^^^^^^^^^^) is estimated, and metrics generator 36A sets the normalized IVC area (^^^^^^^^^^) as equal to a mean of a measured normalized IVC area (^^^^^^^^^^−^^) and an estimated normalized IVC area (^^^^^^^^^^). In these embodiments where normalized IVC area (^^^^^^^^^^) isestimated, the estimated normalized IVC area ^^ )( may be generated by applying a mixed-effects^^^^^^^^regression, such as, but not limited to, a linear regression model to a patient population containing plural vessel area traces for each patient of the patient population. In some embodiments, the mixed- effects regression model is applied to a plot of the measured collapsibility index (^^^^ ) versus the^^measured normalized IVC area (^^ ) for a patient population. Index generator 40 may set the^^^^^^^^−^^Congestion Index as follows: (100−^^^^^^^^^^^^−^^^^^^)+(^^ +^^^^^^^^^^) Index =^^2

[0069] area (^^ ) is equal to the mean of the^^^^^^^^measured normalized IVC area (^^ ) and the estimated normalized IVC area (^^ ), Equation^^^^^^^^−^^ ^^^^^^^^(100−^^^^ )+(^^ )^^^^^^^^^^^^^^ . The adjusted collapsibility index16 may be rewritten as Congestion Index =(^^^^ ) may be set

[0070] Two separate have been conducted using system 10 where index generator40 generated the Congestion Index based on normalized IVC area (^^ ) and adjusted collapsibility^^^^^^^^index (^^^^ ): a First in Human study (FIH) and an Early Feasibility Study (EFS). As described in^^^^^^more detail below, these studies have shown that a Congestion Index based on direct, daily measurement of IVC size and variation, coupled with patient-specific thresholds, provides very early 29 Attorney Docket No.15653-022WOU2indication of elevated risk of hospitalization and demonstrates the opportunity for a remote monitoring system to prevent hospitalization by providing daily, actionable information to patients and their care teams to guide timely medical intervention.

[0071] Between the FIH and the EFS studies, 63 patients were implanted with sensor 12. Patientsunderwent a Right Heart Catheterization (RHC) during the implantation procedure for sensor 12 in order to measure RAP. RAP data were captured using a standard Swan-Ganz catheter. The following Table 1 shows demographic and clinical data at the time of implantation for all 63 patients. Table 1 Variable Data N 61 Sex (M / F) 47M / 14F Age (yrs.) 65.77±9.72 BMI (kg / m2) 30.55±6.02 Height (cm) 171.03±10.27 Weight (kg) 89.67±20.18 6 minute walk (m) 274.70±110.23 NT-proBNP (ng / L) 1872.50±2675.00 BP systolic (mmHg) 117.08±15.33 BP diastolic (mmHg) 70.13±13.15 Arrhythmia (%) 62.30 Hypertension (%) 70.49 Hypercholesterolaemia (%) 49.18 NYHA class (Class III (%)) 77.05 Number HFH (previous 12 months) 1.73±1.04 Race (Black or African American (%)) 4.92Ethnicity (Hispanic or Latino (%)) 4.92 Heat Failure type (HFrEF (%)) 90.16 Dilated (%) 72.13 Aetiology (Ischemic (%)) 42.62

[0072] Patients were asked to take a 60-second daily reading while in a supine position fromsensor 12 throughout the studies, and made clinic visits after 1, 3, 6, and 12 months (as per each study’s clinical investigation protocol). At the end of each 60-second reading in a supine position, patients were asked to perform a breath-hold maneuver. Each daily reading therefore included a supine reading and a breath-hold reading. FIG.12 shows an example vessel area trace 177 from a daily reading from sensor 12, including both a supine reading 177A and a breath-hold reading 177B. Patients had a median of 252 days of recordings (with a range of 14 days to 1517 days). Each of the 61 patients was 30 Attorney Docket No.15653-022WOU2over 18 years old, had HF for at least three months, had a HF hospitalization within 12 months prior to being implanted with sensor 12, had elevated natriuretic peptide (NT-proBNP) (greater than 1000ng / L), and were on optimally tolerated HF treatment. Patients with both reduced ejection fraction (HFrEF) and preserved ejection fraction (HFpEF) were included in the studies. During the studies, a total of 25 adjudicated heart failure events (HFEs) occurred from 15 patients. An HFE is a HF-related hospitalization, HF treatment in a hospital day-care setting, or an un urgent outpatient HF visit. HFEs were adjudicated using a procedure following international guidelines.

[0073] A number of features were calculated from each daily reading obtained from system 10.Only readings obtained in a supine position and recorded in the home environment were included in this following analysis. However, data from the daily breath hold maneuver were used to normalize the IVC area on a daily basis. Feature detector 34 was configured to identify characteristic features within wirelessly received periodic vessel area traces from daily readings representing changes in fluid state of a patient over time. The identified characteristic features included features representative of maximum and minimum vessel areas and maneuver types. For example, referring to the example vessel area trace 177 in FIG.12, feature detector 177 identified supine reading 177A and breath-hold reading 177B, as well as a maximum area 177C associated with supine reading 177A, a minimum area 177D associated with supine reading 177A, and a maximum area 177E associated with breath-hold reading 177B. Feature detector 34 provided the identified characteristic features to metrics generator 36A, which was configured to generate heart function-related parameters for each area trace based on the identified characteristic features. Heart function-related parameters included IVC max area (Amax), which corresponded to maximum area 177C associated with supine reading 177A, and IVC min area (Amin), which corresponded to minimum area 177D associated with supine reading 177A.

[0074] Feature detector 34 provided maximum areas associated with breath-hold maneuvers toupper boundary determination module 134A of boundary generator 38A, which set the upper area boundary (UB) to a largest historically observed maximum IVC area value measured for each patient. This upper value can sometimes be achieved via a breath hold during the implantation procedure but may, in some cases, occur later during the patient’s use of sensor 12. For example, if the first vessel area trace for a patient has a maximum area associated with the breath-hold reading of 640 mm2, then boundary generator 38A begins by setting the upper area boundary (UB) to 640 mm2. If the second vessel area trace for the patient has a maximum area associated with the breath-hold reading of 635 mm2, then upper boundary determination module 134A keeps the upper area boundary (UB) set to 640 mm2. If the third vessel area trace for the patient has a maximum area associated with the breath- 31 Attorney Docket No.15653-022WOU2hold reading of 645 mm2, then upper boundary determination module 134A will reset the upper area boundary (UB) to 645 mm2. Lower boundary determination module 132A of boundary generator 38A set the lower area boundary (LB) to an initial predetermined value of about 150mm2, which approximates the population average. If the smallest historically observed minimum IVC area value measured for a patient fell below 150mm2, lower boundary determination module 132A reset the lower area boundary (LB) to that smallest historically observed minimum IVC area value that is less than 150mm2. Boundary generator 38A provided the upper area boundary (UB) and lower area boundary (LB) to metrics generator 36A.

[0075] Metrics generator 36A normalized area on a daily basis by setting the normalized IVCarea (^^ ) as equal to 100 * (A - LB) ∕ (UB -LB) according to Equation 12. Normalization using^^^^^^^^min the upper area boundary (UB) improves the longitudinal stability of Congestion Scores based on normalized IVC area (^^^^^^^^^^) and lessens the effect of daily variation in maneuver and supine data. Metrics generator 36A also set the adjusted collapsibility index (^^^^^^^^^^) as equal to the mean of a measured collapsibility index (^^^^ ) and an estimated collapsibility index (^^^^ ), and set the measured^^collapsibility index (^^^^ ) as 100 ∗^^^^^^^^−^^^^^^^^^^^^^^^^^^−^^^^ according to Equation 13. The presence of sensor 12 may prevent the IVC fromFor example, a patient may have a true lower area boundary (LB) of abo2ut 150mm , but the presence of sensor 12 in the patient’s IVC may prevent the 2 IVC min area (A ) from ever reaching 150mm . Setting the measured collapsibility index (^^^^ ) asmin^^^^ −^^^^^^^^ ^^^^^^ 100 ∗according to Equation 13, which incorporates the lower area boundary (LB) into the ^^ −^^^^^^^^^^equation, allows the collapsibility index to be adjusted to correct for the presence of sensor 12.

[0076] FIG.13 shows normalized IVC area (^^ ) plotted against measured collapsibility index^^^^^^^^(^^^^ ) for each of the 61 patients, where normalized IVC area (^^ ) is on each x-axis and measured^^ ^^^^^^^^collapsibility index (^^^^ ) is on each y-axis. Compliance curves 178 (only three labeled to avoid^^clutter) show a broadly consistent relationship suggesting that a general compliance curve can be fit to all data.

[0077] FIG. 14A shows data for each of the 61 patients combined into one plot of normalizedIVC area (^^ ) plotted against measured collapsibility index (^^^^ ). Compliance curve 180 shows^^^^^^^^ ^^the global relationship between normalized IVC area (^^ ) plotted against measured collapsibility^^^^^^^^index (^^^^ ) fit to all patient data.^^32 Attorney Docket No.15653-022WOU2

[0078] Because collapsibility index (CI) is affected by how patients breathe and by the rate ofrespiration, there is variation in the measured collapsibility index (^^^^^^) which is reflective of the breathing effort of the patient. To account for this, metrics generator 36A set the adjusted collapsibility index (^^^^^^^^^^) as equal to the mean of the measured adjusted collapsibility index (^^^^^^) and an estimated collapsibility index (^^^^ ), wherein the estimated collapsibility index (^^^^ ) was generated from a mixed- effects regression model. The mixed-effects regression model removed some of the variance from breathing effort.

[0079] Estimating collapsibility index (^^^^ ) based on daily normalized IVC area (^^^^^^^^^^) using alinear mixed-effects model yielded the following equation:

[17] (^^^^ ) = 88.8592 − 1.8764 ∗ A + 0.0119 ∗ A2 ^^^^^^^^ ^^^^^^^^

[0080] In FIG. 14B, adjusted collapsibility index (^^^^ ) is plotted against normalized IVC area^^^^^^(^^ ), wherein adjusted collapsibility index (^^^^ ) is represented by^^^^^^+ ^^^^^^^^^^^^ ^^^^^^2 . Like FIG.14A, all patient data is accumulated in FIG.14B. Compliance curve 182 shows the relationship betweennormalized IVC area (^^ ) plotted against adjusted collapsibility index (^^^^ ) fit to all patient data.^^^^^^^^ ^^^^^^( )100−^^^^ +(100− ^^^^^^^^ + ( )^^^^^^^^2

[0081] Index generator 40 then set the Congestion Index as equal toaccording to Equation 15. Incorporating normalized IVC area (^^provides stability from using context gathered from the implant of sensor 12 to date, and incorporating estimated collapsibility index (^^^^ ) is beneficial because it removes variance from breathing effort. Moreover, as set forth above, incorporating both normalized IVC area (^^ ) and adjusted^^^^^^^^collapsibility index (^^^^ ) is beneficial because IVC area can be used as a surrogate of volume and^^^^^^collapsibility index (CI) can be used as a surrogate of pressure, and together they provide insight into the compliance of the IVC. The resulting Congestion Index is personalized to each patient. For example, two patients may each have a Congestion Index score of 60, but have different IVC max area (Amax) and IVC min area (Amin) values from their supine reading. Patients could also have the same IVC max area (A ) and IVC min area (A ) values from their supine readings but have differentmax min Congestion Index scores, if, for example, they have different upper boundaries (UB) from their breath- hold readings. Continually updating the upper area boundary (UB) to reflect the maximum area associated with all received breath-hold readings means that thresholds applied to the Congestion 33 Attorney Docket No.15653-022WOU2Index (described in more detail below) do not need be updated, as instead the range over which they operate is being updated.

[0082] The IVC Congestion Index produced by index generator 40 can be directly reported byuser interface devices 46, but also can be an input for decision logic 42 as shown in FIG.15, and for alert generator 44. Decision logic 42 comprises one or more instruction sets for determining patient heart health status and treatment recommendations based on heart function parameters generated by metrics generator 36 or metrics generator 36A and the Congestion Index. Decision logic 42 may be embodied in one or more circuits, executed by one or more data processing devices, or distributed across a combination of such platforms. In general, the Congestion Index is input to various diagnostic and treatment algorithms 184, 186 executed in decision logic 42. Other metrics from metrics generator 36 or metrics generator 36A may optionally also be selectable inputs to various diagnostic and treatment algorithms 184, 186 executed in decision logic 42. In one example, the thresholds may be set as follows: ^0 to 15%: Below normal range which could indicate least likely risk of heart failurehospitalization due to volume overload, yet risk of low IVC area, which may indicate hypovolemia, patient data needs attention and review, potentially reduce diuresis ^15% to 70%: Normal range, potentially optimize guideline directed medical therapy^ 70% to 100%: Above normal range which could indicate elevated risk of impending heartfailure hospitalization, patient data needs attention and review, potentially increase diuresis

[0083] In this manner, any one feature of daily Congestion Index with selected cut-off thresholds(i.e., lower and upper) is now scaled onto the final Congestion Index as follows: ^The lower range from 0 to lower threshold will be scaled from 0 to 15%^ The normal range from lower threshold to upper threshold will be scaled from 15 to 70%^ The high range from upper threshold to 100% will be scaled from 15 to 100%

[0084] In some embodiments, alert generator 44 (FIG. 3) is configured to generate alerts for thepatient based at least on the Congestion Index produced by index generator 40, wherein the alert generator: generates an alert for patient attention and review of risk of low IVC area (which may 34 Attorney Docket No.15653-022WOU2indicate hypovolemia) if the Congestion Index is below a predetermined lower threshold; generates a patient notification of patient fluid status in a normal range if the Congestion Index is between the predetermined lower threshold and a predetermined upper threshold; and generates an alert for patient attention and review of risk of hypervolemia if the Congestion Index is above the predetermined upper threshold. As set forth above, in some embodiments, the predetermined lower threshold is 15 and the predetermined upper threshold is 70.

[0085] Utilizing an approach as outlined above, as an example, a priority list algorithm can becreated for moving patients into a higher priority status after a threshold crossing for a defined number of days and removing patients when within normal range for a defined number of consecutive days. As another example, a priority list algorithm can be created for moving patients into a higher priority status after a rate of change of the Congestion Index for a period of days and removing patients when within normal range for a defined number of consecutive days.

[0086] In some embodiments, decision logic 42 is configured to determine a medication changefor the patient based at least on the Congestion Index. The medication change may comprise an up- titration of a heart-failure medication based on time the Congestion Index spends above an upper medication threshold, and the medication change may comprise a down-titration of a heart-failure medication based on time the Congestion Index spends below a lower medication threshold. Heart- failure medications may comprise a diuretic, Guideline Directed Medical Therapy (GDMT), and / or any other heart-failure medication. The medication change may comprise an up-titration of Guideline Directed Medical Therapy when the Congestion Index is between the lower medication threshold and the upper medication threshold. In some embodiments, the medication change is based on a rate of change of the Congestion Index for a period of days. Determining a Congestion Index regularly, such as daily, offers the potential for use of system 10 as a patient self-management tool, whereby patients are automatically instructed to make a medication change based on time their Congestion Index spends above an upper threshold or below a lower threshold. Clinician action would only then be needed in the event the patient’s Congestion Index remains elevated for an extended period, which could trigger an alert to the treating physician.

[0087] For example, in some embodiments, the upper medication threshold is 70 and the lowermedication threshold is 15. If a patient has a Congestion Index above 70 on a given day, then they receive an alert from alert generator 44 instructing them to up-titrate a heart-failure medication (such as a diuretic and / or Guideline Directed Medical Therapy (GDMT)), and if the patient has a Congestion 35 Attorney Docket No.15653-022WOU2Index below 15 on a given day, then they receive an alert from alert generator 44 instructing them to down-titrate the heart-failure medication. In other embodiments, alert generator 44 does not send an alert unless the daily Congestion Index reading is over 70 or under 15 for a number of consecutive days, such as two, three, or four consecutive days.

[0088] In some embodiments, alert generator 44 is configured to generate an alert for physicianattention and review of risk of low IVC area if the Congestion Index is below the lower medication threshold for a predetermined number of days in a predetermined time window; and generate an alert for physician attention and review of risk of hypervolemia if Congestion Index is above the upper medication threshold for the predetermined number of days in the predetermined time window. For example, in some embodiments, if the Congestion Index is above 70 for five days out of a seven-day window, then alert generator 44 sends an alert to the patient’s treating physician. Similarly, in some embodiments, if the Congestion Index is below 15 for five days out of a seven-day window, then alert generator 44 sends an alert to the patient’s treating physician. In other embodiments, alert generator 44 sends an alert to the patient’s treating physician if the Congestion Index is above the upper medication threshold or below the lower medication threshold for any other number of days in a window.

[0089] Selecting a time window and number of days above or below a threshold for triggering analert to the patient’s treating physician may depend on the desired prediction horizon (see FIG.22C). For example, in some embodiments, when the desired prediction horizon is seven days, alert generator 44 sends an alert to the patient’s treating physician if the Congestion Index is above the upper medication threshold or below the lower medication threshold for five days in a seven-day window, and when the desired prediction horizon is fourteen days, alert generator 44 sends an alert to the patient’s treating physician if the Congestion Index is above the upper medication threshold or below the lower medication threshold for three days in a seven-day window.

[0090] In the FIH and the EFS studies, numerous statistically significant models of predictingHF-related hospitalizations were observed. A model was considered to be statistically significant if the odds ratio (OR) for risk of HF-related hospitalization was greater than 1. In the FIH and the EFS studies, for a seven-day prediction horizon, statistically significant models of predicting HFEs included having alert generator 44 send an alert to the patient’s treating physician if the Congestion Index was above 70 for four days out of a seven-day window (OR: 1.53), 11 days out of a 14-day window (OR: 1.32), 22 days out of a 30-day window (OR: 1.11), 42 days out of a 60-day window 36 Attorney Docket No.15653-022WOU2(OR: 1.06), and 52 days out of a 90-day window (OR: 1.03). In the FIH and the EFS studies, for a 14- day prediction horizon, statistically significant models of predicting HF-related hospitalizations included having alert generator 44 send an alert to the patient’s treating physician if the Congestion Index was above 70 for three days out of a seven-day window (OR: 1.3), seven days out of a 14-day window (OR: 1.16), 22 days out of a 30-day window (OR: 1.11), 42 days out of a 60-day window (OR: 1.06), and 52 days out of a 90-day window (OR: 1.03). In the FIH and the EFS studies, for a 30- day prediction horizon, statistically significant models of predicting HF-related hospitalizations included having alert generator 44 send an alert to the patient’s treating physician if the Congestion Index was above 70 for five days out of a seven-day window (OR: 1.31), nine days out of a 14-day window (OR: 1.15), 13 days out of a 30-day window (OR: 1.06), 23 days out of a 60-day window (OR: 1.04), and 52 days out of at 90-day window (OR: 1.02). In the FIH and the EFS studies, for a 60-day prediction horizon, statistically significant models of predicting HF-related hospitalizations included having alert generator 44 send an alert to the patient’s treating physician if the Congestion Index was above 70 for 52 days out of a 90-day window (OR: 1.02). In the FIH and the EFS studies, for a 90-day prediction horizon, statistically significant models of predicting HF-related hospitalizations included having alert generator 44 send an alert to the patient’s treating physician if the Congestion Index was above 70 for four days out of a seven-day window (OR: 1.26), 23 days out of a 30-day window (OR: 1.06), and 44 days out of a 90-day window (OR: 1.02).

[0091] In the FIH and the EFS studies, odds ratio analyses also showed that every day that theCongestion Index was above the upper threshold of 70 increased the probability of a patient having an HFE in the next 0 to as far out as 30 days. Moreover, hazard ratio analyses showed that every day that the Congestion Index was above the upper threshold of 70 in a week increased the hazard (expected rate) of an HFE occurring. The increasing odds ratio and hazard ratio each day support self- management being beneficial every day that the Congestion Index is above the upper medication threshold or below the lower medication threshold. Therefore, in preferred embodiments, if a patient has a Congestion Index above the upper medication threshold on a given day, then they receive an alert from alert generator 44 instructing them to up-titrate a heart-failure medication, and if the patient has a Congestion Index below the lower medication threshold on a given day, then they will receive an alert from alert generator 44 instructing them to down-titrate the heart-failure medication. Additionally, in the FIH and EHS studies, when a patient’s Congestion Index was above the upper medication threshold for five days in a seven-day window, the odds ratio was nearly 18 and the hazard ratio was about 2.5. 37 Attorney Docket No.15653-022WOU2

[0092] Because the Congestion Index is personalized for each patient, it is a valuable tool formanaging HF medications. Sensor 12 and the Congestion Index solve the problems of needing sensitive, timely, and actionable data (1) by personalizing the Congestion Index to each patient through normalization techniques, such as incorporating the upper boundary (UB), (2) by quantifying IVC area, which is an earlier indicator of HF events than changes in atrial pressure, and (3) by having daily area trace readings taken from sensor 12. Empowering patients to manage their fluid status in the home has significant benefits for both patients and providers, as it decreases the odds of having an HFE, and decreases the monitoring burden on providers.

[0093] Also, the Congestion Index as described above can be enhanced by factoring additionalheart function-related parameters as further described below to provide a potentially fuller picture of patient heart health status. Other examples of such algorithms are disclosed in the aforementioned and incorporated USP 11,564,596 and alternative embodiments could inform the up or down titration for specific medications, such as diuretics, vaso-dilators, or other medications, based on a pre- specified, patient-specific prescription. Additional examples of conditions that may be diagnosed and treatment recommendations generated via algorithms 184 and 186 include conditions such as tricuspid regurgitation, atrial / ventricular tachycardia, and bradycardia.

[0094] In other embodiments, maneuvers or similar changes can be utilized to move the patientbetween two or more different states. Signal differences between these states can be indicators of different physiological conditions. One example of this would be in the case of HFpEF patients where deterioration of condition parameters is only evident with exercise. In this case, the no exercise and exercise states could be used to provide a signal to diagnose the presence of HFpEF. Bendopnea is another example where a maneuver can result in shortness of breath and this could also be evidenced via the signal indicating respiration rate / magnitude changes. These maneuvers could be detected from the signal in feature detector 34 or could be inputs from the interface device 46.

[0095] Decision logic 42 may utilize any number of metrics as inputs. In some cases onlyindividual metrics or groups of related metrics, such as area metrics, may be used. In one alternative embodiment, metrics weighting sub-module 188 (FIG.15) applies a weighting factor to one or more of the individual metrics inputs. Weighting factors may be determined via analysis of clinical and pre- clinical data and pre-specified, or could be continually calculated and updated based on user-specific input. Notifications in this case may comprise instructions to take treatment actions, instructions to 38 Attorney Docket No.15653-022WOU2take additional readings, or instructions to a care provider to alert the care provider to possible clinically problematic situations.

[0096] As illustrated in FIG. 16, it is contemplated that interface devices 46 will comprise at leasta user interface 190 accessible to a care provider, for example via care provider device 192, for data inputs 194 such as patient-specific information, prescription details, user settings, maneuver instructions, algorithm updates, and other system hygiene, etc. User interface 190 could also provide notifications and / or alerts generated by decision logic 42, alert generator 44, or other components of diagnostic engine 32, and could be web-based or a mobile application. In some embodiments, interface devices 46 may include processing capabilities in the form of one or more processors 196 and associated hardware / software. Diagnostic or treatment alerts 198 also may be delivered through care provider device 192, for example as described in the aforementioned and incorporated USP 11,564,596. Interface devices 46 also may encompass patient personal devices 200, which may be employed for a variety of functions in the system in addition to patient notifications, as shown, for example, in FIG.1.

[0097] Embodiments of the present disclosure also may be described in terms of a continuousprocess flow, such as example process flow 202 shown in FIG.17, or alternative embodiments of the system may have versions of this formation. As shown therein, different data types invoke different processes. For example, data 204 representing a sensor time series for quiet respiration invokes process 206 to filter (respiration trace, cardiac trace, mean trace, noise trace) and process 208 for feature extraction (Amax, Amin, Anorm, Collapse, CI, respiration collapse, cardiac collapse, heart rate, respiration rate). Similarly, data 210 representing a sensor time series for a maneuver invokes process 212 to filter and transform using BOSS (Bag of Symbolic Fourier Approximation (SFA)-symbols) and process 214 to assess the maneuver trace using features from BOSS, as well as from other common descriptive statistics to assess trace quality and to extract max area achieved. External data 216, such as blood pressure, weight, blood oxygenation, etc., invokes process 218 to incorporate such data representing other externally sensed features from other sensors or direct inputs. Outputs of these processes may be stored in database 220.

[0098] Feature outputs from database 220 (or directly from upstream processes), invoke process222 to generate metrics as described herein, including, for example, Congestion Index 224, Right Atrial Pressure (RAP) 226, Respiration Rate (RR) 228, Cardiac Output (CO) 230, Systemic Vascular Resistance (SVR) 232, IVC tone 234, Heart Rate (HR) 236, Patient Weight 238, Patient Activity 240, 39 Attorney Docket No.15653-022WOU2and other metrics described herein. Metrics as generated are then processed 242 to produce an all metrics-based or enhanced IVC congestion index score 244, which is presented to a user or used as an input to decision logic 42 and / or alert generator 44 as a basis for further diagnostic or treatment decisions.

[0099] In some embodiments, the enhanced indexing process 242 is configured as an AI ormachine-learning-based model that is trained based on all described inputs and determines the best weightings for each metric in order to predict a specific output such as “worsening heart failure”, “increasing congestion”, “likelihood of hospitalization”, etc. One such example would be a multiple logistic regression model that uses a training dataset to develop a model which returns the probability of hospitalization based on the inputs or a subset of the inputs 224-240, where the training dataset comprises data from patients implanted with sensor 12. Another embodiment would be a decision tree model that splits the data from each of the inputs based on the training dataset to categorize into “high likelihood of hospitalization” or “no likelihood of hospitalization” for example, repeating this process for each input parameter until the model is created to accurately predict likelihood of hospitalization.

[0100] FIG.18 shows another embodiment of a continuous process flow, namely example processflow 246, which was used in the above-referenced FIH and EFS studies. At step 248, patients take a daily reading from implanted sensor 12. An example of a daily reading is shown in FIG. 12 and discussed above. At step 250, feature detector 34 detects features in sensor data, such as interval time per respiration cycle, area magnitude of respiration modulation, interval time per cardiac cycle, area magnitude of cardiac modulation, dominant cardiac peaks, second cardiac peaks, respiration related area reduction, maneuver types, and maximum and minimum areas, which in some embodiments are associated with identified maneuvers. In some embodiments, step 250 comprises receiving a vessel area trace containing at least two characteristic features representing vessel area parameters for a patient and identifying the characteristic features representing at least a first maximum area (Amax), a minimum area (Amin), and a second maximum area. In some embodiments, step 250 further comprises identifying a supine portion and a breath-hold portion of the vessel trace; setting the first maximum area (Amax) as a maximum area determined from the first supine portion; setting the minimum area (Amin) as a minimum area determined from the first supine portion; and setting the second maximum area as a maximum area determined from the breath-hold portion. In some embodiments, the vessel area trace is wirelessly received from a sensor implanted in a patient’s inferior vena cava. 40 Attorney Docket No.15653-022WOU2

[0101] At step 251, the upper area boundary (UB) and lower area boundary (LB) are set. In someembodiments, step 251 comprises setting the upper area boundary (UB) as the second maximum area, which may be the maximum area determined from the breath-hold portion; and setting the lower area boundary (LB) as an initial predetermined value, which in some embodiments is about 150mm2. In some embodiments, step 251 further comprises resetting the upper area boundary (UB) as a largest maximum area out of all vessel area traces for the patient; and resetting the lower area boundary (LB) as a smallest minimum area out of all vessel area traces for the patient if the smallest minimum area falls below the initial predetermined value.

[0102] At step 252, metrics generator 36 or metrics generator 36A generates metrics based onidentified features, such maximum IVC area (Amax), minimum IVC area (Amin), normalized IVC area (Anorm), and adjusted collapsibility index (^^^^^^^^^^). In some embodiments, step 252 comprises generating a normalized IVC area (^^^^^^^^^^) based on the upper area boundary (UB), the lower area boundary (LB), and the minimum IVC area (Amin); and generating an adjusted collapsibility index (^^^^^^^^^^) based on the maximum IVC area (Amax), the minimum IVC area (Amin), and the lower area boundary (LB).

[0103] At step 254, normalizing information is input into metrics generator 36 or metricsgenerator 36A, and at step 256, metrics are normalized. The steps of generating metrics 252 and normalizing 256 may go hand in hand, which is represented by a double-headed arrow between step 252 and step 256. While some metrics, such as maximum IVC area (Amax), could potentially be calculated without using normalizing information, other metrics, such as normalized IVC area (Anorm), may be calculated using normalizing information. For example, in the FIH and EFS studies, metrics generator 36A set normalized IVC area (Anorm) as equal to 100 * (Amin - LB) ∕ (UB -LB) according to Equation 12. This calculation for normalized IVC area (Anorm) includes normalizing information comprising the lower boundary (LB) and upper boundary (UB), which helps to improve longitudinal stability of the Congestion Index and lessen the effect of daily variation. In addition, as set forth above, in the FIH and EFS studies, metrics generator 36A set the adjusted collapsibility index (^^^^^^^^^^) as equal to the mean of the measured adjusted collapsibility index (^^^^^^) and the estimatedcollapsibility index (^^^^ ), where the measured collapsibility index (^^^^ ) was equal to 100 ∗^^^^^^^^−^^^^^^^^^^^^^^^^^^−^^^^ according to Equation 13 and the estimated collapsibility index (^^^^ ) was equal to 88.8592 −1.8764 ∗ A ^^^^^^^^ + 0.0119 ∗ A2^^^^^^^^according to Equation 17. 41 Attorney Docket No.15653-022WOU2

[0104] At step 258, index generator 40 determines the Congestion Index. In some embodiments,step 258 comprises generating the Congestion Index based on the normalized IVC area (^^^^^^^^^^) and the adjusted collapsibility index (^^^^^^^^^^). In the FIH and EFS studies, index generator 40 received the normalized IVC area (^^^^^^^^^^), the measured collapsibility index (^^^^^^), and the estimated collapsibility (100−^^^^^^)+(100−^^^^) ^^ + index (^^^^^^^^^( 2 ) ^^^ ), and set the Congestion Index as equal to 2 according to Equation 15.

[0105] At step 260, thresholds are set, such as the upper and lower thresholds discussed above.In some embodiments, step 260 comprises setting a lower predetermined threshold to 15 and setting an upper predetermined threshold to 70. In some embodiments, there is a self-management threshold, a call-clinic threshold, and a hospitalization threshold. At step 262, the thresholds from step 260 are applied to the Congestion Index from step 258.

[0106] At step 264, rules are set for alerts based on the thresholds. For example, a self-management rule may comprise sending a self-management alert if the Congestion Index is above the self-management threshold, and a call-clinic rule may comprise sending a call-clinic alert if the Congestion Index is either above the call-clinic threshold or has been above the self-management threshold for a preset consecutive number of days, such as three days. In other embodiments, the call- clinic rule may comprise sending a call-clinic alert if the Congestion Index has been above the self- management threshold for a predetermined number of days in a predetermined time window, such as five out of seven days.

[0107] FIG. 19 shows an approach to calculating the number of consecutive days above threshold266 comprising aggregating Congestion Indexes for days where the patient did a reading (by excluding from the analysis any days where a reading is missing), and then counting the number of consecutive days of the Congestion Index being above a threshold. Congestion Indexes were calculated based on readings taken during a five-day span for a first patient 268, second patient 270, and third patient 272. First patient 268 never had their Congestion Index above threshold 266 for more than one consecutive day. Second patient had their Congestion Index above threshold 266 for three consecutive days, and third patient had their Congestion Index above threshold 266 for two consecutive days. Even though third patient is missing a reading between those two consecutive days, out of the days where there was a reading, there are still two consecutive days above threshold 266. As another example of setting 42 Attorney Docket No.15653-022WOU2rules 264, a reduce-diuretics rule may comprise sending an alert to reduce diuretics if the Congestion Index has been below the lower threshold for a preset consecutive number of days.

[0108] In some embodiments, there is a rule for a notification pause period post-intervention. Atstep 274, alerts are sent based on the rules from step 264 and the application of thresholds in step 262. In some embodiments, step 274 comprises generating an alert for patient attention and review of risk of low IVC area if the Congestion Index is below a predetermined lower threshold and sending an alert for patient attention and review of risk of hypervolemia if the Congestion Index is above a predetermined upper threshold. Step 274 may further comprise generating a patient notification of patient fluid status in a normal range if the Congestion Index is between the predetermined lower threshold and the predetermined upper threshold. In some embodiments, step 274 may also further comprise generating an alert for physician attention and review of risk of low IVC area if the Congestion Index is below the predetermined lower threshold for a predetermined number of days in a predetermined time window; and generating an alert for physician attention and review of risk of hypervolemia if Congestion Index is above the predetermined upper threshold for a predetermined number of days in a predetermined time window. As set forth above, as an example, the predetermined number of days may be five days and the predetermined time window may be seven days.

[0109] In some embodiments, alerts comprise intervention advice 276, such as advice to call aclinic, go to a hospital, or adjust a medication, such as down-titration or up-titration of a heart-failure medication.

[0110] In some embodiments, alerts are sent at step 274 within about one second to about oneminute of patients taking a daily reading at step 248, which addresses the need for faster response times for observing changes in fluid volume status. At optional step 278, the Congestion Index’s performance in predicting heart failure hospitalization events as well as its association with mortality, RAP, NT-proBNP, and NYHA class are evaluated. Outcomes 280 from the performance evaluation 278 of Congestion Indexes from the FIH and EFS studies are discussed below. Due to the very early indication of elevated risk provided (up to 90 days), the results from the FIH and EFS studies demonstrate that the system 10 can be used to prevent hospitalization by providing daily, usable information to care teams to guide their actions. 43 Attorney Docket No.15653-022WOU2

[0111] Detecting Heart Failure Events and Mortality / Upper Threshold Selection

[0112] As explained in more detail below, there is a strong statistical association of theCongestion Index with HFEs and mortality events. The Congestion Index has been shown to be indicative of increasing risk of hospitalisation with an Odds Ratio (OR) of 1.78 (1.25-2.53, p<0.001) and Hazard Ratio (HR) of 1.61 (1.22–2.12, p<0.001) for each day that the Congestion Index is above 70% in a 30-day period.

[0113] An upper threshold was derived using data from the first adjudicated HFE for all patientswho had an HFE in order to provide a notification to patients and physicians that risk of hospitalization is elevated and that intervention is warranted. Only the first HFE per patient was considered so that they could each contribute an equal weight to the upper threshold, and if a patient experienced a mortality event or had an unadjudicated event pending adjudication, then they were excluded. The Congestion Index was found to be a significant predictor of adjudicated HFEs with the fitted weight value having p<0.001***.

[0114] Mean Congestion Indexes per groupings of days were considered at 3, 7, 14, 30, and 60days. For example, the 3-day mean is the mean of the average Congestion Index measured over three days. FIG. 20A shows the percentage of event detection accuracy vs. the mean false positive rate (FPR) for 3-day group 282, 7-day group 284, 14-day group 286, 30-day group 288, and 60-day group 290. Event detection accuracy is the percentage of HFEs correctly detected. An HFE was considered to be correctly detected if the mean Congestion Index for the grouping of days crossed the upper threshold, which in FIG.20A is 70%. A false positive is a day (or mean across several days, as is the case here), when the upper threshold is crossed outside of an event window. The mean FPR is the average value across all patients of the percentage of days, outside an event window, triggering an alert. 3-day group 282 was found to be the best level of aggregation with the highest geometric mean of event detection accuracy and 100–FPR, where the geometric mean is the Nth root of the product of all values, used to select an event detection threshold by maximizing the distance from line of chance 292. The analysis shown in FIG. 20A (and FIG. 20B described below) was repeated with several different event detection thresholds, and a threshold of 70% yielded the highest geometric mean.

[0115] FIG. 20B shows the percentage of event detection accuracy vs. the unexplained alert rate(UAR) expressed in median days per month for 3-day group 282, 7-day group 284, 14-day group 286, 30-day group 288, and 60-day group 290, where the UAR is the rate of false positive days per patient- month and the upper threshold was set to 70%. FIGS.20A and 20B and Table 2 below show that 3- 44 Attorney Docket No.15653-022WOU2day group 282, 7-day group 284, and 14-day group 286 means perform relatively similarly with the selected thresholds all reaching an event detection accuracy of 91.67%. However, 3-day group 282 has a slightly lower FPR and UAR.

[0116] Table 2 provides examples of thresholds that may be used to generate alerts. Thesethresholds are selected optimizing for event detection accuracy and median UAR or specificity, where specificity is 100-FPR and may be expressed as a percentage. Table 2 Mean Optimized for Threshold Event Detection Specificity Median UAR of Days Sensitivity and Rounded [%] Accuracy [%] [%] [Days / Month] 3 UAR 73 91.67 75.56 1.64 3 Specificity 76 83.33 84.44 0.74 7 UAR 69 91.67 68.89 3.08 7 Specificity 74 83.33 82.22 1.05 14 UAR 70 91.67 68.89 2.90 14 Specificity 74 83.33 77.78 0.94

[0117] FIGS. 21A and 21B show how the Congestion Index may vary in relation to dailynormalized IVC area (^^^^^^^^^^) and measured adjusted collapsibility index (^^^^^^^^^^) (FIG.21A) and to adjusted collapsibility index (^^^^^^^^^^) (FIG.21B). HFEs 294 and deaths 296 (not all labeled to avoid clutter) are overlaid on topIndex values for 7 days prior to HFEs or deaths. HFEs and deaths are seen to almost always occur when the Congestion Index is above a 70% threshold 298. Both HFEs and mortality events (deaths) are also statistically independent of the data used to generate the Congestion Index. In FIG.21B, patients who have had an HFE are marked separately from those patients who had never had an HFE to date.

[0118] Table 3 details performance of the Congestion Index in assessing the risk of HFEs acrossa number of time windows. Results assume a decision threshold of 70%. Table 3 Time window [days] 1 7 30 60 90Patient Accuracy [%] 95.24 85.71 75.00 65.48 51.19Event Accuracy [%] 91.30 82.61 73.91 65.22 52.17UAR patient-month7.50 7.29 6.77 6.43 6.22[ / p-m] Mean FPR [%] 34.01 32.83 29.88 27.94 27.06Decision Threshold 50 60 70 80 9045 Attorney Docket No.15653-022WOU2Patient Accuracy [%] 100.00 92.86 85.71 25.00 0.00Event Accuracy [%] 100.00 95.65 82.61 26.09 0.00UAR patient-month15.08 11.89 7.29 2.25 0.19[ / p-m] Mean FPR [%] 68.76 53.46 32.83 11.25 1.09Based on these results, the optimal decision threshold and time window are 70% and 7 days respectively.

[0119] The Congestion Index provided accurate detection of HFEs with prospective windows upto 90 days in advance of HFEs.

[0120] FIG. 22A shows box blots comparing the Congestion Index in a 7-day event windowagainst the Congestion Index in non-event windows of 7 days on either side of event windows as well as against patients who have not had an HFE. The mean Congestion Index 7 days before an HFE was 77.0%, while mean Congestion Index in non-event periods (random selection of 7 outside event window) for patients with events was 59.3%. The mean Congestion Index in patients with no HFEs was 56.1%. Using a one-way ANOVA test across all three categories found a statistically significant difference (p<0.001), while a Wilcoxon signed rank test found a statistically significant difference between the Congestion Index for patients in the event window compared to those same patients outside the event window. These results suggest a strong statistical association of the Congestion Index with HFEs.

[0121] FIG. 22B shows box blots comparing the Congestion Index in a 90-day event windowagainst the Congestion Index in non-event windows of 90 days on either side of event windows as well as against patients who have not had an HFE. Statistically significant differences were observed between the Congestion Index in the event window and the Congestion Index outside the event window, further supporting the strong statistical association of the Congestion Index with HFEs. Statistically significant differences were also observed between the Congestion Index in the event window and the Congestion Index for patients with no events. A Wilcoxon signed-rank test found that there was a significant difference between the Congestion Index in the event window compared to a random sampling of time periods outside the time window. A random sampling was chosen to ensure that equal amounts of event and non-event data were compared within each patient. Similarly, a Wilcoxon rank-sum test found that there were statistically significant differences (rank-sum: 46 Attorney Docket No.15653-022WOU282.000000, p-value: 0.008057) in the Congestion Index within the event window compared to a random selection of data for patients who had not had an HFE.

[0122] In order to determine how early the Congestion Index can detect HFEs, certainperformance measures were examined across a range of days prior to HFEs and decision thresholds (0 to 100). For example, FIG.22C shows the variation in sensitivity and UAR for HFE detection (for a 30-day data window, with threshold of 70%) across a varying number of days prior to the HFE. For example, the sensitivity and UAR values at 60 days prior to an HFE are based on the 30-day data window from 90 days prior to the HFE to 60 days prior to the HFE. This example may also be referred to as having a 60-day “prediction horizon,” where the prediction horizon is the time from when the prediction is made until the HFE is expected to appear. 22D is a graph showing the relationship between a data window, a prediction horizon, and an HFE.

[0123] FIG. 23 shows variation of UAR 300 and event detection accuracy 302 with decisionthreshold. These data suggest that the optimal decision threshold is 70%, as 70% has the largest spread between event detection accuracy and UAR, meaning that 70% is the best balance between maximizing event detection accuracy while minimizing UAR.

[0124] FIG. 24 shows receiver operating characteristic (ROC) curves obtained from eventdetection accuracy and UAR per patient-month for 1-day time window 304, 7-day time window 306, 30-day time window 308, 60-day time window 310, and 90-day time window 312. These data suggest that the optimal time window is 1 day or 7 days.

[0125] FIGS. 25A and 25B display an ensemble analysis of the Congestion Index prior toadjudicated HFEs, further supporting the strong statistical association of the Congestion Index with HFEs. Shaded areas for each trace are the standard error of measurement (SEM). In FIG.25A, data for all patients who experienced an HFE in a 7-day event window 314 was averaged and compared against a random sampling of data from the same patients outside the 7-day event window 316 as well as against patients who have not experienced an HFE 318. Results suggest that patients who had an HFE showed a statistically significantly higher Congestion Index in the period prior to the HFE 320 compared to those patients who have not experienced an HFE. A variety of decision thresholds and time windows were used to conduct an ensemble analysis of the Congestion Index against HFEs, with analysis suggesting that a threshold of 70% 322 provides an optimal balance between event detection accuracy and unexplained alert rate. 47 Attorney Docket No.15653-022WOU2

[0126] FIG. 25B shows an ensemble analysis of Congestion Index for all HFEs 324, with a 70%decision threshold and a 90-day event window. In FIG.25B, data for all patients who experienced an HFE in a 90-day event window 324 was averaged and compared against a random sampling of data from the same patients outside the 90-day event window 325 as well as against patients who have not experienced an HFE 326.

[0127] Over the course of both the FIH and EFS trials, 5 patients experienced a mortality event.Data shown in FIGS.26A and 26B show a strong statistical association of the Congestion Index with mortality events. FIG. 26A shows comparisons of the Congestion Index in patients 7 days before death (during a 7-day event window) to the Congestion Index prior to this 7-day event window as well as to patients who were still alive. The mean Congestion Index 7 days before death was 74.8% whereas the mean Congestion Index for patients who were still alive was 58.0%. A statistically significant difference was observed in the Congestion Index (rank-sum: 1422.00, p= 0.004416) in the period 90 days prior to death compared to patients who are still alive. No significant difference was observed between the score in the days prior to death and the period prior to this 90-day window.

[0128] FIGS. 27A and 27B display an ensemble analysis of the Congestion Index prior to death328, further supporting the strong statistical association of the Congestion Index with mortality events. Shaded areas for each trace are the SEM. In FIG. 27A, data is for all 5 patients in the 7-day period prior to death 330 and is averaged and compared against a random sampling of data from the same patients outside this 7-day period 332 as well as against patients who were still alive 334. In FIG. 27B, data is for all 5 patients in the 90-day period prior to death 336 and is averaged and compared against patients who were still alive 338 with the 70% threshold 327 identified. Results suggest that patients who experienced a mortality event showed a statistically significantly higher Congestion Index in the period prior to death compared to those patients who were still alive and that the high values may extend beyond the 7 and 90-day periods examined.

[0129] Lower Threshold Selection

[0130] In the FIH and EFS studies, a lower threshold was defined based on analysis of thepopulation distribution in order to provide a notification to the patient and physician that that patient’s IVC area has decreased below the normal or ‘euvolemic’ range and that action may be warranted. The inclusion of a lower threshold adds value to the Congestion Index, as it highlights when a patient is at a higher risk of experiencing volume depletion which may lead to hypovolemia. The occurrence of volume depletion leading to hypovolemia is rare. In the FIH and EFS studies, only one event was 48 Attorney Docket No.15653-022WOU2labelled as being “possibly due to hypovolemia”. Furthermore, studies suggest that the rate of hypovolemia-related adverse events could be as low as 1.1%.

[0131] A lower threshold was calculated in the FIH and EFS studies in at least two general ways.In some embodiments, given the low occurrence rate of hypovolemia-related events, to determine a lower threshold, the distribution of the Congestion Index across patients was assessed, and the lower threshold was considered to be a cutoff for the determination of outliers. Only patients who had successfully completed a minimum of 180 home readings were included in the analysis, and a random sample of 180 readings per patient were chosen for the analysis. FIG. 28 shows the distribution of patients’ Congestion Index from taking this random sample of 180 readings for each patient. Because the distribution of the Congestion Index across patients was non-normal and skewed to the left, one method used to determine this lower threshold was:

[18] ^^^^^^^^^^ ^^ℎ^^^^^^ℎ^^^^^^ = ^^2 − 1.5 × ^^^^^^Where Q2 is the second quartile and IQR is the inter-quartile range. Using the methodology described above, all data below 10% were considered outliers in some embodiments. Applying this 10% threshold 340 across all data used for this analysis would give a low alert rate of 1.36%, which aligns closely with the expected hypovolemia-related events rate of 1.1% described above.

[0132] In other embodiments, the lower threshold was set at a value which would result in aspecified proportion of low alerts, e.g., in 1% or 1.1% of cases, based on the distribution of the underlying Congestion Index data. For example, in the FIH and EFS studies, a lower threshold of 15% corresponded to the first percentile of data, and applying the 15% lower threshold resulted in a low alert rate of 1.06%, which aligns closely with the expected hypovolemia-related events rate of 1.1% described above.

[0133] Right Atrial Pressure

[0134] RAP is a standard invasive method for evaluation of congestion status. As set forth inmore detail below, the Congestion Index can classify patients according to RAP values at the time of implant of sensor 12. The Congestion Index was compared against RAP data obtained at implant for a sample of 50 patients. Comparison of the RAP value at implant and Congestion Index at the nearest time point to implant found a Pearson’s correlation coefficient of 0.46 (p <0.001). Additionally, a Wilcoxon rank-sum test found statistically significant differences in Congestion Index between patients who had high (RAP≥10mmHg) and not-high (RAP<10 mmHg, 705.00 (p=0.022)) RAP 49 Attorney Docket No.15653-022WOU2values at implant, as well as between patients who had high (RAP≥10mmHg) and very high (RAP≥15mmHg, 306.00, p=0.000) RAP values at implant. FIG.29 illustrates the association between RAP values at implant and the Congestion Index at implant. There was a 76% correlation in very- high range (RAP≥15mmHg) 342. There was a 51% correlation outside of very-high range 342. The area outside of very-high range 342 comprises very-low range (RAP≤3mmHg) 344 and median range (RAP 3-15 mmHg) 346.

[0135] FIG. 30A shows a boxplot of Congestion Index between high (RAP≥10mmHg) and not-high (RAP<10 mmHg) groups. FIG. 30B shows a boxplot of Congestion Index between very-high (RAP≥15mmHg) and not-high (RAP<10 mmHg) groups. These data suggest an association of the Congestion Index with RAP.

[0136] In FIG. 31, an ROC curve 348 was used to assess the ability of the Congestion Index toclassify high (RAP>10mmHg) / not-high (RAP<10 mmHg) RAP, and ROC curve 350 was used to assess the ability of the Congestion Index to classify very-high (RAP>15mmHg) / not very-high (RAP<15 mmHg) RAP. An ROC area of 0.64 was found for classifying between high and not-high RAP values, using a threshold of 10mmHg to denote high RAP. An ROC area of 0.89 was found in discriminating very-high RAP, using a threshold of 15mmHg to denote very-high RAP. These data further support the association between the Congestion Index wand RAP.

[0137] Results for classification of high / very-high and low RAP using the Congestion Index arereported in Table 4 below. 60% and 70% Congestion Index thresholds are examined. These data further support using a threshold of 70%. Table 4 Metric RAP>10mmHg RAP>15mmHg RAP>10mmHg RAP>15mmHg (60% (60% (70% (70% threshold) threshold) threshold) threshold) Accuracy [%] 58.0 50.0 64.0 72.0Sensitivity [%] 73.9 100.0 56.5 100.0 Specificity [%] 44.4 41.9 70.4 67.4 Positive 53.1 21.9 61.9 33.3 Predictive Value [%] Negative 66.7 100.0 65.5 100.0 Predictive Value [%] Area Under 0.64 0.89 0.64 0.89 Curve 50 Attorney Docket No.15653-022WOU2

[0138] NT-proBNP

[0139] NT-proBNP is a well-validated biomarker in the assessment and management ofcongestive heart failure. As set forth in more detail below, there is a strong statistical association between NT-proBNP and the Congestion Index across multiple timepoints, showing that the Congestion Index can provide clinically useful information.

[0140] NT-proBNP was captured for each patient at each clinic visit. The relationship of variationin the Congestion Index to variation in NT-proBNP was then examined. FIG.32 shows variation of Congestion Index with log valued NT-proBNP across clinic visits, including 1-month visits 352 (only one labeled to avoid clutter), 3-month visits 354 (only one labeled to avoid clutter), 6-month visits 356 (only one labeled to avoid clutter), and a baseline 358 (only one labeled to avoid clutter). FIG. 32A presents the same data as in FIG.32 in bar chart form with the data bucketed.

[0141] High NT-proBNP is defined as greater than 1000 ng / L, while reduced NT-proBNP isdefined as greater than 30% reduction from baseline per follow-up visit. A Linear Mixed Effects (LME) regression model showed that there is a significant association of Congestion Index with NT- proBNP (p<0.005), accounting for different timepoints with patient ID included as a random effect (see FIG.32).

[0142] 51 of 61 patients had NTproBNP levels above 1000ng / L at implantation / baseline 358. Atthe 6-month follow-up visit 356, 16 of 42 (38%) patients had reduced NTproBNP levels below 1000ng / L. Four of 42 patients showed an increase greater than 30% in NTproBNP levels; 18 of 42 showed a decrease greater 30%; and 13 of 42 showed changes less than 30%.

[0143] Comparison of high and low NT-proBNP with mean values of the Congestion Index foundthat the mean Congestion Index was significantly (p<0.05) associated with NT-proBNP label (i.e., the Congestion Index was higher in patients with high NT-proBNP compared to patients with low NT- proBNP) across time points. This is shown in FIG.33A—at each of the baseline, 1-month visit, 3- month visit, and 6-month visit, the mean Congestion Index was greater in patients with high NT- proBNP 360 compared to patients with low NT-proBNP 362. FIG.33B shows mean Congestion Index for patients with stable NT-proBNP 364 and reduced NT-proBNP 366 across the 1-month visit, 3- month visit, and 6-month visit. 51 Attorney Docket No.15653-022WOU2

[0144] FIG.34 shows variation in Congestion Index with NT-proBNP, and in particular with highNT-proBNP (greater than or equal to 1000 ng / L) 368 and low NT-proBNP (less than 1000 ng / L) 370. Significant association was observed between Congestion Index and NT-proBNP label (high / low).

[0145] In order to determine how well the Congestion Index can classify high and low NT-proBNP, an ROC analysis was conducted and is shown in FIG.35. The relationship was considered separately at each clinic-visit time point (baseline 372, 1 month 374, 3 months 376, and 6 months 378) given that NT-proBNP levels were evolving over the course of the trials as a result of medical interventions. Classification results for high and low NT-proBNP using the Congestion Index at each time point are tabulated in Table 5 and suggest a range in Area Under Curve (AUC) of 0.64-0.73. A 60% Congestion Index threshold was used. Table 5 Metric Baseline Month Month Month (N=61) 1 3 6 (N=61) (N=59) (N=52) Accuracy [%] 57.38 65.57 64.41 63.46Sensitivity [%] 54.90 63.64 62.50 55.88 Specificity [%] 70.00 70.59 68.42 77.78 Positive 90.32 84.85 80.65 82.61 Predictive Value [%] Negative 23.33 42.86 46.43 48.28 Predictive Value [%] Area Under 0.67 0.73 0.64 0.69 Curve

[0146] New York Heart Association Class

[0147] The New York Heart Association (NYHA) heart failure classification is a standard clinicalassessment of heart failure disease status. As with NT-proBNP, NYHA class was assessed for each patient at each clinic visit and the relationship between the Congestion Index and NYHA class was then examined. As set forth in more detail below, trends were observed between the Congestion Index and NYHA class.

[0148] Analysis of the Congestion Index with NYHA class suggests that there is a degree ofconcordance between the Congestion Index and NYHA class at baseline, with the mean value of the Congestion Index increasing with increased NYHA class. Twelve patients showed an improvement 52 Attorney Docket No.15653-022WOU2in NYHA class between baseline and their 6-month follow-up, while two patients deteriorated in NYHA class. The remaining 28 patients were stable in their NYHA class after 6 months. FIG. 36 shows the mean Congestion Index for Class 1 patients 380, Class 2 patients 382, and Class 3 patients 384 at each clinic-visit time point (baseline, 1 month, 3 months, and 6 months).

[0149] FIG.37 examines the change in NYHA class from baseline to 6 months and its associationwith Congestion Index. Patients whose NYHA class deteriorated (e.g., from Class II to Class III) were not found to have a significant increase (p>0.05) in Congestion Index compared to patients who were stable or improved in Congestion Index. However, a clear trend can be observed between deteriorating NYHA class and increased Congestion Index.

[0150] Medication Change

[0151] How the Congestion Index responds to medication, particularly oral and IV diuretics, wasalso examined. As set forth in more detail below, significant changes were observed in the trends of the Congestion Index within 10-20 days of a medication change.

[0152] After obtaining a dataset of medication-related data (including when oral and IV diureticswere taken) and sensor-related data (including daily Congestion Indexes), data was refined in pre- processing. Pre-processing included filtering the data, signal smoothing, and choosing metrics calculated from the collected signals for use in determining if the medication changes prompted a response in any element of the signals. For filtering, only the first sensor reading and resulting Congestion Index were retained each day to minimize noise. For signal smoothing, a moving average filter was applied to smooth the sensor readings, reducing daily variations. Medication events were then identified based on significant changes in diuretic dosages. For each detected medication event, the dataset was segmented into a period of 20 days before the event and a period of 20 days after the event to form a 40-day period. Events with data missing from over 60% of the 40 days, events where the dosage at the end of the 40-day period was lower than at the start of the 40-day period, and events clashing with other events (such as HFEs) were excluded. Remaining medication events were then visualized in ensemble plots showing the average Congestion Index for each day relative to the medication event along with confidence intervals, which provided a clear view of how the Congestion Index behaves before and after medication changes.

[0153] For example, in the ensemble plot shown in FIG. 38, the goal was to investigate how theCongestion Index behaves relative to an increase of oral diuretics, and specifically how the Congestion 53 Attorney Docket No.15653-022WOU2Index changes as the increase of oral diuretics approaches and how it behaves after the increase of oral diuretics. The ensemble plot displays the average trajectory of the Congestion Index across all increases of oral diuretics (events), with the x-axis representing time (days) relative to the increases of oral diuretics (events), and the y-axis showing the change in the Congestion Index. The event onset 392 (increase in oral diuretics) occurs at day 0, dividing the graph into 20-day pre-event period 394 and 20-day post-event period 396. The expected response is a gradual increase towards event 392 in pre-event period 394 and a reduction afterwards in post-event period 396. In FIG.38, the average of 54 medication events from 36 different patients and a 95% confidence interval are captured.

[0154] Then, slope analysis was performed for pre-event period 394 and post-event period 396 toquantify the changes. Slopes of the Congestion Index were computed for each day within the 40-day period to help identify changes in the direction or trend of the Congestion Index before and after medication adjustments. Slope analysis highlights if there is any significant shift in the Congestion Index behavior surrounding the medication event. These slope values are plotted in FIG.39, including slope values before the event 398 and slope values after the event 400. The statistical significance of these changes was tested using a Wilcoxon rank-sum test, which compared slopes before the event 398 and slopes after the event 400 on a daily basis. For example, the slope one day before the event was compared to the slope one day after the event, the slope two days before the event was compared to the slope two days after the event, etc. The results were visualized using boxplots with p-values to clearly indicate statistical significance. In the secondary axis, the p-value of the rank-sum of slopes has been plotted. Days where p-values are lower than 0.05 threshold 402 show that the medians of the before and after populations 398, 400 are significantly different. Comparing the slopes before and after the medication event shows significant differences between the two samples (before 398 and after 400) within 10-20 days of when the oral diuretics are increased.

[0155] Further alternative embodiments for generation of various metrics within components ofmetrics generator 36 are described below with reference to FIGS. 40 through 47. For example, an additional alternative embodiment for respiration rate module 90 of metrics generator 36 provides for extraction of respiration rate from an area trace as shown in FIG.40. As illustrated therein, respiration rate module 90A receives an area trace as input 404. Baseline wander and higher frequency components are then removed from the input signal using a butterworth bandpass filter module 406. An example range for this filter would be 0.08Hz to 0.75Hz. The signal is then split into many windows with a certain lag module 408—an example would be five hundred twelve (512) windows with a four (4) second lag. The mean is then subtracted from each window module 410. These 54 Attorney Docket No.15653-022WOU2windows from module 410 are then used to estimate the power spectral density. This is done using Welch’s method to account for possible non-stationarity in the signal, i.e., a change in respiration rate module 412. For example, this could use a Hann window with a sample size of two hundred fifty-six (256). The power spectral density from module 412 is then filtered to examine only possible respiration rates in module 414—for example eight (8) to forty (40) beats per minute. All peaks in this filtered spectra are then found in module 416 for example using peak prominence. The largest detected peak is extracted in module 418. The signal quality of this peak is then assessed by comparing it with signal quality outside of the range excluding potential harmonics of this peak. If the signal quality is greater than a threshold in module 420, the breaths per minute as estimated by the max peak is accepted for the window else the respiration rate for the window is rejected at module 422. Module 412, 414, 416, 418, 420, and 422 are repeated across all windows given by 410. Module 424 takes all of these respiration rates in breaths per minute as an input. If the number of respiration rates is greater than a threshold, the mean of the accepted respiration rates is returned as the global respiration rate for the input 404. A final output 426 of this respiration rate module 90A returns an individual value for each of respiration rate (RR), signal quality (SQ), and respiration collapsibility index, derived for each area trace input.

[0156] As yet another metrics example, each beat of the heart also perturbs the IVC and thishigher frequency signal can also be extracted from the IVC area trace by spectrally decomposing the area trace using Fast Fourier Transformation (FFT) as shown in FIG.40A. It is well known that heart rate is also a useful metric in the management of patients and this could be included in the generated metrics. Increased or decreased rate or increased rate variability can be predictors of worsening outcomes. Also, pacing related metrics may be determined from the area trace.

[0157] In another example, as illustrated in FIG. 41, assuming an idealized ‘n’ shape curve forIVC area vs. IVC collapse, it may be possible to estimate the ovality of the IVC (hence provide an approximation of the IVC diameter should the existing diameter-based clinical guidelines be used for thresholds). Also, it may be assumed that the IVC shape varies from flattened ellipse to fully circular along that ‘n’ curve, with axis ratio correspondingly varying from low (~ 0.1) to equal to one (1) along the curve. Thus, knowledge of both collapse and area allow approximate determination of location on ‘n’ shape, which in turn allows estimation of the IVC’s axis ratio. The approximated axis ratio can in turn be plugged into the area equation for an ellipse, reducing the number of unknown variables to one thereby permitting an exact solution for IVC major and minor diameters (for comparison with existing clinical guidelines). 55 Attorney Docket No.15653-022WOU2

[0158] In another embodiment, the area change seen in the IVC is used to estimate venous return(VR) / cardiac output (CO) by measuring the change of area per time as the integral under the curve (AUC) and a simple initial calibration to a clinical cardiac output measurement. This measurement is based on volume passing through the IVC causing temporal distension of the IVC relative to its minimum size. FIG.42 illustrates this distortion in an idealized cross-sectional depiction of the IVC with (b) and (d), which corresponds to the same features in the area trace (FIG.2), identifying circle and area shifts. In this example, cardiac output can be determined using the following equation

[0019] :

[0019] CO = VR = corfacresp(Rcol*RR) + corfaccard(Hcol * HR) Correction factors (corfacresp and corfaccard) are derived from calibration step reference cardiac output as measured (e.g., ultrasound to estimate cardiac output). The correction factors can be split into elements contributing to respiration and to cardiac function as the distance of the device to the heart can affect the cardiac component detected relative to the respiration modulation. For example, two-thirds of blood flow into the right atrium originates from the IVC. As blood flows through the IVC, it expands being a compliant vasculature. Similarly, to stroke volume estimates from images of the left atrium, the change in the IVC cross-sectional area is correlated with cardiac output as has been reported in 2004 by Barberi et al. “Respiratory changes in inferior vena cava diameter are helpful in predicting fluid responsiveness in ventilated septic patients” [Intensive Care Med (2004) 30:1740– 1746, DOI 10.1007 / s00134-004-2259-8] (incorporated herein by reference in its entirety). The IVC modulates with respiratory effort and cardiac activity.

[0159] Linkage between IVC area change and cardiac output involves four main features—areachange with respiration and cardiac activity as well as with heart rate and respiration rate. Due to the complex nature of venous return / cardiac output—it is a time integral—it is necessary to calibrate the area time integral measured in m2in a way that it relates to cardiac output measured in liters / min. To do that at least one synchronized cardiac output measurement is required in conjunction with a recording of free respiration IVC sensor area. Assuming 0 m2and 0 liter / min is the second calibration point, a simple linear calibration is possible, ultimately leading to a conversion factor f in units of liters / min / m2so that:

[0020] CO= fcorr * AreaTimeIntegral(Sensor)

[0160] In a different embodiment, multiple calibration points are created through phlebotomy / withdrawal of blood and re-infusion or through saline injection of a known bolus volume. The cardiac 56 Attorney Docket No.15653-022WOU2output accuracy is expected to improve when respiration modulation of area is related in a weighted manner to cardiac modulation of area. Furthermore, tricuspid regurge information can be incorporated to correct for regurge-related area modulations. Correction factor weighting may be included to adjust contribution from cardiac and respiration with seen collapse. This approach requires more than one single reference point for calibration. Cardiac output measurements during provocative maneuver such as, for instance, an inspiration or expiration breath-hold may allow separating the respiratory component from its cardiac component—as during breath-hold the cardiac component is exclusively visible.

[0161] As further shown in FIG. 43, measured area trace and filtered respiration and cardiaccomponent traces can be used to compute the AUC as a metric for cardiac output. The change of area in the vessel is correlated to the volume delivered to the right atrium—that is generally known as venous return. Venous return is expected to equal cardiac output. Venous return is composed of flow from superior and inferior vena cavae. As this sensor is intended in one of these locations, an initial calibration step is required. Common clinical measurement methods, such as ultrasound or thermodilution, and others can provide a reference cardiac output level. In some embodiments, a single cardiac output measurement could be used, or alternatively, a second cardiac output and sensor reading could be obtained via the use of a maneuver or other method. With such a reference and the calculated AUC, the correction coefficient to link both is established. The correction coefficient can be further refined by adding information about the respiration and cardiac contributions to flow in the respective sensor location, for instance, by splitting the single correction factor by a pre-determined ratio applied to cardiac and respiration.

[0162] In another embodiment, as shown in FIG. 44, a cardiac output estimation process 428 maybe used to estimate cardiac output directly from sensor data. Inputs 430 to process 428 include extracted features related to the area under the curve such as cardiac collapse (Hcol) and respiration collapse (Rcol), as well as heart (HR) and respiration (RR) rate. With these inputs, a cardiac component of output is set at 432 as the product of heart rate, cardiac collapse, and sensor length. Similarly, the respiration component of output is set at 434 as the product of heart rate, respiration collapse, and sensor length. Sensor length in this process may refer to the longitudinal length of the sensor coil portion of a wireless resonant circuit sensor implant that is configured and dimensioned to be implanted in a patient blood vessel in contact with the vessel wall. One example of such a sensor implant comprises an expandable and collapsible variable inductance coil with a plurality of adjacent wire strands formed around an open center to allow substantially unimpeded blood flow through the 57 Attorney Docket No.15653-022WOU2open center. In some such embodiments, the coil is configured and dimensioned (i) to extend around an inner periphery of the vessel when implanted therein and (ii) to move with the vessel wall in response to expansion and collapse of the vessel. The sensor implant also comprises a capacitance, which together with the variable inductance coil, forms a variable inductance resonant circuit having a variable characteristic frequency correlated to the diameter or area of the expandable and collapsible variable inductance coil. In further alternatives, the expandable and collapsible variable inductance coil comprises plural wire strands, and the inductance changes based on changes in cross-sectional area or diameter across the coil open center in response to expansion and collapse of said coil. Such sensor implants are described in more detail, for example, in incorporated-by-reference USP 10,806,352 (for example, RC-WVM implant 12 and height (A) as described therein corresponding to sensor length). It is to be noted that steps 432 and 434 may occur in any order or simultaneously. Total cardiac output is then set at 436 as the sum of the two component outputs multiplied by a correction factor. Persons of ordinary skill in the art may derive the correction factor based on the teachings of the present application using an example of measured cardiac output and correlated respiration and cardiac components.

[0163] As further illustrated in FIGS.45(a) through 45(c), the parameters cardiac collapse (Hcol),respiration collapse (Rcol), heart rate (HR) and respiration rate (RR) can be used to calculate the respiration and cardiac traces synthetically (simulated area modulation) to again use the AUC method as another alternative means for assessing cardiac output. In this embodiment, the collapse component is used to inform the magnitude of the oscillation and the rate informs the oscillatory frequency. Thus, the respiration trace (b) in FIG.45 can be simulated as:

[0021] RespirationTrace=Rcol*sin(2*pi*RR*TimeSamplesVector) where the time samples are a linear range of values starting from a starting value to a final time stamp in discrete steps defined by the sampling frequency. In analogue, the cardiac trace (c) can be simulated as:

[0022] CardiacTrace=Hcol*sin(2*pi*HR*TimeSamplesVector)

[0164] In one additional alternative embodiment, RAP can be estimated by correlation to areatrace data. For example, paired data of IVC area trace and RAP as measured by pressure catheter, can be used to create a regression model using training data at an individual or population level which can then be used to predict RAP based on an input of the area metrics produced from the sensor area trace, 58 Attorney Docket No.15653-022WOU2e.g., area mean. Vascular tone, which refers to the degree of constriction experienced by a blood vessel relative to its maximally dilated state, also may be estimated based on area trace data. For example, a change in an extracted feature over time, such as Amax achieved with a maneuver varying between days, may be indicative of changes in IVC tone. Thus, correlation to area trace data using a training model can provide IVC tone estimates.

[0165] In another embodiment, mean arterial pressure (MAP) can be determined using an externalblood pressure cuff and can be combined with the aforementioned cardiac output and RAP (directly measured or estimated as described above) to estimate SVR using the following equation

[0023] :

[0023] SVR=(MAP-RAP) / CO.

[0166] An example of this calculation based on human sensor data is provided in FIGS. 46Athrough 46D showing trend data as a function of days for a heart failure patient experiencing symptoms (dyspnea). Traces of SVR (FIG.46A), area mean from the sensor (FIG.46B), respiration rate from sensor (FIG. 46C), and daily dose of diuretic (furosemide) (FIG. 46D) are shown. Towards the symptom exacerbation (dyspnea / breathlessness) in mid-March, the area increased logarithmically with elevated respiration rate. SVR increased drastically in an exponential shape towards the end of dyspnea. With increase in diuretic dose, area reduced drastically within days and SVR recovered initially to be followed by a brief elevation for several days until medication was ideally titrated to facilitate reduction of patient’s respiration rate to normal values.

[0167] Systems and methods herein described may be implemented using one or more computingdevices and, except where otherwise indicated, may include standard computing components of processors, memory, storage, and communications busses, as well as high resolution graphics rendering hardware and software where required. Such components may be configured and programmed by persons of ordinary skill in the art based on the teachings of the present disclosure. Software programing implementing systems and methods described herein may reside on a non- transient computer readable media as a computer program product. Computing devices in general also include cloud-based computing embodiments and computing systems with elements distributed across networks.

[0168] In various alternative embodiments, signal processing, data extraction, diagnostic, andtreatment-related functions, such as those of trace generator 30, feature detector 34, metrics generator 36, metrics generator 36A, boundary generator 38, boundary generator 38A, index generator 40, 59 Attorney Docket No.15653-022WOU2decision logic 42, alert generator 44, or interface devices 46, including processor 196, as non-limiting examples, may be executed as one or more computing devices, or may be collectively executed in a single or plural computing device. FIG. 47 illustrates one example of such a computing device, wherein computing device 438 includes one or more processors 440, memory 442, storage device 444, high-speed interface 446 connecting to memory 442 and high-speed expansion ports 448, and a low- speed interface 450 connecting to low-speed bus 454 and storage device 444. Each of the components 440, 442, 444, 446, 448, and 450, are interconnected using various buses or other suitable connections as indicated in FIG.47 by arrows connecting components. The processor 440 can process instructions for execution within the computing device 438, including instructions stored in the memory 442 or on the storage device 444 to display graphical information via Graphical User Interface (GUI) 456 with display 458, or on an external user interface device, coupled to high speed interface 446. As will be appreciated by persons of ordinary skill in the art, certain described functions may not require an independent GUI or display. In other implementations, multiple processors and / or multiple buses may be used, as appropriate, along with multiple memories and types of memory. Also, multiple computing devices 438 may be connected, with each device providing portions of the necessary operations (e.g., as a server bank, a group of blade servers, or a multi-processor system).

[0169] Memory 442 stores information within the computing device 438. In one implementation,the memory 442 is a computer-readable medium. In one implementation, the memory 442 is a volatile memory unit or units. In another implementation, the memory 442 is a non-volatile memory unit or units.

[0170] Storage device 444 is capable of providing mass storage for computing device 438, andmay contain information such as timing control, time slice size and / or static color chroma and timing as described herein above. In one implementation, storage device 444 is a computer-readable medium. In various different implementations, storage device 444 may be a floppy disk device, a hard disk device, an optical disk device, or a tape device, a flash memory, or other similar solid-state memory device, or an array of devices, including devices in a storage area network or other configurations. In one implementation, a computer program product is tangibly embodied in an information carrier. The computer program product contains instructions that, when executed, perform one or more methods, such as those described above. The information carrier is a computer- or machine-readable medium, such as the memory 442, the storage device 444, or memory on processor 440. 60 Attorney Docket No.15653-022WOU2

[0171] High-speed interface 446 manages bandwidth-intensive operations for computing device438, while low-speed interface 450 manages lower bandwidth-intensive operations. Such allocation of duties is exemplary only. In one implementation, high-speed interface 446 is coupled to memory 442, display 458 (e.g., through a graphics processor or accelerator), and to high-speed expansion ports 448, which may accept various expansion cards (not shown). In the implementation, low-speed interface 450 is coupled to storage device 444 and low-speed bus 454. The low-speed bus, which may include various communication ports (e.g., USB, Bluetooth, Ethernet, wireless Ethernet) may be coupled to one or more input / output devices as part of GUI 456 or as a further external user interface, such as a keyboard, a pointing device, a scanner, or a networking device such as a switch or router, e.g., through a network adapter.

[0172] Various implementations of the systems and techniques described here can be realized indigital electronic circuitry, integrated circuitry, specially designed ASICs (application specific integrated circuits), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0173] These computer programs (also known as programs, software, software applications, orcode) include machine instructions for a programmable processor, and can be implemented in a high- level procedural and / or object-oriented programming language, and / or in assembly / machine language. As used herein, the terms “machine-readable medium” and “computer-readable medium” refer to any computer program product, apparatus, and / or device (e.g., magnetic discs, optical disks, memory, Programmable Logic Devices (PLDs)) used to provide machine instructions and / or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal. The term “machine-readable signal” refers to any signal used to provide machine instructions and / or data to a programmable processor.

[0174] To provide for interaction with a user, the systems and techniques described herein can beimplemented on a computer having a display device separate from video display 458. LED displays are now most common, however older display technologies (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) may be used. Other interface devices may include a keyboard and a 61 Attorney Docket No.15653-022WOU2pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feed-back, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0175] The systems and techniques described herein can be implemented in a computing systemthat includes a back-end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a client computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of wired or wireless digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0176] The computing system can include clients and servers. A client and server are generallyremote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. Processing capacities and capabilities described herein also may be implemented as cloud-based or other network-based processing data engines and may also be implemented using a software as a service (SaaS) model.

[0177] Further alternative embodiments of the present disclosure include methods of creatingpatient-specific heart failure diagnostic tools. Such methods may comprise monitoring a patient physiological parameter correlated to patient fluid volume, the physiological parameter having a measurable value; instructing the patient to perform a maneuver; recording a change in value of the monitored physiological parameter in response to the patient performing the maneuver; setting the value of the physiological parameter upon the patient performing the maneuver as a maximum value for the patient; determining a minimum value; and setting the maximum value as an upper limit and the minimum value as a lower limit of a standardized scale correlated to a risk of an adverse heart failure event to establish a customized heart failure risk evaluation tool for the patient. With such methodology, measured values of the monitored physiological parameter are comparable to the standardized scale for determination of a level or type of medical intervention. 62 Attorney Docket No.15653-022WOU2

[0178] In another alternative embodiment, a heart failure diagnostic method comprises steps ofmonitoring a patient physiological parameter correlated to patient fluid volume, the physiological parameter having a measurable value; instructing the patient to perform a maneuver; recording a change in value of the monitored physiological parameter in response to the patient performing the maneuver; setting the value of the physiological parameter upon the patient performing the maneuver as a maximum value for the patient; determining a minimum value; setting the maximum value as an upper limit and the minimum value as a lower limit of a standardized scale correlated to a risk of an adverse heart failure event; comparing measured values of the monitored physiological parameter to the standardized scale; and determining a level or type of medical intervention for the patient based on said comparing to the standardized scale.

[0179] The foregoing methods may further comprise collecting IVC area trace data for a pluralityof patients at a plurality of fluid volume statuses; storing the collected data in a database; analyzing the data to determine curve between max and min values for many different max values; and creating the standardized scale based on the determined curve. In other embodiments, the patient physiological parameter correlated to patient fluid volume comprises an IVC dimension; and the monitoring comprises measuring the IVC dimension. The monitoring may further comprise measuring the IVC dimension with an implanted wireless sensor. In another alternative, the monitoring may comprise measuring the IVC dimension and external ultrasound imaging device. The systems and methods described herein provide unique data and novel data interpretation tools to help clinicians manage heart failure, reduce hospitalizations, better manage patient treatments, and improve thus patient outcomes as compared to traditional heart failure management programs.

[0180] The foregoing has been a detailed description of illustrative embodiments of thedisclosure. It is noted that in the present specification and claims appended hereto, conjunctive language such as is used in the phrases “at least one of X, Y and Z” and “one or more of X, Y, and Z,” unless specifically stated or indicated otherwise, shall be taken to mean that each item in the conjunctive list can be present in any number exclusive of every other item in the list or in any number in combination with any or all other item(s) in the conjunctive list, each of which may also be present in any number. Applying this general rule, the conjunctive phrases in the foregoing examples in which the conjunctive list consists of X, Y, and Z shall each encompass: one or more of X; one or more of Y; one or more of Z; one or more of X and one or more of Y; one or more of Y and one or more of Z; one or more of X and one or more of Z; and one or more of X, one or more of Y and one or more of Z. 63 Attorney Docket No.15653-022WOU2

[0181] Various modifications and additions can be made without departing from the spirit andscope of this disclosure. Features of each of the various embodiments described above may be combined with features of other described embodiments as appropriate in order to provide a multiplicity of feature combinations in associated new embodiments. Furthermore, while the foregoing describes a number of separate embodiments, what has been described herein is merely illustrative of the application of the principles of the present disclosure. Additionally, although particular methods herein may be illustrated and / or described as being performed in a specific order, the ordering is highly variable within ordinary skill to achieve aspects of the present disclosure. Accordingly, this description is meant to be taken only by way of example, and not to otherwise limit the scope of this disclosure.

[0182] Exemplary embodiments have been disclosed above and illustrated in the accompanyingdrawings. It will be understood by those skilled in the art that various changes, omissions and additions may be made to that which is specifically disclosed herein without departing from the spirit and scope of the present disclosure. 64 Attorney Docket No.15653-022WOU2

Claims

What is claimed is:

1. A method for determining a heart-failure congestion index, comprising:receiving a first vessel area trace containing at least two characteristic features representing vessel area parameters for a patient; identifying the characteristic features representing at least a first maximum area (Amax), a minimum area (Amin), and a second maximum area; setting an upper area boundary (UB) as the second maximum area; setting a lower area boundary (LB) as an initial predetermined value; generating a normalized IVC area (^^^^^^^^^^) based on the upper area boundary (UB), the lower area boundary (LB), and the minimum area (Amin); generating an adjusted collapsibility index (^^^^^^^^^^) based on the first maximum area (Amax), the minimum area (Amin), and the lower (LB); andgenerating a normalized congestion index for the patient based on the normalized IVC area (^^^^^^^^^^) and the adjusted collapsibility index (^^^^^^^^^^).

2. The method of claim 1, further comprising generating an alert for patient attention and review ofrisk of low IVC area if the normalized congestion index is below a predetermined lower threshold.

3. The method of claim 2, wherein the alert for patient attention and review of risk of low IVC areacomprises a recommended medication change comprising a down-titration of a heart-failure medication.

4. The method of claim 1, further comprising generating an alert for patient attention and review ofrisk of hypervolemia if the normalized congestion index is above a predetermined upper threshold.

5. The method of claim 4, wherein the alert for patient attention and review of risk of hypervolemiacomprises a recommended medication change comprising an up-titration of a heart-failure medication.

6. The method of claim 1, further comprising generating a patient notification of patient fluid status ina normal range if the normalized congestion index is between a predetermined lower threshold and a predetermined upper threshold.

7. The method of claim 2 or claim 3, further comprising generating an alert for physician attention andreview of risk of low IVC area if the normalized congestion index is below the predetermined lower threshold for a predetermined number of days in a predetermined time window. 65 Attorney Docket No.15653-022WOU28. The method of claim 4 or claim 5, further comprising generating an alert for physician attention andreview of risk of hypervolemia if the normalized congestion index is above the predetermined upper threshold for a predetermined number of days in a predetermined time window.

9. The method of claims 2, 3, 6, or 7, wherein the predetermined lower threshold is about 15.

10. The method of claims 4, 5, 6, or 8, wherein the predetermined upper threshold is about 70.

11. The method of any of claims 7-10, wherein the predetermined number of days is five days and thepredetermined time window is seven days.

12. The method of any preceding claim, wherein said receiving a first vessel area trace compriseswirelessly receiving the first vessel area trace from a sensor implanted in a patient’s inferior vena cava.

13. The method of any preceding claim, wherein the initial predetermined value is about 150mm2.

14. The method of any preceding claim, further comprising:receiving at least a second vessel area trace for the patient; resetting the upper area boundary (UB) as a largest maximum area of the first vessel area trace, the second vessel area trace, and any subsequent vessel area trace; and resetting the lower area boundary (LB) as a smallest measured area of the first vessel area trace, the second vessel area trace, and any subsequent vessel area trace if the smallest measured area falls below the initial predetermined value.

15. The method of any preceding claim, wherein generating the normalized IVC area (^^^^^^^^^^)comprises setting the normalized IVC area (^^^^^^^^^^) as equal to 100 * (Amin- LB) ∕ (UB - LB).

16. The method of any preceding claim, wherein generating the adjusted collapsibility index (^^^^^^^^^^)comprises: setting a measured collapsibility index (^^^^^^) as equal to 100 * (Amax- Amin) ∕ (Amax- LB); generating an estimated collapsibility index (^^^^ ) by applying a mixed-effects regression model to a patient population containing plural vessel area traces for each patient of the patient population; and setting the adjusted collapsibility index (^^^^^^^^^^) as equal to a mean of the measured collapsibility index (^^^^^^) and the estimated collapsibility index (^^^^ ). 66 Attorney Docket No.15653-022WOU217. The method of claim 16, wherein generating the normalized congestion index comprises settingthe normalized congestion index as equal to^^^^^^^^^^+ (100−^^^^^^^^^^)2 .

18. The method of any of claims 1-14, wherein generating the normalized IVC area (^^^^^^^^^^)comprises: setting a measured normalized IVC area (^^^^^^^^^^−^^) as equal to 100 * (Amax- Amin) ∕ (Amax- LB); generating an estimated normalized IVC area (^^^^^^^^^^) by applying a mixed-effects regression model to a patient population containing plural vessel area traces for each patient of the patient population; and setting the normalized IVC area (^^^^^^^^^^) as equal to a mean of the measured normalized IVC area (^^^^^^^^^^−^^) and the estimated normalized IVC area (^^^^^^^^^^).

19. The method of claim 18, wherein generating the adjusted collapsibility index (^^^^^^^^^^) comprisessetting the adjusted collapsibility index (^^^^^^^^^^) as equal to 100 * (Amax- Amin) ∕ (Amax- LB).

20. The method of claim 19, wherein generating the normalized congestion index comprises setting(100−^^^^ )+(^^ ) the normalized congestion index as equal to^^^^^^ ^^^^^^^^2 .

21. The method of any preceding claim, wherein said identifying the characteristic features comprises:identifying a first supine portion and a first breath-hold portion of the first vessel trace; setting the first maximum area (Amax) as a maximum area determined from the first supine portion; setting the minimum area (Amin) as a minimum area determined from the first supine portion; and setting the second maximum area as a maximum area determined from the first breath-hold portion.

22. The method of any preceding claim, wherein said method is a computer-based method with saidsteps executed in one or more processing devices, and wherein generating the normalized congestion index for the patient occurs within about one second to about one minute of receiving the first vessel area trace.

23. A heart failure diagnostic system, comprising:a trace feature detector configured to identify characteristic features within wirelessly received periodic vessel area traces representing changes in fluid state of a patient over time, wherein the identified characteristic features include features representative of maximum and minimum vessel areas; 67 Attorney Docket No.15653-022WOU2a metrics generator configured to generate heart function-related parameters for each area trace based on said identified characteristic features, the heart function related parameters including a maximum IVC area corresponding to a first identified maximum vessel area (Amax), and a minimum IVC area corresponding to an identified minimum vessel area (Amin); a boundary generator configured to generate a lower area boundary (LB) for the patient and an upper area boundary (UB) for the patient, wherein the lower area boundary (LB) is set to aninitial predetermined value and the upper area boundary (UB) is set to an identified absolute maximum vessel area for the patient, wherein the metrics generator further comprises a normalized IVC area module configured to set a normalized IVC area (^^^^^^^^^^) based on the upper area boundary (UB), the lower area boundary (LB), and the identified minimum vessel area (Amin), and wherein the metrics generator further comprises an adjusted collapsibility index (^^^^^^^^^^) module configured to set an adjusted collapsibility index (^^^^^^^^^^) based on the lower area boundary (LB), the first identified maximum vessel area (Amax), and the identified minimum vessel area (Amin); and an index generator configured to set a normalized congestion index for the patient based on the normalized IVC area (^^^^^^^^^^) and the adjusted collapsibility index (^^^^^^^^^^).

24. The system of claim 23, further comprising an alert generator configured to generate alerts for thepatient based at least on the normalized congestion index from the index generator, wherein the alert generator: generates an alert for patient attention and review of risk of low IVC area if the normalized congestion index is below a predetermined lower threshold; generates a patient notification of patient fluid status in a normal range if the normalized congestion index is between the predetermined lower threshold and a predetermined upper threshold; and generates an alert for patient attention and review of risk of hypervolemia if the normalized congestion index is above the predetermined upper threshold.

25. The system of claim 24, further comprising a decision logic configured to determine a medicationchange for the patient based at least on the normalized congestion index from the index generator, wherein: the medication change comprises a down-titration of a heart-failure medication if the normalized congestion index is below the predetermined lower threshold; and 68 Attorney Docket No.15653-022WOU2the medication change comprises an up-titration of the heart-failure medication if the normalized congestion index is above the predetermined upper threshold.

26. The system of claim 24 or claim 25, wherein the alert generator is configured to:generate an alert for physician attention and review of risk of low IVC area if the normalized congestion index is below the predetermined lower threshold for a predetermined number of days in a predetermined time window; and generate an alert for physician attention and review of risk of hypervolemia if the normalized congestion index is above the predetermined upper threshold for the predetermined number of days in the predetermined time window.

27. The system of any of claims 24-26, wherein the predetermined lower threshold is about 15 and thepredetermined upper threshold is about 70.

28. The system of any of claims 24-27, wherein the predetermined number of days is five days andthe predetermined time window is seven days.

29. The system of any of claims 23-28, wherein the initial predetermined value is about 150mm2.

30. The system of any of claims 23-29, wherein the boundary generator resets the lower area boundary(LB) to a smallest measured area out of all area traces for the patient if the smallest measured area falls below the initial predetermined value.

31. The system of any of claims 23-30, wherein the normalized IVC area module of the metricsgenerator sets the normalized IVC area (^^^^^^^^^^) as equal to 100 * (Amin - LB) ∕ (UB -LB).

32. The system of any of claims 23-31, wherein the adjusted collapsibility index (^^^^^^^^^^) module ofthe metrics generator sets a measured collapsibility index (^^^^^^) as equal to 100 * (Amax - Amin) ∕ (Amax - LB), is configured to generate an estimated collapsibility index (^^^^ ) by applying a mixed- effects regression model to a patient population containing plural vessel area traces for each patient of the patient population, and sets the adjusted collapsibility index (^^^^^^^^^^) as equal to a mean of the measured collapsibility index (^^^^^^) and the estimated collapsibility index (^^^^ ).

33. The system of any of claims 23-32, wherein the index generator sets the normalized congestionindex as equal^^^^^^^^^^+ (100−^^^^^^^^^^).

34. The system ofwherein the normalized IVC area module of the metricsgenerator sets a measured normalized IVC area (^^^^^^^^^^−^^) as equal to 100 * (Amax- Amin) ∕ (Amax69 Attorney Docket No.15653-022WOU2- LB), is configured to generate an estimated normalized IVC area (^^^^^^^^^^) by applying a mixed- effects regression model to a patient population containing plural vessel area traces for each patient of the patient population, and sets the normalized IVC area (^^^^^^^^^^) as equal to a mean of the measured normalized IVC area (^^^^^^^^^^−^^) and the estimated normalized IVC area (^^^^^^^^^^).

35. The system of claim 34, wherein the adjusted collapsibility index (^^^^^^^^^^) module of the metricsgenerator sets the adjusted collapsibility index (^^^^^^^^^^) as equal to 100 * (Amax - Amin) ∕ (Amax - LB).

36. The system of claim 35, wherein the index generator sets the normalized congestion index as equalto^^^^^^^^^^+ (100−^^^^^^^^^^)2 .

37. The system of any of claims 23-36, wherein:the identified characteristic features further include maneuver types; the features representative of maximum and minimum vessel areas are associated with identified maneuver types; the first identified maximum vessel area (Amax) is a maximum area associated with a first maneuver for each area trace; the identified minimum vessel area (Amin) is a minimum area associated with the first maneuver for each area trace; the identified absolute maximum vessel area for the patient is a maximum area associated with a second maneuver out of all area traces for the patient.

38. The system of claim 37, wherein the first maneuver comprises a supine position and the secondmaneuver comprises a breath-hold.

39. The system of any of claims 23-38, further comprising non-transitory computer storage containinginstructions executable by a processor to cause the trace feature detector to identify said characteristic features, the metrics generator to generate said heart function-related parameters for each area trace, the boundary generator to generate the lower area boundary (LB) for the patient and the upper area boundary (UB) for the patient, the normalized IVC area module to set the normalized IVC area (^^^^^^^^^^), the adjusted collapsibility index (^^^^^^^^^^) module to set the adjusted collapsibility index (^^^^^^^^^^), and the index generator to set the normalized congestion index for the patient. 70 Attorney Docket No.15653-022WOU20. A heart failure diagnostic system, comprising:a trace feature detector configured to identify characteristic features within wirelessly received periodic vessel area traces representing changes in fluid state of a patient over time, wherein identified characteristic features comprise maneuver types and maximum and minimum areas associated with identified maneuvers; a metrics generator to generate heart function-related parameters for each area trace based on said identified characteristic features, wherein the metrics generator comprises a maximum IVC area module to set a maximum vessel area (Amax) as a maximum area associated with a first maneuver for each area trace, and a minimum IVC area module to set a minimum vessel area (Amin) as a minimum area associated with the first maneuver for each area trace; a boundary generator to generate a lower area boundary (LB) for the patient and an upper area boundary (UB) for the patient, wherein the upper area boundary (UB) is set to a maximumarea associated with a second maneuver out of all area traces for the patient, wherein the metrics generator further comprises a normalized IVC area module to set a normalized IVC area (^^^^^^^^^^) based on the upper area boundary (UB), the lower area boundary (LB), and the minimum vessel area (Amin), and wherein the metrics generator further comprises an adjusted collapsibility index (^^^^^^^^^^) module to set an adjusted collapsibility index (^^^^^^^^^^) based on the lower area boundary (LB), the maximum vessel area (Amax), and the minimum vessel area (Amin); and an index generator to set a normalized congestion index for the patient based on the normalized IVC area (^^^^^^^^^^) and the adjusted collapsibility index (^^^^^^^^^^). 71 Attorney Docket No.15653-022WOU2

Images